|
On December 08 2012 07:22 Antylamon wrote:EDIT: Show nested quote +On December 07 2012 02:02 SirPsychoMantis wrote:On December 06 2012 21:57 Coldlogic wrote:On December 06 2012 17:04 Nepherius wrote: I've got a question for Coldlogic. It would be amazing if you could see your SQ after a round in TheStaircase.
Have you considered trying to get your map to calculate it or is that too difficult / impossible to do?
I have very little knowledge about programming and no knowledge whatsoever about the map editor, so the following could be horribly wrong.
I'm thinking that if you could write a function to look at someone's minerals every 5 seconds, store the information away and average it at the end of the game, then you'd be able to get the average unspent minerals of a player. Then another function could look at the number of ressources mined by a player (in a x range around a hatch / nexus / cc, count how many minerals and gas are remaining and substract that to what the mineral patches and geyser had at the start of the game), and average the mining rate. And then a third function could process those two variables to calculate the SQ.
Is that even feasible? I actually have looked very deeply into it. It is something that will take a lot of time though. You are close to what I am thinking about doing, the problem is that I have memory constraints to deal with on the map (doing too much stuff can cause the game to lag). My initial thought was to just get the resource collection rate and average unspent from the spectator interface, but obviously that is for the entire game, so the numbers would be horribly skewed by the 2nd round. That is when I realized I had to calculate it myself. So every so often I will need to sample the unspent resources and save that data. The fear around memory constraints comes in here. Say I do it every 5 seconds. That is 12 times per minute. Now, the longest game I have ever played has been around an hour probably. That is a maximum of 720 sampling states I need to make. And that is on a per player basis, so just for a 1v1 map, 1440 sampling states. I have to declare this memory at the start of the game, so I will have an array sitting around that has 1440 integers worth of memory. Resource collection rate, I AM PRAYING, could just be calculated by dividing the amount of resources collected by the duration of the round. If not, something similar to avg unspent would have to be done by sampling the amount of resources and then calculating the changes between them. Ok, now those are the two big numbers needed to calculate SQ. Here is the third giant problem: there is no mathematical function for calculating the natural log of a value in the editor. Not really a HUGE problem, but definitely a problem. I would then have to calculate the Taylor series or an Infinite series and traverse a certain number of steps to determine the natural log by hand (via computer ). Then use that value to calculate SQ. You can imagine my concern with allocating so much memory, half the time the memory wouldn't even be used. I also considered limiting SQ to be 30 minutes or something, to cut the allocation space in half. There are a bunch of factors to consider and JaKaTaK and myself are currently discussing what to do with it. Can't you do a running average rather than saving every data point? Store how many you have collected then every tick average that into the running number? He already said what I have to say. Although @PsychoMantis you don't have to store the amount of numbers recorded, you just have to add the average to the current unspent resources then divide by two. Otherwise you'd quickly end up with numbers which round down to zero. + Show Spoiler +In case you still don't see the problem, this would be your first handful of results with a constant 100 income:
100 (100+100)/2=100 (100+100)/3=33.33 (33.33+100)/4=33.33 (33.33+100)/5=26.67 (26.67+100)/6=21.11 etc. It seems to be what Blizz does as well, considering how quickly statistics like income, APM, etc. update.
This is the formulas you should use to calculate on the fly the average unspent ressources and the collection rate. It only involves storing a limited number of variables so the map should be fine memorywise.
Average unspent ressources:
The formula is ((n-1)*last average + current data) / n = Current average
M=Minerals in the bank G=Gas in the bank A=Last average N=Number of data points taken up to this point
(A*(N-1)+M+G)/N
So assuming you have a constant 50 mineral and 50 gas in the bank, things would go like this.
Point 1: (0*(1-1)+50+50)/1 = (0*0+50+50)/1 = 100 Point 2: (100*(2-1)+50+50)/2 = (100*1+50+50)/2 = 100 Point 3: (100*(3-1)+50+50)/3 = (100*2+50+50)/3 = 100 etc.
The average income rate is something a little bit harder to calculate. I've done a few tests and I don't think that Blizzard simply divides the total ressources mined by the time elapsed during the game. I've came to this conclusion after I've tried numerous time to mine out a single base in minerals. With 8 mineral patches containing 1500 minerals each, a single base contains 12000 minerals to be had. With terran, I can mine out a whole base in less then 12 minutes. When I stop the game at the twelve minute mark, I should theorically have an average income rate of 12000 ressources mined / 12 minutes = 1000 ressources per minute. The number the graph returns me is closer to 1150 ressources per minute. Not good enough.
What I think that Blizzard does to calculate the average income rate is that it takes the income rate you have in spectator mode at a regular time interval, stores lots of data points (i.e. T=0, Collection = 0; T=5, Collection = 20; etc...) and then calculate at the end of the game the area under the curve made by all those points (Remember all those calculus classes? Now is the time to apply what you've learned!). I tried doing that by hand, but again my ressults have been slightly different from what Blizzard gets. I think the discrepancy is due to my time interval being slightly different from the one used by Blizzard. We could test that theory with different time interval, but I'm not willing to do it by hand as it's quite time consumming. Instead, I'll give you the formula to get the average income rate.
Normally, in college classes, you are given a function's formula to calculate the area under the curve. We don't have this luxury. We only have data points.
Fortunately, the area under the curve can be easilly calculated if you divide it in rectangles and triangles. So assuming that x is your time interval, y is the income rate at the last data point and that z is the income rate at the current data point, your rectangle is going to be y high and x wide. The triangle on top of it is going to be z-y high and x wide. Now I,m sure you guys know the formulas to calculate the area of a rectangle or a triangle.
Do the sum of every rectangle and triangles under a curve, and you get the area of the curve.
With this in mind, here we go:
A=Mineral gathering rate at the last data point B=Gas gathering rate at the last data point C=Current mineral gathering rate D=Current gas gathering rate E=Current time (units = seconds) F=The interval of time at which you're taking data points (in seconds) G=The area of the last rectangle + the area of the last triangle
F * (A + B) + (E / 60) * (C + D - A - B) / 2 = G or once simplified (F / 2) * (A + B + C + D) = G
So with this, we've calculated the area of the rectangle and the triangle at time E-F.
Now if we want to do a running average of this, that's the formula we'd have:
The formula is ((n-1)*last average + current data) / n = Current average
When I plug in those variables, it becomes:
A=Mineral gathering rate at the last data point B=Gas gathering rate at the last data point C=Current mineral gathering rate D=Current gas gathering rate E=Current time (units = seconds) F=The interval of time at which you're taking data points (in seconds) G=The area of the last rectangle + the area of the last triangle H=The last average I=Current average
( (E / F - 1) * H + G / F) / (E / F) = I ***
or if we plug in the formula for G
( (E / F - 1) * H + (F / 2) * (A + B + C + D) / F) / (E / F) = I
once simplified
( (E / F - 1) * H + (1 / 2) * (A + B + C + D) ) / (E / F) = I
Then at the next time interval, I becomes H and you run the process all over again.
***You may ask me why do I divide G by F in this second formula. The answer is quite simply I DON'T FREAKING KNOW AND IT'S DRIVING ME NUTS. I've calculated this using three different ways so I know what the answer should be. Using trial and error, I found out that I have to divide G by F in this unified formula to match the results I get from the other formulas I've used, but I still don't get why. If anyone can enlighten me, I'd be glad for the help!
In my testings, my time interval (F) was of 5 seconds. I think that blizzard uses time intervals that are larger then that, but I don't want to bother proving it. I guess you could plug this formula in a test map and see if the results at the end of a game match what Blizzard comes up with in it's graph. Same for the average unspent ressources.
Anyway, I've been on this for long enough now. Hope you find a good way to calculate that darn sq in real time!
|
On December 08 2012 10:38 Nepherius wrote: What I think that Blizzard does to calculate the average income rate is that it takes the income rate you have in spectator mode at a regular time interval, stores lots of data points (i.e. T=0, Collection = 0; T=5, Collection = 20; etc...) and then calculate at the end of the game the area under the curve made by all those points (Remember all those calculus classes? Now is the time to apply what you've learned!) I'm in Pre-Calculus right now, and only just finished the review stuff from Algebra... so I can't help much. T_T
Not being out of high school yet sucks in this regard.
|
Holy shit Nepherius, thanks so much. I too wish I had taken a Calculus class. In high school I took AP Stats instead because that looked more interesting to me, and I was a music major in college, so I had enough on my plate in those years. No better time to learn than the present though, thanks for all of your help
EDIT:
Tutorial Video #5 is now out, Enjoy
EDIT EDIT:
A new section has been added to the OP, please let me know what you think.
Skill/Power Ratio
Balancing for Skill - Extra Credits
I recently watched this video from extra credits about Skill/Power ratio and how it can be beneficial or detrimental to the game depending on how it is set up. They use the example of the zerg rush in the original starcraft (when the spawning pool was 150 minerals instead of 200). Its definitely popped into my head to develop a strategical improvement system that chooses strategies in the order of less Skill/Power ratio to greater Skill/Power ratio, but that will likely be a long ways away. For now, instead of choosing strategies in order of less to greater Skill/Power ratio, I am going to take a closer look at the Skill/Power ratio of each unit, and arrange them accordingly. Its likely that little to nothing will change as this was the idea I had been going with the whole time (although now it is much more clearly defined for me), but if you'd like to help, please submit your list of units from least skill/power ratio (small skill, big power) to most skill/power ratio (big skill, big/bigger power). Thanks.
|
Lasender Step 4
I seem to be having trouble with the SQ system, because whenever I have an average resource collection of over 2000, I'm practically guaranteed to have an SQ of over 120, As a Terran with mules, and doing a base heavy style as I am, I tend to have anywhere between 2000-3000 average resources per minute.
Any thoughts on this, or where I might get a more accurate number?
|
How many bases are you on usually?
The less bases you have, the easier it is to get ridiculous SQ.
|
@Anty I completely disagree, for me, the more bases I have, the easier it is to get ridiculous SQ.
|
|
1 base is so hard to get super high SQ. In my experience.
|
Yeah, I was saying that, as a 4 base terran, i'm going to have an sq above 120, regardless of how much I actually spend, because of how much I'm taking in.
|
On December 10 2012 06:04 JaKaTaK wrote: 1 base is so hard to get super high SQ. In my experience. It's the opposite for me... huh.
Although I haven't 1base'd since near the start of step 1.
|
On December 10 2012 06:58 Lasender wrote: Yeah, I was saying that, as a 4 base terran, i'm going to have an sq above 120, regardless of how much I actually spend, because of how much I'm taking in. If you can keep your money low on 4 bases, you're macroing properly.
Edit: Unless you are massively queuing up in your production structures. Edit 2: An SQ of 120 is off the charts of the original SQ thread, btw. Most pros average < 100, I believe, according to that thread.
|
Russian Federation42 Posts
|
Can I play this on EU somehow?
Looks nice but I want to try it myself! :D
|
On December 10 2012 18:38 vrumFondel wrote:Hi jak! I have a replay of Step1 protoss 1v1 ranked in hots: http://drop.sc/283298 enjoy!
Gah! I just did step 1 protoss. That's the only replay I have! I literally am looking for everything else right now. Thx for the replay though
|
We now have Protoss and Terran step specific videos (Zerg en route).
Protoss
Terran
|
I think these videos are a really good start, but I also think they aren't going to work well as tutorials / as ways of explaining TheStaircase to people. Keep in mind I think the videos are very helpful.
I'll be as specific and direct as possible:
Your narration for the Terran tutorial starts off like this: "Basically what we're going to be doing here is building scvs and gathering minerals.... keeping money low.... what this does is put us in the most powerful position at one point in time. You always want to be... at your maximum power at the right time... that's what people talk about timing attacks... but to start off with what I suggest is that you be at your maximum power at all points in time.... You are probably in the stuff league.... With Terran, you are going to do that with the marine." One big problem here is that you have a lot of talk that either doesn't really mean anything or isn't relevant to a Step 1 tutorial for Terran (see bolded part).
- You don't mention building marines until 1:35 into the video.
- You don't start going through the "areas of focus" until after 2:00.
- You throw in a lot of extraneous stuff when going through the areas of focus (e.g. "money is power" and the long talk about using mules and gaining energy)
- You don't explain what a hard supply block is until 3:45.
Specific things I'd recommend:
- Begin each step video by opening up TheStaircase spreadsheet, and looking at the row corresponding to that step and explain what it means. So Terran Step 1 means we'll be building only scvs, marines, depots, etc. This way in the first 30 seconds of the video, the viewer will learn what Terran Step 1 means.
- After showing the row for the level, then scroll over to the left and talk briefly about the goals for that step (macro or micro goals).
- Then, briefly highlight what it means to "pass" the level, assuming league goal is, for example, Masters. For these first 3 things, I would recommend writing a script to make it as concise and to the point as possible. I would say you should be jumping into the game by 1:30 into the video at the latest.
- Start the replay at regular playback speed (not 2x), so you have time to go into more depth about the goals of the macro step: using control groups, camera locations, and hotkeys. Building tons and tons of rax, depots, scvs, and expansions. Scv saturation, mule usage, keeping money low, avoiding supply blocks, etc.
- Once the player's basic infrastructure and economy is up, then I think you should feel free to 2x/4x through while they send marines at their opponent until the game ends, since the outcome of the game is pretty much irrelevant.
Another thing you might want to consider is a "Quick Start Guide", either in text (near the beginning of the OP in the TL thread) or video form, that basically just lays out how to get started with TheStaircase ASAP, without all the extra explanatory stuff. For example:
- Pick a race. Find the column on the right side of the spreadsheet corresponding to that race. You start on Step 1.
- At each step, you will only be allowed to build a certain set of units (listed on the left) and buildings (listed in the center). At any given step, you may build any of the units and buildings for that step or for any of the previous steps.
- Each step has either a "Macro" or "Multitask/Micro" focus, listed next to the allowed buildings for that step. The focus determines your goals for that step, listed on the left side of the spreadsheet (e.g. learning to use your control groups, camera locations, and hotkeys are goals for the macro-focused steps.)
- To move to the next step, you need to pass the step you are on 3 times in a row. A pass is defined by the set of benchmarks listed on the left side of the spreadsheet, and depends on your desired league goal. For example, if you wish to reach Masters, use the Masters column benchmarks. See the details below the benchmarks for an explanation of how to read them.
This should be somewhere near the top of the TL thread, or you need to put all the background/psychology stuff inside of spoiler tags so people don't have to scroll a ton to be able to figure out what is going on. As it is now, you don't like to the spreadsheet until after multiple large paragraphs of philosophy and background information which many (or most?) people probably won't want to read. I think starting off with a very brief description and then the link to the spreadsheet, then the quick start guide and then the tutorials would be best, while putting all the background info / cool philosophical thinking inside of spoiler tags so that those who are interested can read it.
The Tutorial videos you have so far are a good start, but I don't think they cut to the chase quickly enough. The introduction tutorial has a lot of talk about "mechanics through strategy" or "strategy through mechanics" which I think would be more clearly explained like "Most people learn SC2 by practicing one particular pro build over and over again. For example, beginning Protoss players may be taught by a high-level coach how to execute a 4-gate, and they perform the same build over and over again and try to rise up the ranks. This approach does not work for some people because [insert your reasoning here]. TheStaircase is an alternative approach that starts players off with a very limited tech tree but total freedom in build order, and focuses on having good macro and slowly opening up the tech tree while maintaining macro benchmarks. This works better for some people because [insert your reasoning here]."
Let me know if I wasn't clear about anything in particular. I know a lot of it is brutally honest / critical, but I'm trying to help.
|
Thank you so much JDub, I really appreciate and agree with pretty much everything you said.
I will start implementing your suggestions asap
|
I was a bit afraid to try this method, but after my last losing streak (9 games in a row) i decided to give The Staircase a chance. I'm silver Protoss and my SQ was around 60 before I tried this method. Now I am in phase 3 - it means I can warp stalkers and zealots. First step was pretty easy, because I had not much to worry about - just making zealots, expanding, burning minerals and trying not to be supply blocked so my SQ skyrocketed into 90 SQ. My goal was Platinum (70 SQ) but after part 1 I decided to make it higher and it is Diamond (75 SQ) now.
Part 3 was not so easy because my gas skyrocketed and crushed my SQ: http://drop.sc/283826. I decided to calculate how many minerals and gas full saturated base can afford and here you are my calculations:
I have used one well saturated base on Daybreak with 16 probes in minerals and 6 and 4 probes on two geysers, then with 3 and 2 probes on one geyser.
After that I have opened Liquipedia and I used information about Protoss units and made 3 scenarios with assumption that Warpgate cooldown is 30 seconds.
Scenario 1:
16 probes on minerals and 2 geysers with 6 probes = 645 minerals + 232 gas 4 stalkers = 500 minerals + 200 gas The rest: 145 minerals + 32 gas After 30 seconds: 465 minerals + 148 gas = 3 stalkers + zealot
2 Warpins: (4 stalkers) + (3 stalkers + 1 zealot) = 7 stalkers + 1 zealot
Scenario 2:
16 probes on minerals and 2 geysers with 4 probes = 645 minerals + 180 gas 3 stalkers + 1 zealot = 475 minerals + 150 gas The rest: 170 minerals + 30 gas After 30 seconds: 490 minerals + 120 gas = 2 stalkers + 2 zealots
2 warpins: (3 stalkers + 1 zealot) + (2 stalkers + 2 zealots) = 5 stalkers + 3 zealots
Scenario 3:
16 probes on minerals and 1 geyser with 3 probes = 645 minerals + 112 gas 2 zealots + 2 stalkers = 450 minerals + 100 gas The rest: 195 minerals + 12 gas After 30 seconds: 520 minerals + 68 gas = 3 zealoty + 1 stalker + 1 pylon
2 warpins: (2 stalkers + 2 zealots) + (1 stalker + 3 zealots) = 3 stalkers + 5 zealots
Conclusion: I don't want to have stalker-only army so I decided to have only one geyser per base in phase 3 and it works fine. I can afford upgrades form 2 forges and twilight council and I burn both resources. You can download my game after adjustment: http://drop.sc/283827
Although I was getting better and bettery I have messed up in my last game and tried to win instead of keeping resources low. After that game I felt so sad even though I had won the game. Am I mad?
http://drop.sc/283828
You can check my macro progress in awesome tool made by Wraithan:
http://spendingq.com/graph/JJThompson
White line assigns goal for Diamond league: 75 SQ. Red rectangle: Steps 1 and 2 SQ. Green rectangle: Step 3 SQ before adjustments Blue rectangle: Step 3 SQ after adjustments.
P.S. If you want to make a tutorial and analyse my games it would be great. Thank you a lot for this awesome method. P.S. 2: I forgot to add that I'm playing with screen scrolling disabled.
Links: http://drop.sc/283826 http://drop.sc/283827 http://drop.sc/283828
http://spendingq.com/graph/JJThompson
|
I really need to add the zooming to the graph so it looks a bit better than that.
So I am practicing for the AHGL in HotS. I played my first preseason match earlier this week and will continue to have matches every Sunday for the foreseeable future. My question is what do folks who are doing this think of mixing and matching this style with more traditional practice?
This quote sticks out in my head, from a funny tech talk: "You don't need that, you can't drive a fucking Formula 1 car, you are on a bicycle." That is commonly how I feel when I tech up to HTs and such. But I feel I need them when playing against higher level players (plat/diamond) if I want to hold my own at all.
I was thinking of doing TheStaircase a couple nights a week and doing full practice the rest of the week. Interested in thoughts and ideas about that.
|
I don't see how using the staircase every once in a while could hurt your game. It's no different than working on micro only every once in a while with a micro trainer or some such. And with the Staircase map it's even easier to do sporadically. Good luck in the AHGL
|
|
|
|