Last weekend me and a friend decided to work on a micro bot to push the potential of the in-game units to the limits.
As of today, we've implemented two common scenarios: bum rushing siege tanks with zerglings, and splitting Marines.
Our bot is now capable of : 40 speed banelings vs. 20 marines (no marines lost) 20 spread siege tanks vs. 100 zerglings (all tanks dead, 19 zergling survivors)
To liven things up we made videos with footage using the bot, combined with some decent music and scenes from GSL.
Here's the scenarios:
It would be awesome to get some ideas on what we should implement next We're currently planning on making a little gang of malicious Stalkers.
If you feel you have a scenario that requires a ridiculous amount of micro, please let us know.
Please note that this is not any form of a third party program, just code in the Starcraft II Editor.
Edit, regarding the zerglings splitting before the shot lands:
The script is capable of determining which zergling that is going to take the next blow. When a tank starts aiming (when its target dies for example), the bot predicts it's gonna fire at the closest zergling and executes a split the AI of Starcraft 2 is very predictable in that meaning..
As for the cool "sphere"-effect in the beginning, we made the tanks prefer zerglings with more friends around, but the ai compensates for this too!
First of all we must say that we're really glad that the community appreciated Automaton 2000!
Question: What is the APM of the bot in the different videos? Answer: We just made a script to check the APM, we got 250 and got kinda disappointed... then we realized that we had 250 APS, which is equal to 15000 APM!
So: 15000 APM for the kiting marines, and 8000 APM for the zerglings breaking the tanks.
Is there a risk that this will be abused on ladder? For everything Automaton 2000 can do we have to thank Blizzard for creating a wonderful editor. Having that said, there's obviously no way Automaton can be used for cheating online.
Would it be possible to code this same micro-scripting into a custom AI without needing to prompt the actions with a chat command? Yes, the chat commands were only implemented because it was convenient. And when we use the command, the AI will be applied to all marines/zerglings that spawn. The reason we had the commands was that we wanted to try out some zergling/marine-micro for ourselves! (...and we failed horribly!)
We've already started with the next implementations, and we're planning on having them finished in a couple of days. One of the scenarios will be more "realistic", and the other one just for funzies. Thank you for all the great ideas!
On the Malicious Stalkers - apparently, if you blink perfectly, you can dodge projectile style attacks (making them do 0 damage). I'd like to see 20 or so Blink Stalkers take on a 30-40 stimmed marauder force.
On April 07 2011 18:01 Aequos wrote: On the Malicious Stalkers - apparently, if you blink perfectly, you can dodge projectile style attacks (making them do 0 damage). I'd like to see 20 or so Blink Stalkers take on a 30-40 stimmed marauder force.
haha yes this :D Not sure if possible with those numbers because stim marauders can shoot faster than the blink rechargers (maybe) but would be cool to see in action
Those zerglings know what's coming.... I also watched the baneling video and uhh.....wow...marines counter banes pretty hard.
As for my suggestion, I would have to say 30 workers against 3? well micro'd blue flame hellions in a 8x8 unit square. Obviously the bot is microing the workers.
Gadd & Blid: When a tank starts aiming (when its target dies for example), the bot assumes it's gonna fire at the closest zergling and executes a split
As for the cool "sphere"-effect in the beginning, we made the tanks prefer zerglings with more friends around, but the ai compensates for this too!
On April 07 2011 18:14 Asha` wrote: Really impressive, though after yesterdays thread about micro bots I'm scared about potential abuse on ladder and stuff
This is just code in the editor, it has no connection to third party bots . So no worries!
On April 07 2011 18:14 theMarkovian wrote: I can see the bane vs. Marine one actually work with some insane micro, but the ling vs. tank one is impossible IRL Awesome vids.
Well you could just do a spread before actually running into the siege tanks, that should work I think. Sure the unit mechanic tries to push them into a small ball, but a few well-placed hotkeys could fix that problem relatively well.
* Colossus vs Marine Micro * Colossus vs Blink Stalker micro * High Templar vs Marine Micro * Blink Micro Stalkers vs Hydra and/or Roaches * Stalker ball vs Baneling drops (with ovi speed ofc). Stalkers microing/blinking out of the way of banelings being dropped on them but the win condition is killing all the overlords/banelings. Could even do the same scenario with Roaches attacking them as well, with or without creep.
On April 07 2011 18:14 Asha` wrote: Really impressive, though after yesterdays thread about micro bots I'm scared about potential abuse on ladder and stuff
This is just code in the editor, it has no connection to third party bots . So no worries!
Ah cool, well I shall resume applauding your quite spectacular efforts then =)
Wow that's really awesome. I have an idea of something that would be cool to see:
- as well all know, you can only drop units one at a time from an overlord - however, you can build as many overlords as you'd like because they don't cost food - so for a doom drop, place 1 unit per overlord (maybe hydras would be good?), and drop the whole force at once - you can use pickup/drop micro to either protect hydras that are getting targeted, or to do hit and runs - you can also drop some queens to transfuse injured units/overlords
Some other ideas: - void ray charge juggling - super fast and precise force fields - a few zerglings picking off a planetary fortress by running faster than the turret can turn - using phoenix as a support unit to pick up your units that are targeted by ground only units - mass neural parasite and then using other unit's abilities
blink stalkers vs banshees. Just make the stalkers blink just before the banshee attacks hits and see if u can kill em with no losses. Might also work if the banshees are cloaked. Anyone got any idea if the shots are cloaked as well :D
On April 07 2011 18:53 mumakil wrote: blink stalkers vs banshees. Just make the stalkers blink just before the banshee attacks hits and see if u can kill em with no losses. Might also work if the banshees are cloaked. Anyone got any idea if the shots are cloaked as well :D
I really you stick with the more realistic examples of micro. It's really cool to see what can happen in the future. Like that bane thing. This ling vs tank isn't so realistic, cool but not realistic. I doubt anyone will be that good that they can predict which ling will be hit.
On April 07 2011 19:05 leveller wrote: Of course you should make phoenix vs muta or phoenix vs anything.
You can micro phoenixes pretty damn well against mutalisks without a bot, so i dunno how that would actually turn out to be amazing as the past videos has been.
After watching the marine split micro video, it kind of reminds me of pro WoW raiders doing progression (Utgard hard mode Mimiron for example). Except it took Stars like 1000+ tries to achieve this with the best gamers in the world and 25 pairs of hands.
And through many years of trials and tribulations, hardships and love, I have finally decided to speak the words I have meant to for so many years:
When u have made enough of these sick ai modifications will you release a map where the ai that u face will be using these :D also is it possible to make a map where the unit ai automaticly does this for a human player. Would be so cool to play with a friend on custom map with both of us having godly micro lol
On April 07 2011 18:39 zylog wrote: Wow that's really awesome. I have an idea of something that would be cool to see:
- as well all know, you can only drop units one at a time from an overlord - however, you can build as many overlords as you'd like because they don't cost food - so for a doom drop, place 1 unit per overlord (maybe hydras would be good?), and drop the whole force at once - you can use pickup/drop micro to either protect hydras that are getting targeted, or to do hit and runs - you can also drop some queens to transfuse injured units/overlords
Some other ideas: - void ray charge juggling - super fast and precise force fields - a few zerglings picking off a planetary fortress by running faster than the turret can turn - using phoenix as a support unit to pick up your units that are targeted by ground only units - mass neural parasite and then using other unit's abilities
those are awesome, like the ling vs planetary the most
I want to see a ridiculous micro fest of stalkers vs marauders.
A move a huge amount of stalkers into marauders, shoot once, blink away before damage and repeat with EVERY stalker >=D
Also, mass workers vs zealots. You can utilize right click on minerals to phase through the zealots. I'd love to see probes taking out zealots with no losses and no health damage.
Zealot vs workers and some units that kill the zealots over time. Theoretically you should be able to not lose any workers because they will run away after first hit and they have the same speed as zealots.
On April 07 2011 19:48 Marcus420 wrote: Great, more cheats to abuse the game. great.
These scripts are made with the map editor. This is no kind of hack nor shall it never be used on ladder. Like the OP said they are making it for fun and if it looks awesome while doing its thing then thats just a bonus :D
(i do hope we get a custom map with these for the ai though :D. maybe make it like the obs maps so your friends can see u get owned by uber micro)
On April 07 2011 17:55 justiceknight wrote: whats the code u used in sc2 editor? is it extremely long??
Probably not..
Event: unit starts attacking (or something like it, don't remember what it's called) Conditions: attacking unit is siege tank (sieged) And in the actions you put: "pick each unit in region (convrt circle to region; point; targeted unit, 2) and do: order unit to move targeting (point; targeted unit) offset by 2 towards angle between position of (targeted unit) and (picked unit) degrees."
Some of the terms are off but the above should more or less work.
Awesome. I'd love to see the Colossus vs various unit micro. Maybe ~20 zergling (or whatever number would not overpower a colossus) microd to lose as few as possible. Hydra would be very interesting too.
Stimmed infantry avoiding multiple storms would be interesting too.
Nice job. Pretty sick for the marines. But the zergling one was unrealistic tho.
On April 07 2011 19:47 Zaphid wrote: Zealot vs workers and some units that kill the zealots over time. Theoretically you should be able to not lose any workers because they will run away after first hit and they have the same speed as zealots.
On April 07 2011 19:47 Zaphid wrote: Zealot vs workers and some units that kill the zealots over time. Theoretically you should be able to not lose any workers because they will run away after first hit and they have the same speed as zealots.
I really really like the marine-split demo, because it actually shows what's possible with godlike micro. Makes me wanna see Flash vs Jaedong marine vs bling battles.
On April 07 2011 19:48 Marcus420 wrote: Great, more cheats to abuse the game. great.
These scripts are made with the map editor. This is no kind of hack nor shall it never be used on ladder. Like the OP said they are making it for fun and if it looks awesome while doing its thing then thats just a bonus :D
(i do hope we get a custom map with these for the ai though :D. maybe make it like the obs maps so your friends can see u get owned by uber micro)
Actually, its very possible that something like this has already been made and implemented.
On April 07 2011 19:48 Marcus420 wrote: Great, more cheats to abuse the game. great.
These scripts are made with the map editor. This is no kind of hack nor shall it never be used on ladder. Like the OP said they are making it for fun and if it looks awesome while doing its thing then thats just a bonus :D
(i do hope we get a custom map with these for the ai though :D. maybe make it like the obs maps so your friends can see u get owned by uber micro)
Actually, its very possible that something like this has already been made and implemented.
All it would take is for someone to make a 3rd party program to be able to utilize the same AI that you have implemented in SC2 custom map editor
Yeah i saw that too and i did not say its impossible i just said the maker of the automaton 2000 did it with map editor scripts and said(i think we can take his word for this :D) that it wont be used as a hack/it is no hack
i'd really love to see the marines vs banelings as well... I've been working on this lately and it could probably give a hint or two on how to do it optimally! thanks for sharing
I'd like to see the computer count how much SC2-APM it is executing to move the zerglings in the way it does =). How many of the 100 have to alter their path to avoid the tank fire and how fast those clicks would have to come.
The second I see something like this I wonder when the impossible becomes possible? Can a human user split 100 zerglings to defeat 20 siege tanks, and what APM did he perform at?
My gut tells me the coding required to actually generate the computer's APM would be as difficult as a human doing what the Automaton 2000 just did.
P.S. Shout out to that little guy, who first saw on Scrap Station and loved.
On April 07 2011 20:54 CursedFeanor wrote: that is scary! O_O
i'd really love to see the marines vs banelings as well... I've been working on this lately and it could probably give a hint or two on how to do it optimally! thanks for sharing
On April 07 2011 21:00 DreamOen wrote: is that a cheat tool post on team liquid??? Awaiting for the hammer
This is not a cheat tool! This is a ai code on the map editor that allows all of this. I have no idea how these things work but im quite sure u cant just use these in a ladder game. If we ever get a chance to play with these/against these it would be on a custom map only.
(btw i got no official info from the maker im just guessing pretty much :D)
This could actually be an immensely useful tool to test human limits (not so much for the zergling vs. tank micro since that is generally impossible).
I would highly suggest try to implement an APM-limiter on the bot. In other words, assign a maximum APM the bot is allowed to use (perhaps by requiring a certain amount of time to pass between actions). Then you could see what kind of APM you need to have to perform intense micro.
The only problem is that you might also want to implement boxed unit selection instead of individually commanding the units. Also I would like to see:
Is there no "cheating" at all involved? Does the script trigger by the AI seeing that the a unit has come in range of a tank and then trigger again as long as a unit is within the tank range sequentially or does it trigger when a tank starts aiming at a target? What would happen if none of the zerglings had vision of the tank when it fired(ie. if it was on a cliff)? I guess my question is, does the AI use any information not available to the player?
On April 07 2011 21:32 DrainX wrote: Is there no "cheating" at all involved? Does the script trigger by the AI seeing that the a unit has come in range of a tank and then trigger again as long as a unit is within the tank range sequentially or does it trigger when a tank starts aiming at a target? What would happen if none of the zerglings had vision of the tank when it fired(ie. if it was on a cliff)? I guess my question is, does the AI use any information not available to the player?
Technically, the information is available to the player; all you have to do is learn how seige tank targeting works through experimentation, right?
Haven't there been some people posting things about losing on ladder to people with 1000 apm and ungodly micro in the last few days? Or were they just complaining about supposed 'map hacks'.
can you please give me a slight hope that you will turn Automaton into a fully working AI bot that we could play on bnet? Like actually "Insane" AI, instead of the cheating crap we got from Blizzard.
can you please give me a slight hope that you will turn Automaton into a fully working AI bot that we could play on bnet? Like actually "Insane" AI, instead of the cheating crap we got from Blizzard.
So that Zergling vs Tank AI isn't cheating?
...right
In theory it isnt actually cheating a human could do it if he knew exactly how the tank targeting works also he would need like 8000 apm or something ridicilous like that so we will actually never get to see it(until that razor april fools machine becomes a reality at least). Oh also the human player would have to have the game sense to see which zergling would be targetted next(as the op says its the closest one so u should be able to deduce it :D)
so not actually cheating just playing really really fast :D
also theres a world of difference between a maphack and a extra minerals hack for the AI when we can make the games more fair by just giving them uber micro. A player cant over come extra minerals and extra info but he can overcome better control(how much better control a player can overcome is debatable but still the point stands :D)
can you please give me a slight hope that you will turn Automaton into a fully working AI bot that we could play on bnet? Like actually "Insane" AI, instead of the cheating crap we got from Blizzard.
So that Zergling vs Tank AI isn't cheating?
...right
In theory it isnt actually cheating a human could do it if he knew exactly how the tank targeting works also he would need like 8000 apm or something ridicilous like that so we will actually never get to see it(until that razor april fools machine becomes a reality at least). Oh also the human player would have to have the game sense to see which zergling would be targetted next(as the op says its the closest one so u should be able to deduce it :D)
so not actually cheating just playing really really fast :D
also theres a world of difference between a maphack and a extra minerals hack for the AI when we can make the games more fair by just giving them uber micro. A player cant over come extra minerals and extra info but he can overcome better control(how much better control a player can overcome is debatable but still the point stands :D)
No human could do that. period.
This would be a CHEAT if used on ladder. Dont argue that its not.
On April 07 2011 21:32 DrainX wrote: Is there no "cheating" at all involved? Does the script trigger by the AI seeing that the a unit has come in range of a tank and then trigger again as long as a unit is within the tank range sequentially or does it trigger when a tank starts aiming at a target? What would happen if none of the zerglings had vision of the tank when it fired(ie. if it was on a cliff)? I guess my question is, does the AI use any information not available to the player?
Technically, the information is available to the player; all you have to do is learn how seige tank targeting works through experimentation, right?
That was not what I was talking about. A player would not be able to dodge fire from a tank he does not see. And even in the case where the information could be derived from information the player has, deriving it yourself instead of hacking into the AI of the opponent is a much harder task.
can you please give me a slight hope that you will turn Automaton into a fully working AI bot that we could play on bnet? Like actually "Insane" AI, instead of the cheating crap we got from Blizzard.
So that Zergling vs Tank AI isn't cheating?
...right
In theory it isnt actually cheating a human could do it if he knew exactly how the tank targeting works also he would need like 8000 apm or something ridicilous like that so we will actually never get to see it(until that razor april fools machine becomes a reality at least). Oh also the human player would have to have the game sense to see which zergling would be targetted next(as the op says its the closest one so u should be able to deduce it :D)
so not actually cheating just playing really really fast :D
also theres a world of difference between a maphack and a extra minerals hack for the AI when we can make the games more fair by just giving them uber micro. A player cant over come extra minerals and extra info but he can overcome better control(how much better control a player can overcome is debatable but still the point stands :D)
No human could do that. period.
This would be a CHEAT if used on ladder. Dont argue that its not.
Yes but no one is talking about using it on the ladder. At least i cant see the point of using it on the ladder when if i dont actually have the skill(that and the apm jumping to stupid numbers would get me banned in a few hours :D)
And i admit no human could do it right now. probably never but what im trying to say is that the game uses all the info the player would have in that situation(if i understood it correctly) so in theory, a super human with more brainpower to process all the movements and info(and better speed and stamina than the normal human) could do it. Of course no one can actually do it but the AI is not actually cheating its just playing really really really fast :D.
PS: If ever used on the ladder in anyway the user should be banned ofc. but remember we are not talking about a hack bot here its just a AI script so unless blizz updates their ladder maps with it. It cannot be used in ladder in its current form
Wow thats amazing Shows how much potential to improve is still untapped
dont agree with the muta vid btw If the mutas are smart i am sure they can kill the phoenix on that map Since the phoenix is allowed godly control, then why are the mutas not?
mutas should spread all across the map aproaching the phoenix from all directions so it cant keep running tailing 50 mutas -.-
can you please give me a slight hope that you will turn Automaton into a fully working AI bot that we could play on bnet? Like actually "Insane" AI, instead of the cheating crap we got from Blizzard.
So that Zergling vs Tank AI isn't cheating?
...right
In theory it isnt actually cheating a human could do it if he knew exactly how the tank targeting works also he would need like 8000 apm or something ridicilous like that so we will actually never get to see it(until that razor april fools machine becomes a reality at least). Oh also the human player would have to have the game sense to see which zergling would be targetted next(as the op says its the closest one so u should be able to deduce it :D)
so not actually cheating just playing really really fast :D
also theres a world of difference between a maphack and a extra minerals hack for the AI when we can make the games more fair by just giving them uber micro. A player cant over come extra minerals and extra info but he can overcome better control(how much better control a player can overcome is debatable but still the point stands :D)
No human could do that. period.
This would be a CHEAT if used on ladder. Dont argue that its not.
You are missing the point. He's saying that the AI isn't breaking the rules of the game. The original post he's responding to calls Insane AI "the cheating crap we got from Blizzard", because to make it stronger, rather than make the AI smarter, Blizzard gave it the ability to break the rules of the game (maphack, mineral hack).
As to whether it would be cheating to put a bot with abilities like these on the ladder (as the original post seems to suggest), I think as long as it was clearly labeled as a bot, that would be fine. Not sure what Blizzard's TOS say though.
can you please give me a slight hope that you will turn Automaton into a fully working AI bot that we could play on bnet? Like actually "Insane" AI, instead of the cheating crap we got from Blizzard.
So that Zergling vs Tank AI isn't cheating?
...right
In theory it isnt actually cheating a human could do it if he knew exactly how the tank targeting works also he would need like 8000 apm or something ridicilous like that so we will actually never get to see it(until that razor april fools machine becomes a reality at least). Oh also the human player would have to have the game sense to see which zergling would be targetted next(as the op says its the closest one so u should be able to deduce it :D)
so not actually cheating just playing really really fast :D
also theres a world of difference between a maphack and a extra minerals hack for the AI when we can make the games more fair by just giving them uber micro. A player cant over come extra minerals and extra info but he can overcome better control(how much better control a player can overcome is debatable but still the point stands :D)
No human could do that. period.
This would be a CHEAT if used on ladder. Dont argue that its not.
You are missing the point. He's saying that the AI isn't breaking the rules of the game. The original post he's responding to calls Insane AI "the cheating crap we got from Blizzard", because to make it stronger, rather than make the AI smarter, Blizzard gave it the ability to break the rules of the game (maphack, mineral hack).
As to whether it would be cheating to put a bot with abilities like these on the ladder (as the original post seems to suggest), I think as long as it was clearly labeled as a bot, that would be fine. Not sure what Blizzard's TOS say though.
Assuming it's anything like the WoW ToS, modifying the game in any way is a violation. The real question is will they enforce it.
Hi there, an engineer myself, I have lots of ideas on this.
Run some iterations to find out the average functions of "X lings vs Y tanks, lings win (find the minimum value of X)" and the same with rines/banes. And/or "X lings can kill Y tanks with Z lings survive", also "X lings vs Y tanks, Z tanks survive".
Well, you could also do a general script where you check the speed, acceleration and range of two units to see if kiting is possible. When this is true, do a "kiting" AI.
Would be pretty cool to see on ladder one day, the other player says: do you object to playing an autonomous AI? If no answer or 'yes' answer, it ggs. If 'no' objection, it says glhf and plays on. How high do you think it could get?
wow insane excuse me for being ignorant but in what concretly this would help ? to play against perfect micro AI ? Or is it just for show ? It seems like this bot can be abused on ladder o.o'?
Really awesome like Severian said is kinda cheating. For a human to do that you would have to calculate which units the Tanks are targeting which would be a godly feat in its self then have the dexterity to split the lings in a fraction of a second I even think Superman would have trouble with it...if he was real
with this bot the units that can kite stuff without being hit (marauders vs anything with range <6 ; banshees vs marines ; marines vs banes etc..) will become unkillable wich i think is pretty interesting for single player practice vs AI.
Anyways that video of the lings sure gave me a boner, sooo beautiful like little flowers =)=)=)
would luv 2 see somethin like this in yabot, perhaps with an ajustable difficulty so its not so surreal. to have computer execute a build and micro properly, to practise against.
That marine vs. baneling micro was ridiculous. It's amazing that you can do this just using the map editor. Reminds me of this awesome AI for SCBW: http://www.youtube.com/watch?v=NcjQhig88wE That one required an external program though.
Wouldn't that zergling micro actually be possible? Or atleast a less insane version of it to a certain degree.
Just an example, you have a group of zerglings running into a tank, and you box an amount of zerglings with the box being about the size of the siege splash, shiftclicking one zergling in the unit selection screen to de-target it, then move ur selection back a notch and shift-a-move back?
That was actually REALLY impressive to watch! There needs to be a whole series created of these for our viewing pleasure.
On April 07 2011 23:11 Marksel wrote: Wouldn't that zergling micro actually be possible? Or atleast a less insane version of it to a certain degree.
Just an example, you have a group of zerglings running into a tank, and you box an amount of zerglings with the box being about the size of the siege splash, shiftclicking one zergling in the unit selection screen to de-target it, then move ur selection back a notch and shift-a-move back?
No because this AI makes only the TARGETTED zergling not run away, we have no idea which zergling the enemy tank is firing at so it wouldn't work at all. Not to mention the fact shift moving commands are really terrible.
On April 07 2011 23:11 Marksel wrote: Wouldn't that zergling micro actually be possible? Or atleast a less insane version of it to a certain degree.
Just an example, you have a group of zerglings running into a tank, and you box an amount of zerglings with the box being about the size of the siege splash, shiftclicking one zergling in the unit selection screen to de-target it, then move ur selection back a notch and shift-a-move back?
Yes I was wondering why pro don't send 3-4 splitted zergling just before their mass zergling in order to absorb the first few blow with minimal loss. It wouldn't require a lot of APM and may reduce the tank effectiveness quite a bit against speedling...
I had an idea to make this a while ago, but unfortunately, I have no idea how to program.
But my idea was that towards the beggining of the game, you bring 8 workers to the opponents base, and using the thing where they cant mine if you're mining their minerals, completly stop them from mining! This would require a ton of micro to keep them alive, as well as not allowing any mining.
The cool thing with this, is even if the micro is not possible to obtain what it can show is "here is what the ideal micro would look like." So you can try your best to mimic it to some degree, how best to split, etc. So pretty cool use because it at least gives you an idea of what you should be attempting to do to avoid splash or split groups even if you only mimic it 10% it is a start.
On April 07 2011 23:11 Marksel wrote: Wouldn't that zergling micro actually be possible? Or atleast a less insane version of it to a certain degree.
Just an example, you have a group of zerglings running into a tank, and you box an amount of zerglings with the box being about the size of the siege splash, shiftclicking one zergling in the unit selection screen to de-target it, then move ur selection back a notch and shift-a-move back?
No because this AI makes only the TARGETTED zergling not run away, we have no idea which zergling the enemy tank is firing at so it wouldn't work at all. Not to mention the fact shift moving commands are really terrible.
I think you misunderstood how the AI know which Zergling will be hit :
The OP states that the tank always targets the nearest Zergling. Of course, in a Zergling blob, it's nearly impossible for an human eye to see which one it is. However, if you move on purpose a single zergling in front of the other, he will be the one targetted by the tank. So by using this, it may be possible to reduce splash damage.
On April 07 2011 23:11 Marksel wrote: Wouldn't that zergling micro actually be possible? Or atleast a less insane version of it to a certain degree.
Just an example, you have a group of zerglings running into a tank, and you box an amount of zerglings with the box being about the size of the siege splash, shiftclicking one zergling in the unit selection screen to de-target it, then move ur selection back a notch and shift-a-move back?
No because this AI makes only the TARGETTED zergling not run away, we have no idea which zergling the enemy tank is firing at so it wouldn't work at all. Not to mention the fact shift moving commands are really terrible.
I think you misunderstood how the AI know which Zergling will be hit :
The OP states that the tank always targets the nearest Zergling. Of course, in a Zergling blob, it's nearly impossible for an human eye to see which one it is. However, if you move on purpose a single zergling in front of the other, he will be the one targetted by the tank. So by using this, it may be possible to reduce splash damage.
Isn't this only true for the first shot? Afterwards it would use smart-fire and spread its targets, so from that point on its pretty much impossible to predict as a non-AI?
The baneling one gave me zerg tears, the better players get at controling the marines the less damage them banes will be doing, really bloody awesome tinkering in the editor man, thumbs up!
On April 07 2011 23:11 Marksel wrote: Wouldn't that zergling micro actually be possible? Or atleast a less insane version of it to a certain degree.
Just an example, you have a group of zerglings running into a tank, and you box an amount of zerglings with the box being about the size of the siege splash, shiftclicking one zergling in the unit selection screen to de-target it, then move ur selection back a notch and shift-a-move back?
No because this AI makes only the TARGETTED zergling not run away, we have no idea which zergling the enemy tank is firing at so it wouldn't work at all. Not to mention the fact shift moving commands are really terrible.
I think you misunderstood how the AI know which Zergling will be hit :
The OP states that the tank always targets the nearest Zergling. Of course, in a Zergling blob, it's nearly impossible for an human eye to see which one it is. However, if you move on purpose a single zergling in front of the other, he will be the one targetted by the tank. So by using this, it may be possible to reduce splash damage.
Isn't this only true for the first shot? Afterwards it would use smart-fire and spread its targets, so from that point on its pretty much impossible to predict as a non-AI?
Judging by the OP answer, it seems it's true for all shots :
Regarding the zerglings splitting before the shot lands. The script is capable of determining which zergling that is going to take the next blow. When a tank starts aiming (when its target dies for example), the bot predicts it's gonna fire at the closest zergling and executes a split the AI of Starcraft 2 is very predictable in that meaning..
Unless he means the closest zergling to the previous target, it seems to always be the closest to the tank
On April 07 2011 23:11 Marksel wrote: Wouldn't that zergling micro actually be possible? Or atleast a less insane version of it to a certain degree.
Just an example, you have a group of zerglings running into a tank, and you box an amount of zerglings with the box being about the size of the siege splash, shiftclicking one zergling in the unit selection screen to de-target it, then move ur selection back a notch and shift-a-move back?
No because this AI makes only the TARGETTED zergling not run away, we have no idea which zergling the enemy tank is firing at so it wouldn't work at all. Not to mention the fact shift moving commands are really terrible.
I think you misunderstood how the AI know which Zergling will be hit :
The OP states that the tank always targets the nearest Zergling. Of course, in a Zergling blob, it's nearly impossible for an human eye to see which one it is. However, if you move on purpose a single zergling in front of the other, he will be the one targetted by the tank. So by using this, it may be possible to reduce splash damage.
Isn't this only true for the first shot? Afterwards it would use smart-fire and spread its targets, so from that point on its pretty much impossible to predict as a non-AI?
There is no such thing as smart-fire. Tanks simply do not overkill their targets because they don't have a projectile - the damage is instant.
What you want to do bumrushing tanks is spread as much as possible (maybe use patrol) and a-click to a point well beyond the tanks so that the lings do not converge.
I would like to see hellion vs speedling micro (currently, speedlings "counter" blue flame hellions because of their speed on creep, though perfect micro of the hellions should mitigate this). Muta vs Archon and Queen vs anything (perfect transfuse should keep all queens alive). anything vs Colossus would be sweet as well
When the zerglings split, don't some overlap (ie. when two tanks are going to hit two zerglings and the zerglings around the ones that are going to be hit clump up). With the clumping, isn't it impossible for perfect micro? or am i wrong in that regard? =S
holy shiz?! This is an excellent example of PERFECT micro which everyone is suppose to aim to get but is impossible lol. I think the pros once said that it is impossible to get perfect micro but it was possible to get perfect macro. This is a great example of perfect micro that no one can ever achieved but it sure does show the potential of having perfect micro is :O
It was very enlightening because now we've SEEN what "perfect" micro is in these situations. With that in our pocket, the pros know what flawless is, and they have something to try to accomplish, even if it will make their mouse explode, or just plain impossible.
I would like to suggest another micro scenario. 1) Attacking Colossi with Marines and using the exact same method of singling out the marine being targetted. 2) A single blink stalker that is up against 100 projectile-shot enemies. The stalker has: Infinite attack power. Can only survive two hits. 3 second Blink cooldown.
I wonder if any of these videos will change the way pro-gamers approach their micro. Obviously you can't replicate the zergling split thingy and the marine micro is pretty much impossible as well. But other types of units might have some ingenious micro trick to them that we haven't thought of yet.
You need to make more! It's pretty insane what you can do when everything is completely perfect. ^^
can you please give me a slight hope that you will turn Automaton into a fully working AI bot that we could play on bnet? Like actually "Insane" AI, instead of the cheating crap we got from Blizzard.
So that Zergling vs Tank AI isn't cheating?
...right
In theory it isnt actually cheating a human could do it if he knew exactly how the tank targeting works also he would need like 8000 apm or something ridicilous like that so we will actually never get to see it(until that razor april fools machine becomes a reality at least). Oh also the human player would have to have the game sense to see which zergling would be targetted next(as the op says its the closest one so u should be able to deduce it :D)
so not actually cheating just playing really really fast :D
also theres a world of difference between a maphack and a extra minerals hack for the AI when we can make the games more fair by just giving them uber micro. A player cant over come extra minerals and extra info but he can overcome better control(how much better control a player can overcome is debatable but still the point stands :D)
No human could do that. period.
This would be a CHEAT if used on ladder. Dont argue that its not.
And who said anything about using it on Ladder? Read more carefully before posting comments like that.
While the marine video I think is crazy to watch, if you think about the micro involved, the zergling vs. siege tank one was far more impressive. I mean holy shit: not one tank did splash damage.
Edit: Regarding the zerglings splitting before the shot lands. The script is capable of determining which zergling that is going to take the next blow. When a tank starts aiming (when its target dies for example), the bot predicts it's gonna fire at the closest zergling and executes a split the AI of Starcraft 2 is very predictable in that meaning..
After reading your edit on the Zergling micro, would it work if the T microed the tanks to fire at specific targets? As they don't overkill, you could basically shift-attack through the bunch, and if aiming for the middle lings, the casualties could be about just as severe as if the lings were unmicroed. You could probably still write a script that slice the lings into a number of groups, two that are 'outside' the target line, and one that is inside. The middle line would then be sliced into smaller groups, as it is not known which zergling will be targeted.
Have an automaton to control the tanks, and another one to control the lings - let havoc ensue!
That rine video is just so awesome, the zling one is just more hilarious with all the zlings "jumping" out of the way(kinda picturing it more visually than just the screen )
In the zerglings vs tanks, it's not micro, your bot is cheating by using the info of which ling is targetted. Even with perfect micro, it is not possible to achieve with normal information.
Marine vs bannelings is different, it really shows the limit of the unit. But to be fair, the bannelings sould also be controlled by a bot.
how about hydras + overlords vs colossi? Since colossi don't deal damage at the beginning of their attack animation, couldn't you pick up the units before the damage is dealt and waste all their shots?
On April 08 2011 02:33 dudeman001 wrote: how about hydras + overlords vs colossi? Since colossi don't deal damage at the beginning of their attack animation, couldn't you pick up the units before the damage is dealt and waste all their shots?
thats a good idea.
Perhaps another one could be perfect phoenix micro vs mutas.
On April 08 2011 02:29 Elean wrote: In the zerglings vs tanks, it's not micro, your bot is cheating by using the info of which ling is targetted. Even with perfect micro, it is not possible to achieve with normal information.
Marine vs bannelings is different, it really shows the limit of the unit. But to be fair, the bannelings sould also be controlled by a bot.
Isn't it just predicting which ling is going to be targeted through distance alone? That's entirely fair.
you guys are so badass for making this. Genius! oh god one day I know people will start marine splitting like that (come on flash!). great work! now let me fight against this AI!
random amounts but in my head i've always thought it would be possible to minimize the amount of workers you lose to zerglings in your mineral line by a lot by clicking the minerals on ones thats are getting hit
This was one of the coolest videos I have ever seen, If a human player could micro like the automaton he would be an instant legend/sc2 bonjwa. Also great music with the videos thanks for posting! :D
If I ever see anybody pull off this kinda stuff, i would totally marry him/her/it!!!!
This is also a very good idea for another videos:
On April 07 2011 23:38 Munk-E wrote: But my idea was that towards the beggining of the game, you bring 8 workers to the opponents base, and using the thing where they cant mine if you're mining their minerals, completly stop them from mining! This would require a ton of micro to keep them alive, as well as not allowing any mining.
Obvious one is the double extractor trick 12 drone rush vs terran. It's already pretty tricky to stop if zerg has human micro, but with automaton 2000 micro, it would maybe be unstoppable ?
My question has been touched upon earlier in the thread but I haven't seen a satisfying answer so far... so here goes:
I notice the scripts are activated by chat commands. Would it be possible to code this same micro-scripting into a custom AI without needing to prompt the actions with a chat command? (I.E. Green Tea custom AI) I'd imagine facing a computer opponent that had this kind of micro skill could be beneficial for improving your own micro.
So in short, are these one-off demonstrations or could they be implemented as a full-fleshed custom AI to play against?
I'd like to see the banelings vs marine fight again, but this time give banelings perfect micro. Can the banelings force the marines to split enough that they can't dps the banelings down fast enough?
On April 08 2011 04:04 ghostsquall wrote: This is interesting, but I don't see the point in creating it...kind of worries me that I'll run into this sort of thing on ladder. :<
It's not a "bot" per se. it's simply a demonstration of the SC2 map editors scripting functions. This isn't a "bot" program that could micro for you during regular play, think of it as a mouse-trap course or an elegant set of dominoes, sure it's amazingly impressive, but it's all scripted on a pre-determined and coded scenario from what I understood.
-not a bot program
-not possible to use it to "cheat" on the ladder
-all actions are scripted from WITHIN the sc2 map editor
On April 08 2011 04:09 ghostsquall wrote: oh ok! :D that makes me feel a bit better
no worries, I was afraid of the same thing before reading the initial responses from the OP, they should probably edit the original post to include a disclaimer addressing these concerns
I didnt know jaedong and flash switched over already o.0. but seriously that was awesome. The marine micro was insane if I ever see that I could get shot after watching that and have a smile.
The zergling micro was insane. At first I expected them to just stay spread out the entire time, but watching them actively "morph" around the one taking the hit was REDONKULOUS! hahaha.
Great job programming the AI. I'd be interested to see the marine/baneling scenario in a more enclosed environment, as we rarely see marines with that much room to more. Throw that scenario with some buildings around or a wall, maybe?
The splitting marines against banelings was pretty awesome. The ling/tank one wasnt as impressive because it is pretty clear the bot knows which ling the tanks are going to fire at(it is cheating).
On April 08 2011 04:12 LittLeD wrote: Wow, actually the game would be quite epic if you had a button that did just that. Like a "Push everything away around a circle"-thingy.
Age of empires had a scatter button if I recall. I think most would argue that such a feature negates the need for a lot of micro so overall isn't a good feature.
I would suggest anything involving blink play in the protoss area. The other micro techniques in protoss are too situational based to be done by a computer. The only possible one besides blink is a FF trap with storm.
Or, possibly using zealot charge to run into something? I'm not quite sure.
Impressive. I expected the lings vs tanks one to be more about having a stagger formation to minimize splash, though. As in, I expected these videos to show what kind of micro we might expect in the future from future bonjwas. A human player would obviously be unable to pull that off. The marines vs banelings was very nice though, and I could certainly see that happening in the future.
On April 08 2011 05:32 Rettet181 wrote: The marines were fantastic. Would perfecetly-microed banelings even be able to do any damage against perfectly microed marines?
On April 08 2011 05:32 Rettet181 wrote: The marines were fantastic. Would perfecetly-microed banelings even be able to do any damage against perfectly microed marines?
pretty sure not.
banelings are slower then marines :>
not on creep! but there should never be creep
sick videos, would love to see many lings/rines/workers/hydras dodge as many blue flame hellions as possible (not taking splash hits)
On April 08 2011 05:32 Kantutan wrote: Impressive. I expected the lings vs tanks one to be more about having a stagger formation to minimize splash, though. As in, I expected these videos to show what kind of micro we might expect in the future from future bonjwas. A human player would obviously be unable to pull that off. The marines vs banelings was very nice though, and I could certainly see that happening in the future.
Its humanly not possible to micro marines like that. Its just not doable, not even if you would spend 24/7 on marine vs baneling micro. Its so many actions. The marine split in the beginning alone was selecting 20 marines and spreading them out perfectly with a move command I guess then individually moving single marines with hit and run while others run etc pp. Too many actions to pull off.
On April 08 2011 05:32 Kantutan wrote: Impressive. I expected the lings vs tanks one to be more about having a stagger formation to minimize splash, though. As in, I expected these videos to show what kind of micro we might expect in the future from future bonjwas. A human player would obviously be unable to pull that off. The marines vs banelings was very nice though, and I could certainly see that happening in the future.
Its humanly not possible to micro marines like that. Its just not doable, not even if you would spend 24/7 on marine vs baneling micro. Its so many actions. The marine split in the beginning alone was selecting 20 marines and spreading them out perfectly with a move command I guess then individually moving single marines with hit and run while others run etc pp. Too many actions to pull off.
I would say it's doable with few marines lost and rest on deep red, cause you would need more stims and more space.
Blink stalkers vs Colossi would be awesome to see.
Marines vs Colossi. Marines targeted by a Colossus would move a hex forward to avoid splash damage to the other marines. Much like the Zerglings vs Tanks.
Also +1 for burrow roaches vs blink stalkers idea.
Perfect force fields against mass zerglings maybe? Or against perfectly microed mass roaches with burrow.
Perfectly split zealots against ultralisks, to avoid ultralisk splash dmg.
Perfect blink micro against mutalisks. Stalkers avoiding the glaive wurm bounce, and blinking back the stalkers with exactly 0 shields.
On April 07 2011 18:02 Assault_1 wrote: doesn't this require knowledge of the future? how does it know which ling the tank will target
awesome vid though
I think it must use a script function to get each tank's current target (or to determine if a unit is being targeted). So it's kind of cheating.
I agree with this, the baneling video is totally possible, but the only way for the lings to work is for the player to know which ling is about to be attacked
On April 08 2011 06:00 EEhantiming wrote: roach vs Automaton 2000 reapers would be nice marauder vs Automaton 2000 blink stalkers hellion vs Automaton2000 speedlings
I don't think it's actually possible to kite roach with reapers--the reaper has a long attack animation that holds it in place for enough time for the roach to hit it. I'd like to see blink stalkers against..well...anything, really. Colossus/zealot/stalker vs. blink stalker would be cool. Splash dodging+blink micro.
This is just sick micro! I wish we could see more people doing the co-op type of thing we saw in TL Attack for SC2 ... if they allow that, which I'm not sure of.
On April 07 2011 18:02 Assault_1 wrote: doesn't this require knowledge of the future? how does it know which ling the tank will target
awesome vid though
I think it must use a script function to get each tank's current target (or to determine if a unit is being targeted). So it's kind of cheating.
I agree with this, the baneling video is totally possible, but the only way for the lings to work is for the player to know which ling is about to be attacked
I don't think so. Someone who is really good at micro might be able to do a decent job, but not like that video. You have to think, a human with 300 efficient APM would be incredible, the computer is capable of millions of APM.
holy fuck, this is crazy. I mean, it's leaving out any kind of human control on both sides so ofc this is not representative of how a real game could be played, but it shows how high the skill ceiling for micro CAN be.
I'd love to see some "sick micro to save marauders from being imprisoned in force fields" tests.
roach vs reapers would have worked before the range buff on the roaches . marauders shouldn't care about blink if they just run in. a move hellions just die against lings, there would be almost no micro need. Just my opinion though :3. So still for ground stuffies against colossi hehe. (as everyone complains about their damage output)
the replay should show the ai apm i guess. Otherwise you could restrict the apm of the ai and work out when the apm you grant isn't enough anymore xD.
Jesus Christ, you guys rock, damn I enjoyed those videos so much. Thank you you rock. Imho this should be played as an introduction to either NASL and/or TSL. They get you in the mood of epic Starcraft.
I think to make this rine vs bling possible u need : -dual core brain -2 plug in and simultaniously running mouses - maybe a extra hand ? (well u can try to use one foot for a mouse)
all in all kind of sad to watch how limited humans are in comparison to a script
The baneling video . . . my jaw is on the floor. Anyone who ends up being able to manage something like that will become the TvZ king and a SC2 legend.
blink stalk vs marauders mutalisk vs thors vikings vs phoenix burrow roach vs gateway units speed reapers killing a base while dodging enemy units mnm vs tank line phoenix vs hydras
On April 08 2011 05:32 Kantutan wrote: Impressive. I expected the lings vs tanks one to be more about having a stagger formation to minimize splash, though. As in, I expected these videos to show what kind of micro we might expect in the future from future bonjwas. A human player would obviously be unable to pull that off. The marines vs banelings was very nice though, and I could certainly see that happening in the future.
Its humanly not possible to micro marines like that. Its just not doable, not even if you would spend 24/7 on marine vs baneling micro. Its so many actions. The marine split in the beginning alone was selecting 20 marines and spreading them out perfectly with a move command I guess then individually moving single marines with hit and run while others run etc pp. Too many actions to pull off.
I would say it's doable with few marines lost and rest on deep red, cause you would need more stims and more space.
Blink stalkers vs Colossi would be awesome to see.
Yeah, pretty sure someone could not lose a single marine in an encounter like that. Obviously they wouldn't be controlling every single marine separately but would still be able to kite the banes like that.
Its not possible for any human to ever micro that perfectly, you would have to have insane APM without missing a SINGLE click just to do the marine vs baneling split because you have to have some on move command and alternating between move and attack command on each individual marine.
Once players get really good, we're going to see more incredible micro like this. I don't think players will ever be capable of this level of micro though.
On April 08 2011 06:48 dmillz wrote: Its not possible for any human to ever micro that perfectly, you would have to have insane APM without missing a SINGLE click just to do the marine vs baneling split because you have to have some on move command and alternating between move and attack command on each individual marine.
I think everyone realizes this, that doesn't change the fact that it's bad ass to watch though
The Lings vs Tanks video really doesn't show much unless you code AI in to the tank to automatically detect the best possible target and/or a random target, accounting for human control. It really is impossible to detect what the next target will be against something that randomly selects targets, you know
Even if those videos were programmed, they were totally badass. I had to re-watch the marine vs baneling video like 5 times to make sure that I was seeing that correctly.
Marines don't counter banelings. Perfect micro does :D
Can't want to see what other types of sick videos you guys come out with!
On April 08 2011 06:48 dmillz wrote: Its not possible for any human to ever micro that perfectly, you would have to have insane APM without missing a SINGLE click just to do the marine vs baneling split because you have to have some on move command and alternating between move and attack command on each individual marine.
I think everyone realizes this, that doesn't change the fact that it's bad ass to watch though
I posted that because I keep reading comments from people saying how in a few years the top pros will totally be microing marines like that vs banelings, and there isn't a chance lol.
And then see a thread like the one that I'm posting in... I mean I don't think that the OP is using his Micro Bot for any malicious intentions but just makes you wonder.
And then see a thread like the one that I'm posting in... I mean I don't think that the OP is using his Micro Bot for any malicious intentions but just makes you wonder.
This is done through the map editor. It's not like you can get that to work in a ladder match.
Maybe we will see players in the future who can come close with their marine micro to the one shown in the video but I highly doubt it. Players are already playing a lot, and got a lot of experience from their BW/WC3 background, and even 800APM don't make it possible to do simultaneous commands like this bot does.
And then see a thread like the one that I'm posting in... I mean I don't think that the OP is using his Micro Bot for any malicious intentions but just makes you wonder.
This is done through the map editor. It's not like you can get that to work in a ladder match.
Yeah that's what I read... Guess it's not as scary now. Doesn't seem as malicious.
On April 08 2011 07:47 AJ- wrote: do blink stalker vs chargelots
Don't think that would work anymore due to the 1.3 changes. Chargelots will always hit once so I'd imagine they'd continuously charge until they make contact with the stalker either before or after they blink. Or rather, it'd be very unimpressive as the stalkers would still take hits.
I remember seeing a thread like this back during broodwars (pretty sure it's BWs.. otherwise i'm going insane).
Someone should find that video where the computer kills sunkens in BW with goliaths/tanks by dropping them and picking them up instantly and add it to the OP for nostalgia/comparison .
Either way awesome to watch what perfect micro does. Hoping for more videos in the future. I expect a phoenix vs muta, or viking vs other air would be easy to make.
The two best ones i can think of are: drop ship pick up micro (preferably something like immortals or tanks) with 5 drop ships and ten tanks/5warp prisms and ten immortals at the same time. Against like 100 roaches or something ridiculous. And mass queens with perfect transfuses.
LOL this is amazing! Although as people have said, its realistically impossible (for the tanks at least) because we can't predict what the tank AI will shoot
one day an RTS game will be fought between computers that can beat humans 100% of the time, their micro will be like this, I think mass marines would be impossible to defeat with this activated against every possible AoE
On April 07 2011 18:01 Aequos wrote: On the Malicious Stalkers - apparently, if you blink perfectly, you can dodge projectile style attacks (making them do 0 damage). I'd like to see 20 or so Blink Stalkers take on a 30-40 stimmed marauder force.
On April 07 2011 18:01 Aequos wrote: On the Malicious Stalkers - apparently, if you blink perfectly, you can dodge projectile style attacks (making them do 0 damage). I'd like to see 20 or so Blink Stalkers take on a 30-40 stimmed marauder force.
This is a great idea. I second this Nothing is impossible, when you use the Automaton 2000. Brilliant job by the way!! The marine split was especially mind blowing to watch.
Another idea for you would be to micro cracking vs zealots attacking and pulling away before the zealot can swipe for damage. Don't know how possible this is..
This is awesome! Some hacker was using roach burrow micro hacks I read about so maybe this isnt impossible to implement after all...This was super friggin cool though
Wow, that's very impressive! Of course, you could have programmed your tank AI to negate this trick very easily if you wanted to by randomly choosing a zergling to target one ms before firing .
On April 08 2011 10:31 FluidTek wrote: Haha that marine micro was so impressive, i think someone could pull off something similar in the near future.
I doubt it's physically possible to perform at ~15,000 APM, regardless of training. Granted, having said that, I imagine MKP's going to get a robot arm specifically to do so.
The first baneling vid made my jaw drop to the floor, then i saw the marines that microed perfectly against the banelings nice job dudes! i wonder if you can: Viking vs BC perfect micro BurrowRoach vs Sentires/stalkers Dodging Colosus hits with lings or marines?
On April 07 2011 17:56 Rb6v King wrote: That was pretty cool - try recreating something that was done in bw, but in sc2.
Shitton of mutas vs shitton of ht with storm, make no muta die.
Gogogoogogo
This was done in starcraft 1 already, not sure if it needs to be done again really.
I guess it could be done in SC2 again... see if things changed much.
Personally I'd want to see an upgraded burrowed roach micro fight against something like marine-medivac, or marine-siege tank.
Marine or hydralisk vs colossus would be nice too. Marines have kinda low health and range though, and hydralisks are kinda slow, so those might be issues— maybe more just challenges though.
On a side note while it's not as amazing, Green tea/Fyn AI seems to have excellent unit micro vs. banelings— they will spread out to make banes rather inefficient.
On April 07 2011 17:56 Rb6v King wrote: That was pretty cool - try recreating something that was done in bw, but in sc2.
Shitton of mutas vs shitton of ht with storm, make no muta die.
Gogogoogogo
This was done in starcraft 1 already, not sure if it needs to be done again really.
Personally I'd want to see an upgraded burrowed roach micro fight against something like marine-medivac, or marine-siege tank.
On a side note while it's not as amazing, Green tea/Fyn AI seems to have excellent unit micro vs. banelings— they will spread out to make banes rather inefficient.
Well, these small scenarios like these are steps towards implementing as close to "perfect" SC2 AI as possible, which I'm all for. There shouldn't be a reason not to aim for one for both BW and SC2. Keep up the amazing work!
ㅋㅋㅋseriously, I don't think pro players could co-op vs an AI like this and win...unless every pro was only controlling one unit..still doubt it though.
Maybe can you make two AIs like this vs each other in different scenarios and see how the game is played in perfect-land^^
Picture yourself in a GSL final and you would do a marine kite vs banelings like that, damn the crowd and commentators would scream so fucking much that the roof would explode. You would be a king for the rest of your life.
I would definitely watch a league of only perfect ai fighting against each other just so I could have my mind blown every 2 minutes at the crazy stuff like this. GSLai make it happen...
These videos made my day . I was literally laughing out loud seeing that perfect Marine micro vs Speed Banes. Oh my goodness - that was AWESOME (but kinda scary at the same time ).
IMO, the Insane AI shouldn't cheat resource-wise, but should have micro like this, for the lols .
Great job, you guys!
EDIT: would also love to see a game with the AIs fighting each other XD.
On April 08 2011 16:57 WhiteLight wrote: I would definitely watch a league of only perfect ai fighting against each other just so I could have my mind blown every 2 minutes at the crazy stuff like this. GSLai make it happen...
There is an ai vs ai competition in broodwar, i dont see why it wouldnt be possible in sc2... Microing is one thing, they have to macro and such and make decision vs pro players too.. Some pretty amazing games happen in ai vs pros
I'm not sure if its been mentioned before but it would have been nicer to see the marines constrained walls, for instance in the main base of the opponent.
Amazing stuff. To be honest, I liked the zergling vs. tanks scenario a lot more than the marine vs. banelings, simply because the marines do what everybody tries to do in that situation (although obviously with much more success), while the zergling attack is very original, because not only does it overburden the APM limits of human players, but also their analytic capabilities.
Another scenario I'd like to see would be a couple of zerglings taking out a line of SCVs while avoiding the shots from the planetary fortress the SCVs are mining at. I'd guess that this should actually be a lot easier than the tank assault, and so I'm really curious whether this could theoretically be done within APM-constraints that would allow it to be reproduced by a human player.
Also with the APM mentioned (15,000 for the marines and 8,000 for the ling), it would be impossible for a human to achieve. It would be interesting to see what can be done using APM humans could achieve, eg set a 300-400 APM cap and run the AI.
how about marines vs colossi splash??? I think that's vertical splitting micro required - so have to keep units aligned at a right angle to the colossus' attack!
if you do that one... i will marry you. for reals.
On April 08 2011 19:26 Poffel wrote: Another scenario I'd like to see would be a couple of zerglings taking out a line of SCVs while avoiding the shots from the planetary fortress the SCVs are mining at. I'd guess that this should actually be a lot easier than the tank assault, and so I'm really curious whether this could theoretically be done within APM-constraints that would allow it to be reproduced by a human player.
This has essentially already been done by humans (not attacking scvs though, not that that's much harder). It's extremely difficult to do, but certainly very humanly possible. The biggest issue is comming from the right angle at the right time... something that is much less viable in real games where the space is limited
I would like to see a perfect 4-unit surround, on for example speedlings vs stalker. (A 4 unit surround is when you position units all around an enemy unit with a perfect timing so that the attacked unit cannot move).
A "perfect" bioball of marines/marauders that all stuttermoved individually would also be impressive. Most players, even pros, stutter with one controlgroup which is sometimes ineffecient.
Next, pair the perfect zerglings up against the perfect bioball.
Oh, and as a zerg, some epic roach-splitting versus colossi would be AWESOMESAUCE.
(Note that I do realize this is just for showoffs, which is good - because then I can "safely" appreciate the work done without worring about abuse)
Overpowered!!! But the only one thats doable (well not maybe doable but anyway) is the marinesplitting one, the other one to make the lings live is impossible first there is no such command to make the lings spread out to all directions from the one whos getting hit (sure, you could take each ling for itself and move away with it, but no thats not something you could ever dream of to be able to do in such minimal time), and secondly you dont know in beforehand which ling is gonna be attacked (if you werent a computer).
the marine vs baneling vid actually makes me think it might be fun if someone made a custom map for ladder where every player controls 1 marine in a squad and you just see how many waves of banelings you can survive. it's like marine micro with multiple players to increase APM
Not sure how to express how awestruck i am right now. that zergling micro was just AMAZING!. of course the marine thing was sexy but i am still implressed by the sling stuff. GG
The speedlings is impressive, given the targeting system and such, but the fluidity of the marines around those banes was gorgeous. Just gorgeous. Like water around the mob of blings. So funny and awesome to watch 20 marines stutter step at different times all together xD
On April 08 2011 16:55 Termit wrote: Picture yourself in a GSL final and you would do a marine kite vs banelings like that, damn the crowd and commentators would scream so fucking much that the roof would explode. You would be MarineKing for the rest of your life.
how about trying to implement something like this for blink stalkers? where the stalkers blink away before the projectile hits them. God that would be so frustrating...
I find it interesting that these AI hacks are appearing on the ladder now. I also find it hard to believe that this was just a little weekend project that you and your buddy threw together for giggles.
On April 09 2011 04:08 hpTheGreat wrote: I find it interesting that these AI hacks are appearing on the ladder now. I also find it hard to believe that this was just a little weekend project that you and your buddy threw together for giggles.
Just saying. Read between the lines...
Maybe instead of reading between the lines you should read the actual lines first? This is made using the map editor, no third party programs and as such cannot be used on ladder.
On April 09 2011 04:08 hpTheGreat wrote: I find it interesting that these AI hacks are appearing on the ladder now. I also find it hard to believe that this was just a little weekend project that you and your buddy threw together for giggles.
Just saying. Read between the lines...
Maybe instead of reading between the lines you should read the actual lines first? This is made using the map editor, no third party programs and as such cannot be used on ladder.
It's a video promoting displaying the capabilities of AI and automated actions. I am still giving the OP the benefit of doubt but what hptheGreat said was also the first thing that crossed my mind.
On April 09 2011 04:08 hpTheGreat wrote: I find it interesting that these AI hacks are appearing on the ladder now. I also find it hard to believe that this was just a little weekend project that you and your buddy threw together for giggles.
Just saying. Read between the lines...
Maybe instead of reading between the lines you should read the actual lines first? This is made using the map editor, no third party programs and as such cannot be used on ladder.
It's a video promoting displaying the capabilities of AI and automated actions. I am still giving the OP the benefit of doubt but what hptheGreat said was also the first thing that crossed my mind.
Advertising doesn't have to be specific.
I'm convinced that there are micro bots used on ladder and there are even threads on Bnet SC2 forums full of people noticing something strange is going on: http://us.battle.net/sc2/en/forum/topic/2325612728
Whoever is stupid enough to run them will probably get banned promptly. And yet, the sceptic I am, thought of OP being an advertisement never crossed my mind. If you two were able to really 'read between the lines', you would see that in the OP a moderator even added more Youtube goodness. Which means what? No foul play.
i love the marine vs baneling video the most. this kind of illustrates a point i've been trying to make that banelings are very easy to micro effectively but they are very limited. they are essentially micro capped while the skill ceiling on marines is limitless which can be very frustrating as a zerg player.
I'd like to see the automation bot affecting both the baneling player, and the marine player, the baneling split i think could have been more effective in an ideal situation.
Sentries vs Roaches! How powerful are perfect forcefields. Vikings? Kiting and/or landing/lifting micro Banshee cloaking shenanigans? Using banshee to uncloak/cloak to eat shots? Ravens? Perfect Seeker missiles or perfectly kiting them. Blink Stalkers. Muta micro. Splitting for Thors, attack and fade vs marines. Colossus + Void Prism? Drop play in general. The ling AI with banelings would be pretty cool. Crazy stuff with Phoenix? Perfect 3 beam forever Void rays. Killing DTs with no detection?
On April 09 2011 23:10 Benjef wrote: Mind = blown. If someone could something like that marine micro it would be godly.
MKP comes scarily close with his marine micro in a game on Shakuras vs. KyrixZenith. Excerpts of the micro in first part of this vid: + Show Spoiler +
Also, IMMVP does pretty well, as illustrated in the vid. MKP does move back into the blings again to attack them, something I didn't see IMMVP do (from memory, so I might be wrong). Once the mechanics are there, and pros have practiced enough, I think we will see Automaton-ish practices, at least regarding the marines. The sling-splitting is just insane, but maybe we can learn something from it regardless.
Concerning the 18k APM... Every single marine on the map is constantly polling and adjusting it's distance to other marines, to nearby banelings and simultaneously kiting. 18k APM, that's 300 APS, for 18 marines, so almost 17 actions per second per marine. Anyone can see that a human can come very close to executing this (say, with losing 5 marines) with substantially less actions per marine per second (but still insane APM for humans though). Mechanics like patrolsplitting and magic-boxing for marines are only the start to 'tricks' which can lower the APM requirements for automaton-ish miracles.
This might be stating the obvious, but I'm pretty optimistic about the skill ceiling of SC2 .
Vs the Tanks you can use a small force of Zergling to run in, just say 6 with a big enough distance to each other, then run your big mob after them.
Most the time that will do the job allready, not perfect, but its good enough to have a big gain from this.
DropShip Actions should be crazy with allmost anything.
A Szenario could be:
A DropShip with 2 Marauders is moving around, drops them on the fly, the Marauders "stim" to shoot and run in the moving Dropship (or don't need stim at all).
Doing that with Marines should work aswell.
*With bunkers that should work too, unload Marines, before the melee/projectile hits, back in the bunker. Body Block Melee with the Marines so they can't touch the bunker. Should be totally impossible for humans, but crazy cool for an AI.
Perfect roach with Burrow against Stalkers with Blink would be spectacular. Don't even know who would win actually.
I think Roaches should win, as they can pre-emptive go in with burrow and have no cooldown on burrow/unburrow ; they also regenerate to 100% and not just shield.
*But yes, pretty much everything would be just plain cool to watch, and the outcome of a Perfect AI is pretty much unpredictable.
On April 07 2011 18:01 Aequos wrote: On the Malicious Stalkers - apparently, if you blink perfectly, you can dodge projectile style attacks (making them do 0 damage). I'd like to see 20 or so Blink Stalkers take on a 30-40 stimmed marauder force.
This is a great idea. I second this Nothing is impossible, when you use the Automaton 2000. Brilliant job by the way!! The marine split was especially mind blowing to watch.
Another idea for you would be to micro cracking vs zealots attacking and pulling away before the zealot can swipe for damage. Don't know how possible this is..
This idea sounds fun ( Stalker vs Stimmed Rauders ).
Also : Kill an Ultralist with Drones/Probes/SCVs by avoiding the Splash damage with the help of Automaton :D
On April 08 2011 02:29 Elean wrote: In the zerglings vs tanks, it's not micro, your bot is cheating by using the info of which ling is targetted. Even with perfect micro, it is not possible to achieve with normal information.
Marine vs bannelings is different, it really shows the limit of the unit. But to be fair, the bannelings sould also be controlled by a bot.
Is it really THAT hard to read? It has been stated multiple times that it doesn't cheat to see which ling is being targeted. It just predicts it and according to the OP it is rather easy to do so.
Would be very interesting to see what could this BOT do with a limited amount of APM. I mean if you would give this bot an APM limit, one could get a lot of insight into important parts of micro. For example: could the program achieve same survival ratio with 250 APM?
I assume that right now the program just does what it has to do in order to achieve the best possible results, but if it had to cut unnecessary actions or less important actions to achieve same results, it could potentially find the key actions required for micro. It could even find something that SC2 Pro's could incorporate in their game.
All in all, I am super curious what could this BOT achieve with low APM.
On April 10 2011 22:36 Katagiri wrote: Would be very interesting to see what could this BOT do with a limited amount of APM. I mean if you would give this bot an APM limit, one could get a lot of insight into important parts of micro. For example: could the program achieve same survival ratio with 250 APM?
I assume that right now the program just does what it has to do in order to achieve the best possible results, but if it had to cut unnecessary actions or less important actions to achieve same results, it could potentially find the key actions required for micro. It could even find something that SC2 Pro's could incorporate in their game.
All in all, I am super curious what could this BOT achieve with low APM.
I'm also curious to see how a bot would react with limited APM or in simple scenarios encountered frequently in game such as the probe/zealot/stalker encounter in a PvP. I feel like we could learn a lot by watching a micro bot.
It's be cool to see the bot use perfect FF's. MC force field donuts anyone??
On April 10 2011 22:36 Katagiri wrote: Would be very interesting to see what could this BOT do with a limited amount of APM. I mean if you would give this bot an APM limit, one could get a lot of insight into important parts of micro. For example: could the program achieve same survival ratio with 250 APM?
I assume that right now the program just does what it has to do in order to achieve the best possible results, but if it had to cut unnecessary actions or less important actions to achieve same results, it could potentially find the key actions required for micro. It could even find something that SC2 Pro's could incorporate in their game.
All in all, I am super curious what could this BOT achieve with low APM.
I'm also curious to see how a bot would react with limited APM or in simple scenarios encountered frequently in game such as the probe/zealot/stalker encounter in a PvP. I feel like we could learn a lot by watching a micro bot.
I agree here 100%. Playing out scenarios, with a variety of APMs could produce some very interesting results.
On April 10 2011 22:36 Katagiri wrote: Would be very interesting to see what could this BOT do with a limited amount of APM. I mean if you would give this bot an APM limit, one could get a lot of insight into important parts of micro. For example: could the program achieve same survival ratio with 250 APM?
I assume that right now the program just does what it has to do in order to achieve the best possible results, but if it had to cut unnecessary actions or less important actions to achieve same results, it could potentially find the key actions required for micro. It could even find something that SC2 Pro's could incorporate in their game.
All in all, I am super curious what could this BOT achieve with low APM.
I'm also curious to see how a bot would react with limited APM or in simple scenarios encountered frequently in game such as the probe/zealot/stalker encounter in a PvP. I feel like we could learn a lot by watching a micro bot.
I agree here 100%. Playing out scenarios, with a variety of APMs could produce some very interesting results.
Might be hard to teach the ai how to priorize the commands.
Without priorization the ai would look plenty stupid with 250 apm probably.
Although not a starcraft pro I can give you some insight into the problem of limiting apm from an AI coders perspective ..
The bot at the moment can simply read in the targets of the tanks (or work them out itself) and then generate the list of doomed lings, and simply send a move command to all of the nearby lings to move away.
It moves not only the lings under the splash effect, but also the lings around those lings to make space. Essentially issuing move / attack commands to all of the lings multiple times a second.
If you limit the APM, you are constraining the output of the bot to a much smaller number of commands, so instead of just executing the logically - deduced 'perfect micro' you have to create instead an 'as good a micro as I can with only x100 APM'. This kind of optimisation problem is considerably harder than a solving problem. For example comparing the aiming of a gun in a fps (point at the head correcting for movement) to playing a game of chess (choosing a move given a large amount of possibilities and consequences) gives you an idea in the jump in complexity.
In order to perform the calculations necessary to reduce the perfect tank micro to a list of just some 200 move commands per second the simulations would need to be done incredibly quickly and is probably way out of the realms of possibility of galaxy code running on a high end machine / small cluster.
Even just deducing the minimum apm - if you have a tank hitting ~15 zerglings you need 14 move commands just to save them from the splash, assuming perfect planning that is another 14 moves to make room for them. You then need to correct all of these moves so they go back to trying to hit tanks, so you are already up to 56 move+attack commands to dodge one tank splash. assuming a 'sensible' apm limit of 500 you can only avoid a couple of tank shots a second assuming that EVERY ling not directly affected by the splash is given no commands at all.
It might be possible for a apm-limited bot to perfectly micro against a small number of tanks (3-4) but any higher and you have to accept a non-perfect (i.e. optimised) set of instructions, and you run into the computation complexity problems mentioned above.
I hope this wasn't too geeky but a few people seemed interested in the AI side of the consequences of limiting, so I thought I would share
On April 11 2011 01:11 MrCeeJ wrote: Even just deducing the minimum apm - if you have a tank hitting ~15 zerglings you need 14 move commands just to save them from the splash, assuming perfect planning that is another 14 moves to make room for them. You then need to correct all of these moves so they go back to trying to hit tanks, so you are already up to 56 move+attack commands to dodge one tank splash. assuming a 'sensible' apm limit of 500 you can only avoid a couple of tank shots a second assuming that EVERY ling not directly affected by the splash is given no commands at all.
To be fair you could have the AI move a group of lings as one "action", which would mimic (to some extent) the human ability to box select groups of units.
So even with an apm limit it would get pretty good spread on the zerglings with minimal actions, more apm would allow smaller groups of units to be given the same command scaling to each zergling individually getting it's own separate move command as the apm limit rises.
It would give some variance and artificial noise to the spreads, still un-humanly fast but with reasonable flaws.
i'm thinking pull a spanishiwa style and have mass queens transfusing vs something buff. perhaps voids? or is that alrdy one sided in queens favor? maybe carriers then :D
Excellent first post, a very well edited video showing that maybe there is no skill cap to sc2 at least micro wise.
On related to op note, burrow micro roaches would be interesting, while we commonly see blink micro, burrow micro tends to be neglected, I would like to see it!
OMFG, that Marine micro was freaking crazy. The zergling thing not possible at all as a play will never know what is going to be targeted but the marine might be what future koreans are :D.
The dropship control actually isn't too difficult vs melee units/roaches. Obviously you can't do it quite as well as the AI, but a medivac/tank or warpprism/immortal or colossi is pretty easy. I actually believe there is a replay somewhere of a prime member doing it, but I can't recall
While watching those Zerglings jump away from Tank splash was probably the most beautiful thing I've ever seen, this worries me somewhat...how long till these micro bots are all over the ladder at the lower leagues?
On April 11 2011 11:03 Ridiculisk wrote: While watching those Zerglings jump away from Tank splash was probably the most beautiful thing I've ever seen, this worries me somewhat...how long till these micro bots are all over the ladder at the lower leagues?
Its an AI built into the map. SC2 Map editor and nothing else.
This is pretty nice to see nice job guys seriously But if I see this king of automation in a ladder game I will kick your ass so hard that you wont be able to play SC2 again
Be carefull with this king of cheating guys. As long it stay off-ladder that ok This king of things seem to be more and more used over SC2 and I really hope that Blizzard will have the software so track those macro/soft
On April 11 2011 11:18 MadCow911 wrote: This is pretty nice to see nice job guys seriously But if I see this king of automation in a ladder game I will kick your ass so hard that you wont be able to play SC2 again
Be carefull with this king of cheating guys. As long it stay off-ladder that ok This king of things seem to be more and more used over SC2 and I really hope that Blizzard will have the software so track those macro/soft
On April 11 2011 11:18 MadCow911 wrote: This is pretty nice to see nice job guys seriously But if I see this king of automation in a ladder game I will kick your ass so hard that you wont be able to play SC2 again
Be carefull with this king of cheating guys. As long it stay off-ladder that ok This king of things seem to be more and more used over SC2 and I really hope that Blizzard will have the software so track those macro/soft
On April 11 2011 12:15 Jotoco wrote: I would like to see responses from pro-players that said SC2 skill ceilling was too low...
This proves it is sky high, and probably impossible to ever reach.
Lets not be too critical of the pros though, the Zergling and tank one seems near impossible to do as it requires the player to separate all the other lings from the one being targeted. The drop micro was nicely done! Hope to see part 2 soon.
So skill ceiling is high because with 15000 apm every ranged unit would be unstoppable? Its not really an argument...
I mean, it really is a perfectly valid argument... Because you agree that this 15000 "ceiling" is basically impossible to achieve for a human, it is essentially as if there is no skill ceiling at all for humans.
On April 08 2011 02:29 Elean wrote: In the zerglings vs tanks, it's not micro, your bot is cheating by using the info of which ling is targetted. Even with perfect micro, it is not possible to achieve with normal information.
Marine vs bannelings is different, it really shows the limit of the unit. But to be fair, the bannelings sould also be controlled by a bot.
Is it really THAT hard to read? It has been stated multiple times that it doesn't cheat to see which ling is being targeted. It just predicts it and according to the OP it is rather easy to do so.
Their AI predicts assuming a bunch of Tanks that are on Hold Position (i.e. automatic targeting and/or targeting closest clumps of units).
Granted, targeting the most dense groups of units would theoretically be optimal, but it's also incredibly obvious. A "randomized" targeting among, say, the upper 80% of "ling groupings" is much more difficult to predict, effectively becoming more optimal in practice.
Obviously, if you know how the Tank AI works, you can program accordingly to counter it >.>
Wow that was pretty interesting. I wonder if we'll ever see mini versions of the zergling type micro in pro games. We already have mini versions of the marine micro, but staggering lings and microing them to minimize splash is something I could actually see being implemented by gosu players.
On April 07 2011 21:56 Rassy wrote: Wow thats amazing Shows how much potential to improve is still untapped
dont agree with the muta vid btw If the mutas are smart i am sure they can kill the phoenix on that map Since the phoenix is allowed godly control, then why are the mutas not?
mutas should spread all across the map aproaching the phoenix from all directions so it cant keep running tailing 50 mutas -.-
You can't put two machines up against each other...there'd be a crapton of variables since each AI has a ton of "if...then.." in it.
So skill ceiling is high because with 15000 apm every ranged unit would be unstoppable? Its not really an argument...
That video demonstrates optimal patrol micro. I don't think pro players use that much or have practiced doing it for star2 yet. You can do that with banshees.
i love your videos! keep up the good work! i think its not that long until we see dropmicro like that in a real game. (maybe by tankking.prime.we or GoOdy :D in GSL 27) artosis and tasteless would go crazy
That was SICK to say the least, I can't believe I've been ignoring this thread for so long.
I would love to see this bot put up in a show match or something of the like against some pros. Kind of like a supercomputer that plays against chess grandmasters.
I hope you make the full AI and battle other known AIs like SKYNET, Berkeley Overmind and Behemoth AI.
P.S.: To all haters thinking this is a bot, please learn how to read and understand what you are reading before wasting a couple of keystrokes of ignorance. And also, get a life.
Banshees vs marines, reapers vs roaches, stalkers vs roaches and vikings vs corruptors. Great script, it's really amazing, I hope in a few years players will be able to pull that off.
It pisses me off seeing so many people that have blatantly not even read just the OP, let alone any of the thread. And how many of those who have been corrected have even commented to say "woops, my mistake, I guess I need to read properly in future"? They don't even revisit the thread after throwing their misguided and ignorant opinion around! GRRR.
Great job OP, the micro is really stunning to watch and I'm really looking forward to more vids. I love that you've mixed in footage from tournaments and 'cinematics' etc, rather than just showing the bare bot micro - it really adds a LOT to the videos. Have sent a link to these to my friend, who I fully expect is going to 'ZOMG' too!
On April 07 2011 17:49 blid wrote: If you feel you have a scenario that requires a ridiculous amount of micro, please let us know.
Aside from the obvious "ultra blink stalkers", I'd like to see a big group of JUST infestors VS a big mixed army from Z/T/P. I'd love to see them kiting with perfect fungal growths while NPing special units (and perhaps using their abilities too? Storms etc).
I would love to see perfect blink stalker micro in which the stalkers never take any damage (blink before shot hits) OR in which stalkers blink away for splash (tanks). Then again, just like everyone else is saying you could just blink them before their shield are gone. Be like Pro Protoss players blink micro, but better.
Something that would amaze me would be viking micro against carriers. I'm not sure if it's possible, but it could happen with your programming power!
If you can manage some kind of creative hydra micro off creep, I would be amazed. At the very least, you could try something like the zergling vs tank bot, and split them so that only one ever gets hit. It's obviously harder for hydras though...
I'm just going to say this...amazing work with the automaton micro Off topic: I also recommend a 1v4....Automaton vs NaDa, July, MC and MVP. Would make for an epic match.
I'd love to see scenarios with 2 Micro Bots going at it against each other. Like create 2 situations where it would appear to be a 50/50 even battle and just watch the most insane 2 way micro. like blink stalker/pheonix/Sentry/Colosssi micro vs. Roach/hydra/Corrupter/bling drops or something. Not sure if that is possible.
That would be wild to watch.
These are very cool, thanks for doing them looking forward to the future.
On April 11 2011 22:02 FLuE wrote: I'd love to see scenarios with 2 Micro Bots going at it against each other. Like create 2 situations where it would appear to be a 50/50 even battle and just watch the most insane 2 way micro. like blink stalker/pheonix/Sentry/Colosssi micro vs. Roach/hydra/Corrupter/bling drops or something. Not sure if that is possible.
That would be wild to watch.
These are very cool, thanks for doing them looking forward to the future.
I'd like to see a Blink Stalker v Blink Stalker battle just see what the outcome would be.
I'd love to see an AI programmed to execute perfect speedling surrounds to moving helions.
I can almost picture 6 lings running in in a hexagon formation and crashin down on the helion all at once. To up the difficulty the helion could be moving around randomly. You could test it on and off creep.
On April 11 2011 22:02 FLuE wrote: I'd love to see scenarios with 2 Micro Bots going at it against each other. Like create 2 situations where it would appear to be a 50/50 even battle and just watch the most insane 2 way micro. like blink stalker/pheonix/Sentry/Colosssi micro vs. Roach/hydra/Corrupter/bling drops or something. Not sure if that is possible.
That would be wild to watch.
These are very cool, thanks for doing them looking forward to the future.
I'd like to see a Blink Stalker v Blink Stalker battle just see what the outcome would be.
On April 11 2011 22:02 FLuE wrote: I'd love to see scenarios with 2 Micro Bots going at it against each other. Like create 2 situations where it would appear to be a 50/50 even battle and just watch the most insane 2 way micro. like blink stalker/pheonix/Sentry/Colosssi micro vs. Roach/hydra/Corrupter/bling drops or something. Not sure if that is possible.
That would be wild to watch.
These are very cool, thanks for doing them looking forward to the future.
I'd like to see a Blink Stalker v Blink Stalker battle just see what the outcome would be.
On April 07 2011 18:08 RoyalCheese wrote: I agree that this is kind of cheating, still cool to watch, though.
what do you mean its *kind of cheating*? it's not possible to use it in ladder (or even in any other custom games other than the one its made for).... and if it was possible to use it in regular games, it wouldn't be *kind of* cheating, it would definitely be cheating.
lolz talking zerglings micro is easier than marine micro therefore terran is harder race! lolz jaykay zerg <3 I play random and toss is just too easy with a + move!
Not sure if I like the seige tank drop one. Every zerg knows they should be trying to stand underneath the medivacs, instead you let them retarget the nearest ground army and end up in retargeting hell.
I'd like to see some zergling vs zealot action. By baiting the zealot and attacking with another zergling I think you could have some good results there.
On April 12 2011 02:31 ZeaL. wrote: 50 stalkers vs 100 marauders. Blink while the grenades are in the air and dont get hit!
Awesome Idea.
On a different note, the drop video was AMAZING! I'm actually curious, does picking up a unit negate the cooldown on their attacks?
No, if that were true the medivacs would drop the tanks instantly, creating something like infinite damage, the key is during the cooldown the tank is protected in the medivac.
On April 12 2011 01:29 Mulletarian wrote: oh damn, the medivac tank micro. . . should have had thors and Flight of the Valkyries playing, would have been glorious!
Thors have a significant delay before attacking. Tanks are much faster.
I'd like to see marine split(video 1) WHILE doing the dropship micro with tanks(video 3) against a 300+ supply worth of banelings/roaches ^.^ THUMBS UP IF U AGREE!
On April 12 2011 01:29 Mulletarian wrote: oh damn, the medivac tank micro. . . should have had thors and Flight of the Valkyries playing, would have been glorious!
Thors have a significant delay before attacking. Tanks are much faster.
Flight of the Valkyries, however, is still needed for the video! haha
does picking up at the correct time avoid projectile damage like blink? never had a chance to see it as they were picked up every time a roach got near.
On April 12 2011 07:00 turdburgler wrote: does picking up at the correct time avoid projectile damage link blink? never had a chance to see it as they were picked up every time a roach got near.
yes it does. Same thing with other stuff like burrow (except burrow has a long cast time so it doesn't happen easily)
The part 2 of the medivac micro should have a camera on the faces of the players, so we can see their reactions and hear their swearing while automaton2000 tears them a new one
Is it possible to do the same zergling split micro, but against, say... 5 Colossuses, or like 10 archons or even vs banelings? Or even ultralisk/planetary fortress/hellion splash? Very interesting videos
Hmm...although I guess this could be one of the reasons why the reaver is no longer in SC2. It seems that there's practically no delay between when the unit is dropped and when the unit fires. Reavers would be (almost) OP in SC2 with smarter targeting AI along with the near-instant firing. Shuttles that fire scarabs anyone? :[
On April 12 2011 02:31 ZeaL. wrote: 50 stalkers vs 100 marauders. Blink while the grenades are in the air and dont get hit!
They will all be on blink CD very quickly and get torn apart
maybe roach burrow micro to burrow and heal whenever they hit below a certain HP, though that isn't too too spectacular.
Hydras or marines vs mass colossus would be interesting to see.
Or to include both together, do ling+bling vs siege tank+marine and see how that ends up.
Marauders only shoot every 1 second on stim. Assuming they auto-acquire the nearest target, and acquire a maximum of 5 stalkers at any given moment (which can be arranged with positioning) they'll be fine.
On April 08 2011 05:32 Rettet181 wrote: The marines were fantastic. Would perfecetly-microed banelings even be able to do any damage against perfectly microed marines?
pretty sure not.
banelings are slower then marines :>
If the banelings can trap the marines in a confined space it could be possible.
On April 14 2011 08:08 TheOnlyOne wrote: Drop Shop Action is something you could reliable see in a Pro Match.
Warp-Prismn with an Immortal is pretty good aswell; especialy as the Immortal regs shield if it should somehow get a hit.
No ground hitting unit can beat the DropShip combos, which is allready a pretty big point.
*Maybe Blizzard will fix this with a "delay" or "cooldown" on unit loading ; until then rape as much DropShip Action as you can.
Are you actually suggesting a nerf? Anyone with the insane skills and APM to pull something like that off deserves it. Seriously l2p and stop asking for nerfs everytime someone does something awesome.
On April 14 2011 08:08 TheOnlyOne wrote: Drop Shop Action is something you could reliable see in a Pro Match.
Warp-Prismn with an Immortal is pretty good aswell; especialy as the Immortal regs shield if it should somehow get a hit.
No ground hitting unit can beat the DropShip combos, which is allready a pretty big point.
*Maybe Blizzard will fix this with a "delay" or "cooldown" on unit loading ; until then rape as much DropShip Action as you can.
Are you actually suggesting a nerf? Anyone with the insane skills and APM to pull something like that off deserves it. Seriously l2p and stop asking for nerfs everytime someone does something awesome.
Well to be fair, that was the case in BW and Blizzard 10 years ago had no problem with implementing a delay between when a unit was dropped and when it started firing.
Admittedly, the reasoning behind that change was different. In BW, it was done because prior to that patch, certain pros were so good at reaver-shuttle micro it was essentially as if the shuttles were firing scarabs (which do massive aoe damage on hit if it hits).
Even after the nerf/change, dropship play was still very possible and pros still used incredible dropship micro. It was just a little harder and more risky than it was before.
On April 14 2011 08:08 TheOnlyOne wrote: Drop Shop Action is something you could reliable see in a Pro Match.
Warp-Prismn with an Immortal is pretty good aswell; especialy as the Immortal regs shield if it should somehow get a hit.
No ground hitting unit can beat the DropShip combos, which is allready a pretty big point.
*Maybe Blizzard will fix this with a "delay" or "cooldown" on unit loading ; until then rape as much DropShip Action as you can.
Are you actually suggesting a nerf? Anyone with the insane skills and APM to pull something like that off deserves it. Seriously l2p and stop asking for nerfs everytime someone does something awesome.
Well to be fair, that was the case in BW and Blizzard 10 years ago had no problem with implementing a delay between when a unit was dropped and when it started firing.
Admittedly, the reasoning behind that change was different. In BW, it was done because prior to that patch, certain pros were so good at reaver-shuttle micro it was essentially as if the shuttles were firing scarabs (which do massive aoe damage on hit if it hits).
Even after the nerf/change, dropship play was still very possible and pros still used incredible dropship micro. It was just a little harder and more risky than it was before.
Just so u know, this is completely unit dependant. Go try to "thor drop micro" and you will see it's not as simple as "units fire immediately after unload".
I would like to see what kind of optimum can be achieved in terms of roach burrow play.
I heard of kb macro (not as in the play style, but as in programmable sequences) players abusing the unborrow -> attack -> burrow -> move of roach burrow play.
As you are asking for ideas, I think this might be worth demonstrating this with the Automaton2000 if interesting results are found in experimentation?
Love the videos, keep it up!
(Plus a Tasteless2000 doing an optimum Tasteless build would be great. I can't think of anything for the Artosis2000 to do sadly... Hmmm..)
Why are people making such a big deal out of an unimpressive display of hard scripting ai... Not like its hard to do...
Before you call me troll, I research ai swarm control for uavs as a job... and those videos don't even show the slightest hint of an applied evolutionary algorithm, neural network or any sort of real intelligence.
The author has done nothing but showing how to reduce a huge problem space domain right down to an almost linear problem. Not to mention they use a dumber than shit melee ai (a crappy state machine) to control the enemy...
Show me video where you display real tactical intelligence with a mixture of different units, taking into consideration the terrain, and an actual human playing as the enemy.
As for people saying blizzard should hire them.. Don't make me laugh. This work won't even cut it as a submission to a C ranked ai conference.
On April 14 2011 19:48 Kalafina wrote: Why are people making such a big deal out of an unimpressive display of hard scripting ai... Not like its hard to do...
Before you call me troll, I research ai swarm control for uavs as a job... and those videos don't even show the slightest hint of an applied evolutionary algorithm, neural network or any sort of real intelligence.
The author has done nothing but showing how to reduce a huge problem space domain right down to an almost linear problem. Not to mention they use a dumber than shit melee ai (a crappy state machine) to control the enemy...
Show me video where you display real tactical intelligence with a mixture of different units, taking into consideration the terrain, and an actual human playing as the enemy.
As for people saying blizzard should hire them.. Don't make me laugh. This work won't even cut it as a submission to a C ranked ai conference.
:D dude you should quit matrix.. like seriously neural network? real inteligence ? haha i want to smoke your stash man
On April 14 2011 19:48 Kalafina wrote: Why are people making such a big deal out of an unimpressive display of hard scripting ai... Not like its hard to do...
Before you call me troll, I research ai swarm control for uavs as a job... and those videos don't even show the slightest hint of an applied evolutionary algorithm, neural network or any sort of real intelligence.
The author has done nothing but showing how to reduce a huge problem space domain right down to an almost linear problem. Not to mention they use a dumber than shit melee ai (a crappy state machine) to control the enemy...
Show me video where you display real tactical intelligence with a mixture of different units, taking into consideration the terrain, and an actual human playing as the enemy.
As for people saying blizzard should hire them.. Don't make me laugh. This work won't even cut it as a submission to a C ranked ai conference.
Troll.
I don't really care whether what you say is true or not, there's no reason to point this out the way you do.
Nor do I care whether or not a monkey could do this, the videos were entertaining and that's what matters to me.
OT: Keep it up, looking forward to part 2 of that last video!
Why? Because it looks cool, it's entertaining, it's presented in a nice way, it's exploring the game mechanics somewhat. I hope they continue testing out different things so that we can learn from it and maybe one day see humans execute similar tactics. Maybe not on the scale the machine does it but it sure has potential.
Love it, hope you guys will continue experimenting, feels like you can pioneer some interesting tactics using the Automaton.
On April 14 2011 19:48 Kalafina wrote: Why are people making such a big deal out of an unimpressive display of hard scripting ai... Not like its hard to do...
Before you call me troll, I research ai swarm control for uavs as a job... and those videos don't even show the slightest hint of an applied evolutionary algorithm, neural network or any sort of real intelligence.
The author has done nothing but showing how to reduce a huge problem space domain right down to an almost linear problem. Not to mention they use a dumber than shit melee ai (a crappy state machine) to control the enemy...
Show me video where you display real tactical intelligence with a mixture of different units, taking into consideration the terrain, and an actual human playing as the enemy.
As for people saying blizzard should hire them.. Don't make me laugh. This work won't even cut it as a submission to a C ranked ai conference.
ooh jealous little basement geek coming out to cry his bitterness. why don't you do it? no one else has done this or at least published so how could anyone compare it to anything.
On April 14 2011 19:48 Kalafina wrote: Why are people making such a big deal out of an unimpressive display of hard scripting ai... Not like its hard to do...
Before you call me troll, I research ai swarm control for uavs as a job... and those videos don't even show the slightest hint of an applied evolutionary algorithm, neural network or any sort of real intelligence.
The author has done nothing but showing how to reduce a huge problem space domain right down to an almost linear problem. Not to mention they use a dumber than shit melee ai (a crappy state machine) to control the enemy...
Show me video where you display real tactical intelligence with a mixture of different units, taking into consideration the terrain, and an actual human playing as the enemy.
As for people saying blizzard should hire them.. Don't make me laugh. This work won't even cut it as a submission to a C ranked ai conference.
The work they did was never related to AI-research, your criticism misses the target by a mile.
On April 14 2011 19:48 Kalafina wrote: Why are people making such a big deal out of an unimpressive display of hard scripting ai... Not like its hard to do...
Before you call me troll, I research ai swarm control for uavs as a job... and those videos don't even show the slightest hint of an applied evolutionary algorithm, neural network or any sort of real intelligence.
The author has done nothing but showing how to reduce a huge problem space domain right down to an almost linear problem. Not to mention they use a dumber than shit melee ai (a crappy state machine) to control the enemy...
Show me video where you display real tactical intelligence with a mixture of different units, taking into consideration the terrain, and an actual human playing as the enemy.
As for people saying blizzard should hire them.. Don't make me laugh. This work won't even cut it as a submission to a C ranked ai conference.
You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
Here is a quote for you
"It is not the critic who counts: not the man who points out how the strong man stumbles or where the doer of deeds could have done better. The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood, who strives valiantly, who errs and comes up short again and again, because there is no effort without error or shortcoming, but who knows the great enthusiasms, the great devotions, who spends himself for a worthy cause; who, at the best, knows, in the end, the triumph of high achievement, and who, at the worst, if he fails, at least he fails while daring greatly, so that his place shall never be with those cold and timid souls who knew neither victory nor defeat."
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
On April 14 2011 19:48 Kalafina wrote: Why are people making such a big deal out of an unimpressive display of hard scripting ai... Not like its hard to do...
Before you call me troll, I research ai swarm control for uavs as a job... and those videos don't even show the slightest hint of an applied evolutionary algorithm, neural network or any sort of real intelligence.
The author has done nothing but showing how to reduce a huge problem space domain right down to an almost linear problem. Not to mention they use a dumber than shit melee ai (a crappy state machine) to control the enemy...
Show me video where you display real tactical intelligence with a mixture of different units, taking into consideration the terrain, and an actual human playing as the enemy.
As for people saying blizzard should hire them.. Don't make me laugh. This work won't even cut it as a submission to a C ranked ai conference.
Their videos also show no insight into new research in the field of micro biology or nanotech advancements. You know why? Because that's not the point of it.
The point of these videos is to show (I'm guessing) what you are "capable" of doing in the game. In a sense this shows the unreachable roof for micro techniques that no one will likely every reach.
But you're very skilled at missing the point completely. I'd ask you if your fancy UAVs can make a good cup of coffee but I'm sure they can't because you didn't build them for that purpase.
On April 14 2011 19:48 Kalafina wrote: Why are people making such a big deal out of an unimpressive display of hard scripting ai... Not like its hard to do...
Before you call me troll, I research ai swarm control for uavs as a job... and those videos don't even show the slightest hint of an applied evolutionary algorithm, neural network or any sort of real intelligence.
The author has done nothing but showing how to reduce a huge problem space domain right down to an almost linear problem. Not to mention they use a dumber than shit melee ai (a crappy state machine) to control the enemy...
Show me video where you display real tactical intelligence with a mixture of different units, taking into consideration the terrain, and an actual human playing as the enemy.
As for people saying blizzard should hire them.. Don't make me laugh. This work won't even cut it as a submission to a C ranked ai conference.
:D dude you should quit matrix.. like seriously neural network? real inteligence ? haha i want to smoke your stash man
Neural networks in computer science are very real, usefull and different from what ScienceFiction uses. You can google/wikipedia them if you are intrested.
And regarding Kalafina's post, people do this kind of AI scripting with a mapeditor just for fun, because its fun. If it gives the community some moments of awesomeness its worth it. I think majority of people recognise that the AI's used in games are not really "intellegent" or really learning because its just not cost efficient to use that much time in them or viable to implement in a game.
Why are people making such a big deal out of an unimpressive display of hard scripting ai... Not like its hard to do...
Before you call me troll, I research ai swarm control for uavs as a job... and those videos don't even show the slightest hint of an applied evolutionary algorithm, neural network or any sort of real intelligence.
The author has done nothing but showing how to reduce a huge problem space domain right down to an almost linear problem. Not to mention they use a dumber than shit melee ai (a crappy state machine) to control the enemy...
Show me video where you display real tactical intelligence with a mixture of different units, taking into consideration the terrain, and an actual human playing as the enemy.
As for people saying blizzard should hire them.. Don't make me laugh. This work won't even cut it as a submission to a C ranked ai conference.
This might be one of the greatest posts ever. INTERNETZ and SC 2 IS SERIOUS BUSINESS GUYS!
If I ever run a D ranked AI conference you guys will def. be invited. Keep up the good work lets see some more videos! Fun stuff.
Also, reducing a huge problem space down to an almost linear one is pretty cool imo.
Something cool would be roach vs Thor micro, the ai could maybe analyze the thor's unit animation and burrow the targeted roach right before the Thor fires, forcing it to change targets. Idk if this would work but it would be cool.
On April 14 2011 19:48 Kalafina wrote: Why are people making such a big deal out of an unimpressive display of hard scripting ai... Not like its hard to do...
Before you call me troll, I research ai swarm control for uavs as a job... and those videos don't even show the slightest hint of an applied evolutionary algorithm, neural network or any sort of real intelligence.
The author has done nothing but showing how to reduce a huge problem space domain right down to an almost linear problem. Not to mention they use a dumber than shit melee ai (a crappy state machine) to control the enemy...
Show me video where you display real tactical intelligence with a mixture of different units, taking into consideration the terrain, and an actual human playing as the enemy.
As for people saying blizzard should hire them.. Don't make me laugh. This work won't even cut it as a submission to a C ranked ai conference.
Yes it is not some high level AI project. So what ? I am not looking for good AI algorithms here, nor was this attempt IMHO to create AI player that can compete with real players. What they did is cool because it shows what can be achieved with SC2 units with close to unlimited APM. Do not ridicule something because it does what authors intended and not what you intended. And yes some people may exaggerate what this can do, but such is internet, it is no reason to post angry messages.
On April 14 2011 19:48 Kalafina wrote: Why are people making such a big deal out of an unimpressive display of hard scripting ai... Not like its hard to do...
Before you call me troll, I research ai swarm control for uavs as a job... and those videos don't even show the slightest hint of an applied evolutionary algorithm, neural network or any sort of real intelligence.
The author has done nothing but showing how to reduce a huge problem space domain right down to an almost linear problem. Not to mention they use a dumber than shit melee ai (a crappy state machine) to control the enemy...
Show me video where you display real tactical intelligence with a mixture of different units, taking into consideration the terrain, and an actual human playing as the enemy.
As for people saying blizzard should hire them.. Don't make me laugh. This work won't even cut it as a submission to a C ranked ai conference.
Best first post ever. It's impossible to gauge the ego that you have, I can't even comprehend how large it must be!
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
While it is true that the coding of this script is not particularly difficult, the fact remains that it was done, and that it looks pretty cool.
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
Show me video where you display real tactical intelligence with a mixture of different units, taking into consideration the terrain, and an actual human playing as the enemy.
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
The ego on this guy is astounding.
Ignore the ego - he's completely missing a) the point of the microbot, and b) the point of everyone telling him the point of the microbot.
It's just there to show what can, theoretically, be achieved with the units with through the roof apm. A theoretical maximum to challenge your ability against. You don't need to make supergenius code to do that. We don't need predator to micro lings against tanks - the micro bot does it well enough. It's humanly impossible, but theoretically doable - that's the point of it.
make the bot be able to win the phoenix vs hydra matchup in mirco tournament (custom map) :D (anyone who has ever played this one as toss knows like me that its impossible to win lol).
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
Wow. You made a bot that micros worse than a human, and one that can send workers to separate mineral patches. Absolutely astounding. Truly, your work must be destroyed before it becomes self-aware.
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
I'm sorry but after the pretension of the post you made I was expecting something at least marginally impressive from the videos you linked. Pretty underwhelming performance compared to the OP regardless of implementation differences.
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
You really seem to be missing the entire point of this post/thread. Perfected AIs, while fun, are NOT what this guy is doing, he is instead creating situational response algorithms that can then be used as part of a larger whole. Simply look at the old BWAPI AI competitions, most of it revolved around using 'impossible micro tricks' along side some fairly basic AI build orders/responses.
Get off your high horse, your videos are very very basic compared to what the OP is showing, either come back when you have more to show for yourself so we can take your claims seriously, or stop trying to belittle others achievements for the hell of it.
There's no way this guy (kalafina) isn't trolling. I just lost it during the videos.
AI videos are pretty awesome. Reminds me of a guy who made similar impossible feats w/ AI in SSBM. Releasing a bunch of them at once in a flashy video would be sick. Alternatively complete the campaign using a combination of the AI as fast as possible.
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
All you have proven is that you have less skill than the OP. Thanks for the verification.
You guys are a bit too harsh on Kalafina cause he actually knows what he's talking about. He's right in that Automaton 2000 is purely a hard coded script. It's all done in Galaxy editor and ONLY works in this finite context. If you put things in another map/scenario it could possibly break the script entirely.
Kalafina on the other hand has a LIVE implementation of bots,, The worker split video is garbage but the marine vs zealot video is magnificent. You have to understand that all Kalafina's bot has to work with is exactly what is seen on the screen and has to execute commands through the actual UI just like a "human". You can't do simultaneous actions and there are delays built in. I've never used Galaxy editor or know what Kalafina is doing, so I may be totally off, but an example could be spam warping 100 zealots all at once. In Galaxy, this should be able to be done simultaneously, all at once. Kalafina on the other hand needs to act like a human. He needs to actually move the mouse to the location, press W, press Z, Mouseclick. Galaxy warp(zealot, x_pos, y_pos)
vs. UI Bot mousePos(x_pos, y_pos) delay Z keyClick(w) delay Z keyClick(z) delay Z mousClick delay Z keyClick()
Now of course the delay might not even be noticeable in many cases, but it will still exists nonetheless.
I'm not trying to download Automaton 2000 in anyway. I think it's done an excellent job thus far and the videos are really entertaining. But Kalafina also has something with just as much (if not more) effects to the community. Give him some slack
On April 14 2011 19:48 Kalafina wrote: Why are people making such a big deal out of an unimpressive display of hard scripting ai... Not like its hard to do...
Before you call me troll, I research ai swarm control for uavs as a job... and those videos don't even show the slightest hint of an applied evolutionary algorithm, neural network or any sort of real intelligence.
The author has done nothing but showing how to reduce a huge problem space domain right down to an almost linear problem. Not to mention they use a dumber than shit melee ai (a crappy state machine) to control the enemy...
Show me video where you display real tactical intelligence with a mixture of different units, taking into consideration the terrain, and an actual human playing as the enemy.
As for people saying blizzard should hire them.. Don't make me laugh. This work won't even cut it as a submission to a C ranked ai conference.
On April 15 2011 03:13 kagemucha wrote: You guys are a bit too harsh on Kalafina cause he actually knows what he's talking about. He's right in that Automaton 2000 is purely a hard coded script. It's all done in Galaxy editor and ONLY works in this finite context. If you put things in another map/scenario it could possibly break the script entirely.
Kalafina on the other hand has a LIVE implementation of bots,, The worker split video is garbage but the marine vs zealot video is magnificent. You have to understand that all Kalafina's bot has to work with is exactly what is seen on the screen and has to execute commands through the actual UI just like a "human". You can't do simultaneous actions and there are delays built in. I've never used Galaxy editor or know what Kalafina is doing, so I may be totally off, but an example could be spam warping 100 zealots all at once. In Galaxy, this should be able to be done simultaneously, all at once. Kalafina on the other hand needs to act like a human. He needs to actually move the mouse to the location, press W, press Z, Mouseclick. Galaxy warp(zealot, x_pos, y_pos)
vs. UI Bot mousePos(x_pos, y_pos) delay Z keyClick(w) delay Z keyClick(z) delay Z mousClick delay Z keyClick()
Now of course the delay might not even be noticeable in many cases, but it will still exists nonetheless.
I'm not trying to download Automaton 2000 in anyway. I think it's done an excellent job thus far and the videos are really entertaining. But Kalafina also has something with just as much (if not more) effects to the community. Give him some slack
No one tracked him down and started calling him an idiot. No one found where he lived, tied him to his computer desk and forced him to post in this thread.
Unfortunately*, when you go around being extremely negative and calling people noobs, occasionally someone will be harsh back to you in response.
It doesn't matter if "he actually knows what he's talking about." He came into this thread and decided to call the OP an idiot with no skill (and calling everyone who disagreed with him "ignorant" and himself "elite".) It's somehow surprising that some people (myself included) decided to criticize his work in response?
On April 15 2011 03:13 kagemucha wrote: Give him some slack
You don't walk into other people's threads, acting high and mighty about your accomplishments which are completely unrelated to the conversation at hand.
On April 15 2011 03:13 kagemucha wrote: You guys are a bit too harsh on Kalafina cause he actually knows what he's talking about. He's right in that Automaton 2000 is purely a hard coded script. It's all done in Galaxy editor and ONLY works in this finite context.
The point is that everything you mentioned above was also clearly posted by the OP, they never claimed for it to be anything more than a series of videos demonstrating scripted events within the galaxy editor.
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
well you are actually ignorant if you think this thread is about praising the op's coding ability, think a little, it helps. notice how most of the comments are something along the lines "wow looks cool/awesome". yeah LOOKS.
On April 15 2011 03:13 kagemucha wrote: You guys are a bit too harsh on Kalafina cause he actually knows what he's talking about. He's right in that Automaton 2000 is purely a hard coded script. It's all done in Galaxy editor and ONLY works in this finite context. If you put things in another map/scenario it could possibly break the script entirely.
Kalafina on the other hand has a LIVE implementation of bots,, The worker split video is garbage but the marine vs zealot video is magnificent. You have to understand that all Kalafina's bot has to work with is exactly what is seen on the screen and has to execute commands through the actual UI just like a "human". You can't do simultaneous actions and there are delays built in. I've never used Galaxy editor or know what Kalafina is doing, so I may be totally off, but an example could be spam warping 100 zealots all at once. In Galaxy, this should be able to be done simultaneously, all at once. Kalafina on the other hand needs to act like a human. He needs to actually move the mouse to the location, press W, press Z, Mouseclick. Galaxy warp(zealot, x_pos, y_pos)
vs. UI Bot mousePos(x_pos, y_pos) delay Z keyClick(w) delay Z keyClick(z) delay Z mousClick delay Z keyClick()
Now of course the delay might not even be noticeable in many cases, but it will still exists nonetheless.
I'm not trying to download Automaton 2000 in anyway. I think it's done an excellent job thus far and the videos are really entertaining. But Kalafina also has something with just as much (if not more) effects to the community. Give him some slack
Dude did you read his posts, he's a douchebag. Even if he was the most skilled AI coder in the world, why does that matter if he's a massive dbag.
On April 15 2011 03:13 kagemucha wrote: You guys are a bit too harsh on Kalafina cause he actually knows what he's talking about. He's right in that Automaton 2000 is purely a hard coded script. It's all done in Galaxy editor and ONLY works in this finite context. If you put things in another map/scenario it could possibly break the script entirely.
No one said it wasn't a hard coded script.
Kalafina on the other hand has a LIVE implementation of bots,, The worker split video is garbage but the marine vs zealot video is magnificent. You have to understand that all Kalafina's bot has to work with is exactly what is seen on the screen and has to execute commands through the actual UI just like a "human". You can't do simultaneous actions and there are delays built in. I've never used Galaxy editor or know what Kalafina is doing, so I may be totally off, but an example could be spam warping 100 zealots all at once. In Galaxy, this should be able to be done simultaneously, all at once. Kalafina on the other hand needs to act like a human. He needs to actually move the mouse to the location, press W, press Z, Mouseclick.
Sure he's not using the editors scripting and is instead working through the SC2 interface which would be much more difficult; however, his attitude in this thread is that he thinks that the algorithms used to control the units are poor and stupid. I'm not claiming they're complicated but he then posts a link to a video in which he does the exact same thing but worse. The marine vs zealot micro is no less situational and limited than the marine vs baneling micro and I don't really see how you can believe differently. Furthermore, the micro bot he linked to didn't do a great job, it was pretty painfully slow (probably algorithm inefficiency) and lost units which to me indicates that it doesn't even do a very limited thing very well.
I'm not trying to download Automaton 2000 in anyway. I think it's done an excellent job thus far and the videos are really entertaining. But Kalafina also has something with just as much (if not more) effects to the community. Give him some slack
It's cool you can download teh automaton 2000, with the power of the Internet we can all share it.
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
All you have proven is that you have less skill than the OP. Thanks for the verification.
I do not mean to burst your bubble. But what this guy has done is way more advanced than the one posted in the OP.
Because this is a program that reads the screen, then makes actions based on scenarios that it interprets by itself.
The OPs bot is one that uses the AI scripting, as in it works within the bubble of the application. Therefore everything that the bot needs to know is redily available to it.
This bot actually reads picture data off of a screen and makes mouse movements and keystrokes.
In otherwords. This bot could be used in ladder, where the OPs could not.
Although the OPs bot is pulling off much more advanced micro, this guys bot is doing much more advanced calculations.
I do hate the elitest attitude though.
edit: Everything I just said was already said, my bad.
On April 15 2011 03:13 kagemucha wrote: You guys are a bit too harsh on Kalafina cause he actually knows what he's talking about. He's right in that Automaton 2000 is purely a hard coded script. It's all done in Galaxy editor and ONLY works in this finite context. If you put things in another map/scenario it could possibly break the script entirely.
Kalafina on the other hand has a LIVE implementation of bots,, The worker split video is garbage but the marine vs zealot video is magnificent. You have to understand that all Kalafina's bot has to work with is exactly what is seen on the screen and has to execute commands through the actual UI just like a "human". You can't do simultaneous actions and there are delays built in. I've never used Galaxy editor or know what Kalafina is doing, so I may be totally off, but an example could be spam warping 100 zealots all at once. In Galaxy, this should be able to be done simultaneously, all at once. Kalafina on the other hand needs to act like a human. He needs to actually move the mouse to the location, press W, press Z, Mouseclick. Galaxy warp(zealot, x_pos, y_pos)
vs. UI Bot mousePos(x_pos, y_pos) delay Z keyClick(w) delay Z keyClick(z) delay Z mousClick delay Z keyClick()
Now of course the delay might not even be noticeable in many cases, but it will still exists nonetheless.
I'm not trying to download Automaton 2000 in anyway. I think it's done an excellent job thus far and the videos are really entertaining. But Kalafina also has something with just as much (if not more) effects to the community. Give him some slack
Dude...you realize that Starcraft AIs have been in the works for years now right? There have been regular BW AI competitions that work with actual bots not the map editor and every single one of those BW AI/bots are much more impressive than the shit Kalafina posted. Not to mention those bots worked with a shittier engine as opposed to the SC2 engine.
In all honesty, we're hard on Kalafina 'cause he made an account to come into a thread and talk shit about the OP when the OP himself said that his work isn't AI and is simply map editor. Add on to the fact that his micro bot is complete garbage (how the hell did one of the marines die??? A human can micro better than that) all culminates in him getting same treatment he gave the OP/TL.
EDIT: I'm not saying the stuff Kalafina did isn't impressive or hard to do. It definitely is. But it's not nearly as impressive as you think it is and it definitely isn't so impressive as to warrant his douchebaggery. Go look up some BW AIs (not just the wraith micro video, but actual AIs that can play a game from beginning to end).
On April 15 2011 05:00 TheNikeYork wrote: Wow that tank dropping vs that sunken was amazing.
That was really impressive, i want to see the dropping against stalkers or marauders so that the laser/bullet hits where the tank was, while the tank is being lifted up into the air. Would be so sweet to watch.
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
All you have proven is that you have less skill than the OP. Thanks for the verification.
I do not mean to burst your bubble. But what this guy has done is way more advanced than the one posted in the OP.
Because this is a program that reads the screen, then makes actions based on scenarios that it interprets by itself.
The OPs bot is one that uses the AI scripting, as in it works within the bubble of the application. Therefore everything that the bot needs to know is redily available to it.
This bot actually reads picture data off of a screen and makes mouse movements and keystrokes.
In otherwords. This bot could be used in ladder, where the OPs could not.
Although the OPs bot is pulling off much more advanced micro, this guys bot is doing much more advanced calculations.
I do hate the elitest attitude though.
edit: Everything I just said was already said, my bad.
I don't care how sophisticated his application is. He demonstrated a poor AI. It is a worse solution than OP's because it performed worse at its task. It doesn't matter how "advanced" it is if it sucks.
Reading the screen is an obstacle that is unnecessary and unrelated to the problem of creating a game playing AI. I don't congratulate someone that hammered a nail into some wood using a rube goldberg machine. I facepalm and say "why the hell are you doing this the hard way"?
edit: So I reread Kalafina's first post and it now seems like a clear case of not feeling recognized at his job. He see's OP's post and immediately thinks "wtf I do shit more complex then this all day long and nobody cares, bawww!" So, I kinda feel sorry for him now. That's assuming that anything he posted is true, I guess.
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
All you have proven is that you have less skill than the OP. Thanks for the verification.
I do not mean to burst your bubble. But what this guy has done is way more advanced than the one posted in the OP.
Because this is a program that reads the screen, then makes actions based on scenarios that it interprets by itself.
The OPs bot is one that uses the AI scripting, as in it works within the bubble of the application. Therefore everything that the bot needs to know is redily available to it.
This bot actually reads picture data off of a screen and makes mouse movements and keystrokes.
In otherwords. This bot could be used in ladder, where the OPs could not.
Although the OPs bot is pulling off much more advanced micro, this guys bot is doing much more advanced calculations.
I do hate the elitest attitude though.
edit: Everything I just said was already said, my bad.
I don't care how sophisticated his application is. He demonstrated a poor AI. It is a worse solution than OP's because it performed worse at its task. It doesn't matter how "advanced" it is if it sucks.
Reading the screen is an obstacle that is unnecessary and unrelated to the problem of creating a game playing AI. I don't congratulate someone that hammered a nail into some wood using a rube goldberg machine. I facepalm and say "why the hell are you doing this the hard way"?
edit: So I reread Kalafina's first post and it now seems like a clear case of not feeling recognized at his job. He see's OP's post and immediately thinks "wtf I do shit more complex then this all day long and nobody cares, bawww!" So, I kinda feel sorry for him now. That's assuming that anything he posted is true, I guess.
Why feel sorry for someone that acts like that? He could be the best programmer in the world but with that attitude he won't get any credit or fans anyway.
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
All you have proven is that you have less skill than the OP. Thanks for the verification.
I do not mean to burst your bubble. But what this guy has done is way more advanced than the one posted in the OP.
Because this is a program that reads the screen, then makes actions based on scenarios that it interprets by itself.
The OPs bot is one that uses the AI scripting, as in it works within the bubble of the application. Therefore everything that the bot needs to know is redily available to it.
This bot actually reads picture data off of a screen and makes mouse movements and keystrokes.
In otherwords. This bot could be used in ladder, where the OPs could not.
Although the OPs bot is pulling off much more advanced micro, this guys bot is doing much more advanced calculations.
I do hate the elitest attitude though.
edit: Everything I just said was already said, my bad.
I don't care how sophisticated his application is. He demonstrated a poor AI. It is a worse solution than OP's because it performed worse at its task. It doesn't matter how "advanced" it is if it sucks.
Reading the screen is an obstacle that is unnecessary and unrelated to the problem of creating a game playing AI. I don't congratulate someone that hammered a nail into some wood using a rube goldberg machine. I facepalm and say "why the hell are you doing this the hard way"?
edit: So I reread Kalafina's first post and it now seems like a clear case of not feeling recognized at his job. He see's OP's post and immediately thinks "wtf I do shit more complex then this all day long and nobody cares, bawww!" So, I kinda feel sorry for him now. That's assuming that anything he posted is true, I guess.
My point was that the OPs bot doesnt demonstrate what a human could ever hope to do. It demonstrates how cool it looks when a game plays against itself.
The other guys bot simulates taking sensory info and making decisions. Which shows what is possible for a human to do.
edit: retracted insult... the internet is pissing me off today.
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
All you have proven is that you have less skill than the OP. Thanks for the verification.
I do not mean to burst your bubble. But what this guy has done is way more advanced than the one posted in the OP.
Because this is a program that reads the screen, then makes actions based on scenarios that it interprets by itself.
The OPs bot is one that uses the AI scripting, as in it works within the bubble of the application. Therefore everything that the bot needs to know is redily available to it.
This bot actually reads picture data off of a screen and makes mouse movements and keystrokes.
In otherwords. This bot could be used in ladder, where the OPs could not.
Although the OPs bot is pulling off much more advanced micro, this guys bot is doing much more advanced calculations.
I do hate the elitest attitude though.
edit: Everything I just said was already said, my bad.
I don't care how sophisticated his application is. He demonstrated a poor AI. It is a worse solution than OP's because it performed worse at its task. It doesn't matter how "advanced" it is if it sucks.
Reading the screen is an obstacle that is unnecessary and unrelated to the problem of creating a game playing AI. I don't congratulate someone that hammered a nail into some wood using a rube goldberg machine. I facepalm and say "why the hell are you doing this the hard way"?
edit: So I reread Kalafina's first post and it now seems like a clear case of not feeling recognized at his job. He see's OP's post and immediately thinks "wtf I do shit more complex then this all day long and nobody cares, bawww!" So, I kinda feel sorry for him now. That's assuming that anything he posted is true, I guess.
My point was that the OPs bot doesnt demonstrate what a human could ever hope to do. It demonstrates how cool it looks when a game plays against itself.
The other guys bot simulates taking sensory info and making decisions. Which shows what is possible for a human to do.
edit: retracted insult... the internet is pissing me off today.
I still disagree with you. A screen reading AI will still have infinite potential to be better than any human because it can click the mouse in the exact location and press keys in the correct order instantaneously. Humans will have a fixed upper limit for how fast their fingers work, how fast they can perform an action after interpreting what they see. Any AI can include a forced limit, it doesn't have to be reading the screen to do it.
There are two problems being solved: Interpret input and determine action. Interpret input can be solved by screen reading or by hooking into the game itself. It is my opinion that screen reading is utterly fucking useless because an objectively better set of input is available to you.
On April 15 2011 06:00 artanis2 wrote: There are two problems being solved: Interpret input and determine action. Interpret input can be solved by screen reading or by hooking into the game itself. It is my opinion that screen reading is utterly fucking useless because an objectively better set of input is available to you.
Not useless, it's harder for blizzard to detect it
On April 15 2011 03:13 kagemucha wrote: You guys are a bit too harsh on Kalafina cause he actually knows what he's talking about. He's right in that Automaton 2000 is purely a hard coded script. It's all done in Galaxy editor and ONLY works in this finite context. If you put things in another map/scenario it could possibly break the script entirely.
Kalafina on the other hand has a LIVE implementation of bots,, The worker split video is garbage but the marine vs zealot video is magnificent. You have to understand that all Kalafina's bot has to work with is exactly what is seen on the screen and has to execute commands through the actual UI just like a "human". You can't do simultaneous actions and there are delays built in. I've never used Galaxy editor or know what Kalafina is doing, so I may be totally off, but an example could be spam warping 100 zealots all at once. In Galaxy, this should be able to be done simultaneously, all at once. Kalafina on the other hand needs to act like a human. He needs to actually move the mouse to the location, press W, press Z, Mouseclick. Galaxy warp(zealot, x_pos, y_pos)
vs. UI Bot mousePos(x_pos, y_pos) delay Z keyClick(w) delay Z keyClick(z) delay Z mousClick delay Z keyClick()
Now of course the delay might not even be noticeable in many cases, but it will still exists nonetheless.
I'm not trying to download Automaton 2000 in anyway. I think it's done an excellent job thus far and the videos are really entertaining. But Kalafina also has something with just as much (if not more) effects to the community. Give him some slack
Dude...you realize that Starcraft AIs have been in the works for years now right? There have been regular BW AI competitions that work with actual bots not the map editor and every single one of those BW AI/bots are much more impressive than the shit Kalafina posted. Not to mention those bots worked with a shittier engine as opposed to the SC2 engine.
In all honesty, we're hard on Kalafina 'cause he made an account to come into a thread and talk shit about the OP when the OP himself said that his work isn't AI and is simply map editor. Add on to the fact that his micro bot is complete garbage (how the hell did one of the marines die??? A human can micro better than that) all culminates in him getting same treatment he gave the OP/TL.
EDIT: I'm not saying the stuff Kalafina did isn't impressive or hard to do. It definitely is. But it's not nearly as impressive as you think it is and it definitely isn't so impressive as to warrant his douchebaggery. Go look up some BW AIs (not just the wraith micro video, but actual AIs that can play a game from beginning to end).
I agree with you in that he's presenting himself as a dick and the criticism for that is perfectly understandable. I just don't think the people saying his marine micro video is terrible compared to Automaton's is correct.
I am aware of the BWAPI stuff that went on during BW days. It was definitely amazing how good those AIs were and I don't think any SC2 AI right now can even come close to them (not even Green Tea AI. Although I haven't checked out Green Tea since the beta so who knows). I'm sure with more time, SC2 AI will catch up and surpass BW simply because theres so much more power in Galaxy Editor.
But you're wrong in comparing BW bots to SC2 UI Bots (Kalafina's). The bots in BW that used the BWAPI (which most did) are much more similar to the Galaxy Editor scripts than SC2 UI bots. You can still do simultaneous commands with no delay amongs other days. You aren't necessary limited to what you see on the screen. In fact, there is no "screen" in BW bots, it's all commands.
In the end though, like many have said already, these are just simply two different types of AI that is being developed in Starcraft 2 right now. Neither is "better" than the other, they're just different.
On April 15 2011 03:13 kagemucha wrote: You guys are a bit too harsh on Kalafina cause he actually knows what he's talking about. He's right in that Automaton 2000 is purely a hard coded script. It's all done in Galaxy editor and ONLY works in this finite context. If you put things in another map/scenario it could possibly break the script entirely.
Kalafina on the other hand has a LIVE implementation of bots,, The worker split video is garbage but the marine vs zealot video is magnificent. You have to understand that all Kalafina's bot has to work with is exactly what is seen on the screen and has to execute commands through the actual UI just like a "human". You can't do simultaneous actions and there are delays built in. I've never used Galaxy editor or know what Kalafina is doing, so I may be totally off, but an example could be spam warping 100 zealots all at once. In Galaxy, this should be able to be done simultaneously, all at once. Kalafina on the other hand needs to act like a human. He needs to actually move the mouse to the location, press W, press Z, Mouseclick. Galaxy warp(zealot, x_pos, y_pos)
vs. UI Bot mousePos(x_pos, y_pos) delay Z keyClick(w) delay Z keyClick(z) delay Z mousClick delay Z keyClick()
Now of course the delay might not even be noticeable in many cases, but it will still exists nonetheless.
I'm not trying to download Automaton 2000 in anyway. I think it's done an excellent job thus far and the videos are really entertaining. But Kalafina also has something with just as much (if not more) effects to the community. Give him some slack
Dude...you realize that Starcraft AIs have been in the works for years now right? There have been regular BW AI competitions that work with actual bots not the map editor and every single one of those BW AI/bots are much more impressive than the shit Kalafina posted. Not to mention those bots worked with a shittier engine as opposed to the SC2 engine.
In all honesty, we're hard on Kalafina 'cause he made an account to come into a thread and talk shit about the OP when the OP himself said that his work isn't AI and is simply map editor. Add on to the fact that his micro bot is complete garbage (how the hell did one of the marines die??? A human can micro better than that) all culminates in him getting same treatment he gave the OP/TL.
EDIT: I'm not saying the stuff Kalafina did isn't impressive or hard to do. It definitely is. But it's not nearly as impressive as you think it is and it definitely isn't so impressive as to warrant his douchebaggery. Go look up some BW AIs (not just the wraith micro video, but actual AIs that can play a game from beginning to end).
I agree with you in that he's presenting himself as a dick and the criticism for that is perfectly understandable. I just don't think the people saying his marine micro video is terrible compared to Automaton's is correct.
I am aware of the BWAPI stuff that went on during BW days. It was definitely amazing how good those AIs were and I don't think any SC2 AI right now can even come close to them (not even Green Tea AI. Although I haven't checked out Green Tea since the beta so who knows). I'm sure with more time, SC2 AI will catch up and surpass BW simply because theres so much more power in Galaxy Editor.
But you're wrong in comparing BW bots to SC2 UI Bots (Kalafina's). The bots in BW that used the BWAPI (which most did) are much more similar to the Galaxy Editor scripts than SC2 UI bots. You can still do simultaneous commands with no delay amongs other days. You aren't necessary limited to what you see on the screen. In fact, there is no "screen" in BW bots, it's all commands.
In the end though, like many have said already, these are just simply two different types of AI that is being developed in Starcraft 2 right now. Neither is "better" than the other, they're just different.
Ah, I see waht your trying to say. Yea kalafina bot is indeed different but honesty if you're going to make an ai why not take the most effeicient and effective route? Using screen reading as input seems kind of pointless when scripting is more powerful and can be just as, if not more complex. Only advantage to screen reading is that it's harder for blizzard to detect but that's just gaming
we should stop derailing the thread tho seeing how we agree on the issue that kalafina is a dick lol
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
All you have proven is that you have less skill than the OP. Thanks for the verification.
I do not mean to burst your bubble. But what this guy has done is way more advanced than the one posted in the OP.
Because this is a program that reads the screen, then makes actions based on scenarios that it interprets by itself.
The OPs bot is one that uses the AI scripting, as in it works within the bubble of the application. Therefore everything that the bot needs to know is redily available to it.
This bot actually reads picture data off of a screen and makes mouse movements and keystrokes.
In otherwords. This bot could be used in ladder, where the OPs could not.
Although the OPs bot is pulling off much more advanced micro, this guys bot is doing much more advanced calculations.
I do hate the elitest attitude though.
edit: Everything I just said was already said, my bad.
I don't care how sophisticated his application is. He demonstrated a poor AI. It is a worse solution than OP's because it performed worse at its task. It doesn't matter how "advanced" it is if it sucks.
Reading the screen is an obstacle that is unnecessary and unrelated to the problem of creating a game playing AI. I don't congratulate someone that hammered a nail into some wood using a rube goldberg machine. I facepalm and say "why the hell are you doing this the hard way"?
edit: So I reread Kalafina's first post and it now seems like a clear case of not feeling recognized at his job. He see's OP's post and immediately thinks "wtf I do shit more complex then this all day long and nobody cares, bawww!" So, I kinda feel sorry for him now. That's assuming that anything he posted is true, I guess.
My point was that the OPs bot doesnt demonstrate what a human could ever hope to do. It demonstrates how cool it looks when a game plays against itself.
The other guys bot simulates taking sensory info and making decisions. Which shows what is possible for a human to do.
edit: retracted insult... the internet is pissing me off today.
Here's the thing though, the OP didn't walk in here and say "Hey guys look at this AI that demonstrates what a human can do and look how cool it is, and how hard it was to do, and look at how amazing a programmer I am, praise me!" the OP came and said "Hey guys look at this map I made using the editor that shows some really cool looking micro and notice how I'm not claiming it's anything it's not or that it can do anything it can't I'm just showing you something cool." And you know what? he did what he was aiming to do really damn well, and it looks sick, Kalafina on the other hand came in and said "Hey! I'm doing harder work in my real job than this guy is to make a cool looking thing on a video game, and I attempted something that's harder than what he did even though he never said his was hard just that it looked cool, PRAISE ME!" (I say attempted because he was attempting to make a good marine micro program when it actually would have been mediocre for a human.)
On April 15 2011 06:00 artanis2 wrote: There are two problems being solved: Interpret input and determine action. Interpret input can be solved by screen reading or by hooking into the game itself. It is my opinion that screen reading is utterly fucking useless because an objectively better set of input is available to you.
Not useless, it's harder for blizzard to detect it
That's not true though. Blizz can't detect when another application reads SC2's memory.
Love these videos, bravo, u guys could really make a ton of viewers doing this with all sorts of diff scenarios or even looking at pro games when better micro couldve won someone the game but they failed to be gosu like automoton2K!
I have an idea for you guys, how about a script that would spread Mutalisks against lots and lots of Thors that would dodge the splash perfectly? I assume it's doable and hopefully will look as good, good luck guys.
On April 15 2011 05:00 TheNikeYork wrote: Wow that tank dropping vs that sunken was amazing.
At least it proved hits should always at least land once where they were intended to land.
No, I think it works fine the way it is... first off in a realistic sense, if there is nothing there to hit by the time the hit would land then it shouldn't hit anything, but also if someone has the micro to do something like that without losing out too much on their macro at the same time then more power to them
On April 17 2011 05:30 ObliviousNA wrote: A friend showed this to me, very impressive. I know that collosus damage only occurs after the animation finishes... perhaps blink vs collosus?
I've been using it.
It's freakin' awesome.
The collosi beam will actually follow the blinking stalker (if originally targeted), and this causes a HUUUUUGE distance of travel of the beam, which looks scary at first sight, but later I realized that don't do ANY splash damage any to other unit.
That stalker that got targeted will still get hurt no matter what. (Warp prism anyone?)
Speak of warp prism, it's freakin' amazing against zerg too.
On April 15 2011 05:00 TheNikeYork wrote: Wow that tank dropping vs that sunken was amazing.
At least it proved hits should always at least land once where they were intended to land.
Why you say that? Not to make it realistic, because the whole starcraft and its always-hit system is unreal. You have some analysis to show why its better? I know games (RTS) where every bullet is a moving object on its own which collides or not with its target (or something else), its realistic but im not sure its better as a mechanic.
Can we please see a Ling/Bling/Muta Automaton 2000 micro'd army vs. a Tank/Thor/Marine/Medivac Automaton 2000 micro'd army? That would be the most epic fucking shit ever. Their APM would be over 900000 probably.
On April 14 2011 19:48 Kalafina wrote: Why are people making such a big deal out of an unimpressive display of hard scripting ai... Not like its hard to do...
Before you call me troll, I research ai swarm control for uavs as a job... and those videos don't even show the slightest hint of an applied evolutionary algorithm, neural network or any sort of real intelligence.
The author has done nothing but showing how to reduce a huge problem space domain right down to an almost linear problem. Not to mention they use a dumber than shit melee ai (a crappy state machine) to control the enemy...
Show me video where you display real tactical intelligence with a mixture of different units, taking into consideration the terrain, and an actual human playing as the enemy.
As for people saying blizzard should hire them.. Don't make me laugh. This work won't even cut it as a submission to a C ranked ai conference.
While I agree that the author has done nothing clever, what you're saying is like if someone put together a fireworks display and someone decided to criticize it by saying that those fireworks would never cut it at a rocketry conference.
Then again, you've provoked so many people who have no idea what they're talking about to reveal how little they know what they're talking about so you've done good.
Though, count the number of "OOH I WANT TO SEE AUTOMATON VS. AUTOMATON" posts when it's clear there's nothing intelligent going on in the background (automaton vs. automaton isn't actually possible for multiple reasons, including the fact that the automaton can only play one side of each "matchup" -- there's no automaton 2000 for tanks microing against zerglings or banelings against marines" and it's clear that some people think that this is a really clever breakthrough in AI so I can definitely see how you would get frustrated.
And yes, anyone who says the OP's work is more impressive is wrong. It's definitely more entertaining and visually pleasing, and it's what I would ooh and ahh at.
I was wondering, is it possible to use infinite creep tumor placement/canceling to draw away enemy fire. I was thinking something like constantly placing creep tumors in front of lings so that tanks target it and then canceling it to avoid losing the tumor.
In my opinion, the best thing about the videos is that you chose Dream Theater as backing music for the zergling micro. My favorite game of all time + my favorite band of all time = so much win it's uncanny!
This was awesome. Not sure why some rocket surgeon had to come in here and talk about programming and AI and shit. That's not the point. The point was, as anyone who even watched the videos would know, to show that Zerglings can (in theory) rape siege tanks. It's clearly not doable by humans... (if it is it'd be the greatest play in the history of SC2 at least) but it's something to go "oh wow, that's badass!".
Hell, even the first sentence "push the potential of the in-game units to the limits.". That's exactly what was shown!
Thank God that rocket surgeon stopped at 2 posts. What a crock.
I'd be really interested to see how well humans could do in such a situation though. Very interesting....
On April 20 2011 09:47 Goibon wrote: This was awesome. Not sure why some rocket surgeon had to come in here and talk about programming and AI and shit. That's not the point. The point was, as anyone who even watched the videos would know, to show that Zerglings can (in theory) rape siege tanks. It's clearly not doable by humans... (if it is it'd be the greatest play in the history of SC2 at least) but it's something to go "oh wow, that's badass!".
Hell, even the first sentence "push the potential of the in-game units to the limits.". That's exactly what was shown!
Thank God that rocket surgeon stopped at 2 posts. What a crock.
I'd be really interested to see how well humans could do in such a situation though. Very interesting....
I don't think you understand something :/ He's not coming here to talk about 'programming'. I'm sure many users wouldn't understand if he did.
He's simply just showing off what he and his friend are able to do in the editor, something that the average user can't do.
On April 14 2011 19:48 Kalafina wrote: Why are people making such a big deal out of an unimpressive display of hard scripting ai... Not like its hard to do...
Before you call me troll, I research ai swarm control for uavs as a job... and those videos don't even show the slightest hint of an applied evolutionary algorithm, neural network or any sort of real intelligence.
The author has done nothing but showing how to reduce a huge problem space domain right down to an almost linear problem. Not to mention they use a dumber than shit melee ai (a crappy state machine) to control the enemy...
Show me video where you display real tactical intelligence with a mixture of different units, taking into consideration the terrain, and an actual human playing as the enemy.
As for people saying blizzard should hire them.. Don't make me laugh. This work won't even cut it as a submission to a C ranked ai conference.
While I agree that the author has done nothing clever, what you're saying is like if someone put together a fireworks display and someone decided to criticize it by saying that those fireworks would never cut it at a rocketry conference.
Then again, you've provoked so many people who have no idea what they're talking about to reveal how little they know what they're talking about so you've done good.
Though, count the number of "OOH I WANT TO SEE AUTOMATON VS. AUTOMATON" posts when it's clear there's nothing intelligent going on in the background (automaton vs. automaton isn't actually possible for multiple reasons, including the fact that the automaton can only play one side of each "matchup" -- there's no automaton 2000 for tanks microing against zerglings or banelings against marines" and it's clear that some people think that this is a really clever breakthrough in AI so I can definitely see how you would get frustrated.
And yes, anyone who says the OP's work is more impressive is wrong. It's definitely more entertaining and visually pleasing, and it's what I would ooh and ahh at.
Hahaha Kalafina is acting like a child that is super jealous of someone else getting attention. Seriously, the OP did something cool, who gives a shit if it is really technically hard or advanced? I don't know shit about AI but I still think it's cool. Why do you have to be a fun hater and come in "omg you guys are stupid this isn't hard or anything, I could do it". It just makes you look like fucking children. Grow up and stop getting jealous over others' attention.
Guess OP didnt find any pro terran who wanted to be humiliated by the bot But seriously, we need update, please update, at least with the "no you dont get anything nubs"
Well, you know you dont actually need 15k apm to do this. Its just the apm the bot produced. The human mind makes a much more specialized solution, so you could achieve this with several thousand apm, and a less perfect, but still very good is totally possible (just watch that mvp scene in the vods, and we are at the beginning still).
What i dont understand is why we dont see this kind of dropship micro? I saw it once or twice, but not too much. Ok, not 5 dropships at the same time, but 1 or 2, it would be still unbeatable by certain units, and tanks shoot hard even unsieged, same for immortals and colos.
I just watched it for the first time, and... I really have to focus on my zergling micro na but seriously, coll stuff, keep it coming and if you ever need some more ideas:
Reaper vs Roaches (4.5 range vs 4range) phoenix vs mutalisk/infestor blink stalker vs a big fat splash unit like colossi or Ultras burrow roaches (single burrow micro) and maybe update marines vs banelings with the banelings splitting perfectly too
On April 14 2011 22:18 mechavoc wrote: You made an account to post this? kind of sad. Yes we are sorry you didn't think of this first.
OP had both the ability (which you say many people have), the Idea, and the follow through to acomplish this cool project. Having those 3 things is impressive. You on the other hand have only have contributed critizism.
So take this as a challenge go out and make some really cool AI ! show us what you can do and we will give you all the praise you deserve!
It makes me laugh at what low standards this community has. It truely shows the impassable gap between the ignorant and the elite.
"I didn't think of this first"?? Why would I limit myself to something this simple? The SC2 Hacking scene has been pushing the boundaries of bots and ai's since the beginning of the beta. Heard of the Green Tea ai?
"OP had the ability"?? Give a code monkey a book on implementing a state machine and/or fuzzy logic and it can code this within a day. ANYONE can do it.
"Make a cool AI"?? You underestimate me. The caliber of the projects I work on are many times greater than this. If we release these publicly, it'll be exploited by the masses. Proof? Become one of the elites and you maybe able to join our cabal, a true gathering of minds.
I'll leave you with these 2 videos. A taste of our progress from 3 months ago. I have said what I felt needed to be said, and now I shall say no more. Video 1 Video 2
All you have proven is that you have less skill than the OP. Thanks for the verification.
I do not mean to burst your bubble. But what this guy has done is way more advanced than the one posted in the OP.
Because this is a program that reads the screen, then makes actions based on scenarios that it interprets by itself.
The OPs bot is one that uses the AI scripting, as in it works within the bubble of the application. Therefore everything that the bot needs to know is redily available to it.
This bot actually reads picture data off of a screen and makes mouse movements and keystrokes.
In otherwords. This bot could be used in ladder, where the OPs could not.
Although the OPs bot is pulling off much more advanced micro, this guys bot is doing much more advanced calculations.
I do hate the elitest attitude though.
edit: Everything I just said was already said, my bad.
I don't care how sophisticated his application is. He demonstrated a poor AI. It is a worse solution than OP's because it performed worse at its task. It doesn't matter how "advanced" it is if it sucks.
Reading the screen is an obstacle that is unnecessary and unrelated to the problem of creating a game playing AI. I don't congratulate someone that hammered a nail into some wood using a rube goldberg machine. I facepalm and say "why the hell are you doing this the hard way"?
edit: So I reread Kalafina's first post and it now seems like a clear case of not feeling recognized at his job. He see's OP's post and immediately thinks "wtf I do shit more complex then this all day long and nobody cares, bawww!" So, I kinda feel sorry for him now. That's assuming that anything he posted is true, I guess.
Why feel sorry for someone that acts like that? He could be the best programmer in the world but with that attitude he won't get any credit or fans anyway.
I'm a fan of his works. He has some skill in programming and uses it quite well.
You guys should do one about micro against hellions. Keeping units in an arc facing away from hellions (thus negating any splash damage) would be awesome.
On September 12 2011 09:34 Keo wrote: AI is cheating because it knows which unit is being targeted before it is shot.
No it's not they explain somewhere that tanks default to closest units so even from the Z side you could tell, they did change it for this implementation because it looks cooler if it fires in the middle of the pack but they could have done it from Zerg's info only.
well, people think this can be use in a ladder game, and are complainig the russians IN THE SELLING THING, and the black market, is it true that you can use this in a ladder game for 30USD month to play like a GRANDMASTER? THANK YOU VERY MUCH FOR YOUR TIME