|
Catyoul
France2377 Posts
Update: image for terran @ 3 was wrong, corrected now
Goal Mining rates had already been investigated in the past for Terran, thoroughly on a tl.net blog, and also on a gosugamers thread. This work had not been done for Protoss and Zerg to my knowledge.
Curiosity about the different mining rates per race and per patch finally got the best of me and I decided to investigate it thoroughly. No longer will you wonder where to send your first workers for that little extra edge ! Details about the methodology and the results can be found below. For quick reference here is the illustrated guide :
Methodology The map will be Python 1.3. The test games are played in LAN multiplayer on fastest game speed with no opponent. The initial mineral investment (to build additional bases and workers) are gathered at the natural to avoid touching the main minerals. To avoid differences rising from split differences, all nexus, hatchery or command center are first destroyed or lifted. One worker is sent at each patch of the base. The mining resumes synchronously for all workers when the nexus/hatch/cc is complete/landed. All workers are stopped after a certain amount of time. The process is repeated for all maine bases on the map and all races. The whole process is rerun to check for consistency and provide averaging of the results on a longer period of time.
Results As can be seen in the pictures above, the results are similar while not identical for Zerg and Protoss and very different for Terran. This could be attributed to differences in pathing, speed or acceleration of their respective workers.
Below are the detailed results (averaged over several runs, don't be surprised if you see some numbers that are not dividable by 8). Each patch of each base is on a separate line, starting with the top patch (1) to the bottom one (9). For each race, 3 columns are shown : 1. the amount of minerals mined over 10 minutes, 2. the mining rate per second, 3. the comparison of the mining rate for this patch with the average mining rate at this base. More than 100% is higher than average, less is lower (obviously).
After the 9 patches, a line gives the total mining rate (per 10 minutes and per second). You will notice the values are significantly different for each race/base combo. The values are pretty consistent for terran across all bases, but for protoss and zerg, 9 mines slighlty slower than the others. You can also compare the mining rate across races if you so wish.
In the spoiler below are the above results in graphic format :
+ Show Spoiler [Protoss] ++ Show Spoiler [Terran] ++ Show Spoiler [Zerg] +
Now that you have all the results, if you don't like the choices I made for the patches to highlight or the colors in the first images, you can modify them yourselves : Protoss Terran Zerg
Conclusions and perspectives As you can see, the mining rate can be extremely different from one patch to the other. It is really only useful at the very beginning of the game of course, but since you don't have anything else to do but care about your workers, you might as well do it properly. I haven't tested it enough yet to check for consistency, but preliminary results on a 9 pool with gas build at 3 o'clock show over 5 seconds improvement on the time the zerglings hatch when filling the patches from best to worst compared to a more naive bottom to top.
To my knowledge, the path the workers will take (and thus the time they will take to return the minerals) depends on the angle of approach of the patches. It is possible that by microing each worker to attack the minerals from the right angle, you could further improve the mining rate early game and shave more precious seconds off your fine tuned build order.
|
Awesome work man. Some interesting points: - Protoss seems to have the advantage in mining - The base at 3 is imba. Something i was thinking is how to make a script that run in everymap and enables us to make the same analisis in less time? Im thinking about letting the BW IA running some time and then return the data. It could be a mod or something im not too good at starcraft programming but could be nice to have a script or triggers to do it in all the maps.
|
For protoss, at the 9 and 6 o'clock, there are two patches (5th and 7th patches) which are above average. These patches seem to be the same distance away as the 5th patch at the 12 o'clock position, but at 12 it isn't above average. I had assumed these patches would be the same. Is there any explanation that you could offer? Is the pathing so differerent when the probe is mining on the left or right side of the nexus?
Overall great work! thank you
|
konadora
Singapore66060 Posts
I'm surprised at some of the results o_O
I always thought the closest = best, furthest = worst.
|
You should put the nexus hatcherys and comand centres in all the pictures to help orientate people
|
The benefits to be gained from making use of these results are negligible even at the professional level. It seems like a complete waste of time to me.
|
It's interesting how different the Terran results are from those for the Zerg and Protoss. SCV behavior is really strange, they'll hesitate in front of certain patches. This is actually really helpful, since five seconds early game can easily translate into twenty seconds or even more later on. Thanks a lot, Catyoul!
|
On May 09 2009 19:21 Malongo wrote: Awesome work man. Some interesting points: - Protoss seems to have the advantage in mining - The base at 3 is imba. Something i was thinking is how to make a script that run in everymap and enables us to make the same analisis in less time? Im thinking about letting the BW IA running some time and then return the data. It could be a mod or something im not too good at starcraft programming but could be nice to have a script or triggers to do it in all the maps.
Base 3 i in fact not imba
The color of a patch shows the mining rate relative to the base, not the absolute mining rate. If you look at the sum of minerals mined over 10 minutes, it is almost exactly the same for the different bases for each race.
Btw. great job by the OP. Super easy to read and understand with the layout and all. Interesting read
|
Your result for Terran @3 is different than the result here, who should I believe?
|
excellent work on gathering and sorting out all the info, have been looking for stuff like this for quite a while, i'm ust too lazy to do all these by myself
|
|
this is awesome and its imo showing how BW is great... if u watch bw u can see that all economy rules works here this game is so deep.. i miss fact that i cant follow that BW lecture at that university in usa
|
Smix
United States4549 Posts
Wow, thank you so much for this. Nicely organized and incredibly easy to follow.
|
On May 09 2009 19:47 Swarmy wrote: The benefits to be gained from making use of these results are negligible even at the professional level. It seems like a complete waste of time to me.
I don't think anything can be disregarded as such, every little advantage you can get you should take, also its not like this takes much effort, all you have to do is a little memorisation, enact it at the beginning of the match and then your done.
Awesome OP by the way, extreme good quality and clearly presented, your quality post star is well deserved.
|
Thanks so much for making this, I knew the results would be different than with terran!
I honestly can't believe pro gamers don't do this stuff on every new map. It'd be great if there was a database of every map and good mining locations. It makes such a difference for early game rushes, for example.
Some of those mineral patches I really wouldn't've expected, and as a scrub that plays python way too often it'll help my early game economy to know which patches to split to first.
|
pretty cool ^^
ive always wished sombody would research at which point stop making scvs would be ideal (ignoring other aspects from the game but to have as best mining with as few scvs as possible)
maybe your the man to do the job?
|
Wow, nice work, how long did all this take?
|
Good job mate.
I had done this myself once and the problem I came across was that mining times differ significantly sometimes.
For example iirc the 3rd patch on 3 o'clock for protoss would mine extra fast sometimes and normally/slow some other times. I know this happens for both protoss and terran ( i didn't test zerg) on every starting position on python (the only map I've tested). I am guessing that the reason is the initial path the workers take the first time they mine but this is just speculation.
I remember that if a worker got into "fast mining mode" the difference in mining time was huge, so a more important question for me would be: Can we find the way to cause a worker to take the optimal mining path each time?
|
nice, but I really don't think the little things like this make that big of a difference. It may, if it was a early game rush and micro war between few units, but unless its that short a game, it really won't make that big of a difference
i know some people like to emphasize the importance of such procedures, saying how it is really important to split correctly, set the larvae on the left, etc. but it really doesn't make that much of a difference; they just like to exaggerate
nice nice work though
|
i dont mean to shit on the OP for his greatly executed experiment, but for all the players or newbs who want to take time in studying this, you'd be better off knowing how to scout and defend rushes so these things won't matter much, and studying other things like build orders, counters, and perfecting macro
|
On May 10 2009 05:49 SlayerS_`HackeR` wrote: i dont mean to shit on the OP for his greatly executed experiment, but for all the players or newbs who want to take time in studying this, you'd be better off knowing how to scout and defend rushes so these things won't matter much, and studying other things like build orders, counters, and perfecting macro
|
Nice work, Catyoul. Always cool to see someone with a science (or was it math) background apply their analytical skills to BW.
|
Wow, this is really cool. I've messed around with things like this myself, and I know how much effort it takes. Amazing job
|
A pretty interesting study you made here. I always thought the mineral patches closest to the Nexus/CC/hat provide the best mining rate but apparently I've been proved wrong.
Amazing job and thanks for sharing.
|
Great analysis. Hopefully in the future we can get some more fun and current maps .
|
Haha great work op. I'm going to set the zerg one as desktop background Really useful because I play python a lot since I'm noob like that. Once again great work thanks.
|
nice work and effort thanks a lot! very good information?
|
United States3824 Posts
Holy Fucking Christ
Nice job
|
So, I guess it really sucks to be a Protoss at the 3:00 position, eh?
|
Interesting, my splits will now be that much better :D
|
This is really impressive. Great job.
|
Korea (South)3086 Posts
Wow, this is some intense research. Props.
|
On May 09 2009 23:01 -orb- wrote: I honestly can't believe pro gamers don't do this stuff on every new map. It'd be great if there was a database of every map and good mining locations. It makes such a difference for early game rushes, for example.
Pretty sure progamers do this. They may not know exactly which one is the fastest, but they know which ones are relatively fast and which aren't. I also believe even if they don't test this, from mass gaming, they would get a feel for which patch is slow and thus avoid that. (eg for those who play luna, it is very obvious to discover which patch is slow at 11 main).
On May 09 2009 23:59 stet_tcl wrote: Good job mate.
I had done this myself once and the problem I came across was that mining times differ significantly sometimes.
For example iirc the 3rd patch on 3 o'clock for protoss would mine extra fast sometimes and normally/slow some other times. I know this happens for both protoss and terran ( i didn't test zerg) on every starting position on python (the only map I've tested). I am guessing that the reason is the initial path the workers take the first time they mine but this is just speculation.
I remember that if a worker got into "fast mining mode" the difference in mining time was huge, so a more important question for me would be: Can we find the way to cause a worker to take the optimal mining path each time?
The reason why a worker mine the same patch at different rates when you test during different trials usually depends on the angle of which you approach the same mineral. Below is an illustration For the white patch, a probe would actually mine at different speeds based on the angle, but the difference is not too big. For the yellow patch, based on the pathing, you can clearly see one takes a longer route to reach the same patch. The pathing can be manually corrected by forcing the worker to mine at the correct angle. Sometimes the fix is permanent and sometimes it isn't.
But the longer route doesn't always mean it's slower. Usually this applies to terran, where the longer route has less deacceleration than the shorter route therefore a faster mining time.
When you pay attention to these things enough, you can automatically know which patch is fast even if the map is new.
|
United States17042 Posts
Great Job op. Nicely presented, clear format. I'll try to remember it ^^
On May 09 2009 23:37 MorroW[MB] wrote:pretty cool ^^ ive always wished sombody would research at which point stop making scvs would be ideal (ignoring other aspects from the game but to have as best mining with as few scvs as possible) maybe your the man to do the job?
Saturation is different for every race, and it matters more how many mineral patches you have access to, as well as when you're going to choose to cut workers depending on your build. Search, there are several threads on this topic, and the general consensus is that "it depends"
|
wow, i'm glad you did this, but the fact that you missed 3rd-from-bottom mineral for Python 3:00 terran makes me sad. it's one of the few that I know that allows terrans to make "~8.75" supply while not interrupting SCV production at all (mineral wise or supply wise).
now I don't know whether your stuff is trustable or not.
|
Catyoul
France2377 Posts
On May 09 2009 20:07 Garnet wrote:Your result for Terran @3 is different than the result here, who should I believe? Thanks for pointing it out, my numbers were correct but I made a mistake in coloring the minerals (they were shifted one patch to the bottom). Updated version :
On May 09 2009 19:47 Swarmy wrote: The benefits to be gained from making use of these results are negligible even at the professional level. It seems like a complete waste of time to me. Not that I needed a reason to do it other than curiosity, but I believe it might be useful in low-econ games, especially in 2v2.
Also, it might have been a waste of my time, but now that the results are there, it doesn't cost you any to apply it
On May 09 2009 23:01 -orb- wrote: I honestly can't believe pro gamers don't do this stuff on every new map. It'd be great if there was a database of every map and good mining locations. It makes such a difference for early game rushes, for example. I'm 100% sure I have seen progamers actually micro their workers in the very early game, possibly to force them to take faster mining paths on the slow minerals.
On May 09 2009 23:54 TheFoReveRwaR wrote:Wow, nice work, how long did all this take? About 2 days, thank God for windowed SC :D
On May 09 2009 23:59 stet_tcl wrote: Good job mate.
I had done this myself once and the problem I came across was that mining times differ significantly sometimes.
For example iirc the 3rd patch on 3 o'clock for protoss would mine extra fast sometimes and normally/slow some other times. I know this happens for both protoss and terran ( i didn't test zerg) on every starting position on python (the only map I've tested). I am guessing that the reason is the initial path the workers take the first time they mine but this is just speculation.
I remember that if a worker got into "fast mining mode" the difference in mining time was huge, so a more important question for me would be: Can we find the way to cause a worker to take the optimal mining path each time?
I measured every race/base combo several times to check for consistency. Of course I used the same method to send them to the minerals, so the approach angles were similar (though not exactly the same). Still, the results were very consistent, with 2 exceptions. Once the 2nd mineral patch at the 6 base with protoss was very heavily undermined (huge outlier in the data). I believe I might have made a mistake when sending the probe to the patch and miscliked. The other time, the 2nd mineral patch at the 3 base with terran was slighltly undermined, I don't know why, the rest of the patches were pretty consistent. So, weird things happen sometimes, but they're pretty rare.
I don't know if the workers "lock" on a specific mining path according to where you sent them first. Microing the workers to force their path towards the minerals could be done. To confirm my numbers I watched a couple of runs to see for myself why some patches are slow. The path the workers take for those are... let's say they could be optimized :D That would be the subject for a further study.
On May 10 2009 13:10 rwong48 wrote:wow, i'm glad you did this, but the fact that you missed 3rd-from-bottom mineral for Python 3:00 terran makes me sad. it's one of the few that I know that allows terrans to make "~8.75" supply while not interrupting SCV production at all (mineral wise or supply wise). now I don't know whether your stuff is trustable or not. I have my raw numbers for Terran @ 3, done 3 times over 5 minutes, 15 minutes and 15 minutes. 3rd patch from bottom is consistently... average. It hovers between 99.7% and 100.01% compared to the average. I'm sure you'd have better results if you sent an scv to the 3rd patch from top instead. Also, my results for Terran are consistent with the results Dead9 had already shown before. Feel free to check it yourself anyway.
|
United States1865 Posts
Hey just as a support for the OP:
GGplay's winner's interview for the Daum OSL finals, for example, specifically mentions the fact that he thought he was gonna lose Game 5 on Python because his spawn (3:00) was known for gathering less minerals (a fact which correlates with Catyoul's findings).
|
woah dude, the difference is very noticeable.
NERF PROTOSS!
---
Great work btw, must have taken a lot of time <_<
|
I fucking knew terran spawning at 3 sucked balls!
those tho in the middle that are red always tricked me! I never had 104 at 9 SCVs when i split there so I stopped.
im surprised theyre red
|
On May 11 2009 01:11 Atrioc wrote: Hey just as a support for the OP:
GGplay's winner's interview for the Daum OSL finals, for example, specifically mentions the fact that he thought he was gonna lose Game 5 on Python because his spawn (3:00) was known for gathering less minerals (a fact which correlates with Catyoul's findings). Back then, wasn't the old version of Python still being used? The one with completely screwed up pathing. And correlates? According to OP, 3:00 mines the fastest out of all locations for Zerg.
|
Nice work. It seems like you could have saved some effort by modifying the map to give you enough money to build your first nexus/hatchery/command center rather than mining the expo for it.
I think the difference in mining rates is less than the data suggest. For example, zerg at 9 o'clock gets 768 from patch 7 and 592 from patch 3 after 10 minutes. But you won't leave just one worker gathering for that time, you'll start with four and soon have more workers than patches. So what really matters is the mining rate from the top four patches at each base versus the bottom four patches, and even that will only apply until you get extra workers going.
This is still useful for telling you where to send the first workers and it might put you a couple seconds ahead in economy. That doesn't matter if your other skills are weak, but this does give an edge against someone who is otherwise matched with you.
I wonder if Blizzard knew about and intended the difference in mining rates between races. And I wonder if they're doing tests like these on Starcraft 2 to keep it balanced (or intentionally unbalanced in certain aspects).
|
great work! hmm that seems like a great experiment to try out for myself..
i wonder if anyone has thought about testing this on the expansion minerals?
|
On May 09 2009 20:06 EvoChamber wrote: It's interesting how different the Terran results are from those for the Zerg and Protoss. SCV behavior is really strange, they'll hesitate in front of certain patches. This is actually really helpful, since five seconds early game can easily translate into twenty seconds or even more later on. Thanks a lot, Catyoul! seriously
ur post is wonderful.... i will try this right away!
|
Great post man! Don't let all the people saying this is negligible get to you. This is not only more than negligible, but more importantly, incredibly interesting!
I also have to applaud you on your careful and exacting methodology too. Thanks!
|
p so imba. but yeah this is pretty cool.. i fail to see how it improves ones gameplay
|
On May 09 2009 20:07 Garnet wrote:Your result for Terran @3 is different than the result here, who should I believe? Neither, really.
Mining speed has to do with the pathing efficiency of the worker... The angle of return and acceleration/deceleration of workers varies based on building placement and the initial worker placement on the mineral patch. That this is explained in the OP, but not considered or included in the actual test, is confusing to me. The test that would need to be done to truly prove patch mining efficiency is whether alternative initial worker placements vary the worker's traveling speed. If all worker placements result in the same pathing then you know for sure what that patch's mining rate is. Otherwise, results from tests like this will be inconsistent.
Also, to those who say this doesn't make a big difference, I learned this because on a UMS map I've played to death, spawn defense, two of the positions have a single probe that returns slower than the others. If you don't re-angle this probe, after a minute it's five seconds slower than the other probes. This means you don't get enough money in time to send the first wave of units, and as a result will be at such a disadvantage that unless you're a far superior player to your opponent you'll likely lose. Now, say a progamer splits all four to a bad angle. After about a minute of mining, compared to an ideal start, it's like they waited five seconds at the start before starting to mine.
If you can't see how significant that is, you probably don't understand starcraft very well.
|
On May 09 2009 19:38 pRo9aMeR wrote: For protoss, at the 9 and 6 o'clock, there are two patches (5th and 7th patches) which are above average. These patches seem to be the same distance away as the 5th patch at the 12 o'clock position, but at 12 it isn't above average. I had assumed these patches would be the same. Is there any explanation that you could offer? Is the pathing so differerent when the probe is mining on the left or right side of the nexus?
Overall great work! thank you
I read all the posts, and nobody offered an answer...patiently waiting ^^
|
|
|
|