Although I agree with most of those remarks, rock-paper-scissors proof would not necessarily be 90+%. Not unless you're doing some sort of advanced statistical analysis factoring in weighting based on skill and maps being played.
There is no viable build order that is an auto-loss even against its "hard counter." Even if we look at 4 pool vs 14cc, if the Terran has superior SCV control he can win. Even among progamers, there is a clear and vast difference between someone like Jaedong and your "average" progamer. It's like comparing the sun and the moon.
On June 30 2012 08:55 Miwyfe wrote: Awesome work nice one! The data from 2 player maps is much more valuable than 3 and 4 player maps because overlord scouting is not recorded in the data and players know where to send lings.
What does all this mean. Well unfortunately it means nothing. The 12 hatch might have beat the 12 pool everytime on one map and lost 5 times on another map.
Maps and spawn positions dictate build orders and these dictate the winner in broodwar! When doing statistics on broodwar, always separate data by the map then by spawn position. Only then can you start to analyze the data.
So as you can see I would personally group the data primarily by the map it was played on. Also I would only use 2 player maps and I would record where each player spawned (yes even on 2 player maps, and especially in zvz). If I were to do 3 or 4 player maps, I would only use games where both players scouted first time with overlords (or both players scouted second etc), and I would indicate which starting positions each player was in.
How about you though Crunchums? You said you would also do things differently if you were to do it again.
Data from 2 player maps is definitely more useful if what you are are interested in is some platonic ideal of how certain builds fare against each other. But if what you are interested in is how to win games, it is useful to about games like this one, and silly to fetishize a specific subset of the data (games where both players OL scout each other).
I didn't track things like spawn positions / OL scouting because doing so would have been more work. But I also believe that spawn positions would only be useful with some thinking*, and that if I was going to track OL scouting I'm not exactly sure how I would do it as there are actually a lot of different things a player can do with their first two overlords.
*Consider the in the following set of spawn positions {bottom left vs bottom right on Circuit Breaker, top left vs bottom right on Fighting Spirit, bottom right vs top right on Circuit Breaker, bottom left vs bottom right on La Mancha, and bottom right vs top right on Fighting Spirit} the last two are the only ones that are analogous.
On July 01 2012 00:24 Mortality wrote: Lately it seems overpool is out of fashion, but for a long time it was a staple of ZvZ, and probably the most common build.
My BW experience overlapped with some of that time, but the data's doesn't.
On July 01 2012 20:56 Chef wrote: ty for the stats. I think you should have included weighting for when a player was simply favoured to win in ZvZ. I also don't know if you chose the best season to gather stats, since by this time so many teams were disbanding and there were a lot of things to hurt players skills. It's difficult to read into this data but I'm still glad you made the effort to collect it. IMO rock paper sissors proof would look closer to 90% + wins in build order 'counters' Where this isn't much more extreme than someone getting away with 14 nex vs a cautious player.
Such a weighting would be possible if you had a lot of data, and would be useful and cool. Like maybe If 12h > 12p and 9p > 12h were both at 66% you could look at the weights to determine which matchup was harder to win from the side of the unfavored build. Or maybe you could look at how good each player is and be like daaaaaaamn player X is good or lol player Y is lucky.
I agree that games played while players were also practicing SC2 are suboptimal data points.
On July 03 2012 07:10 ForTenPoints wrote: I'm curious as to what a Chi-Square test or something similar will show about BO distribution for 2 vs 3 vs 4 player maps.
I don't know how to do one, but the data is there if you wanted to.
Yeah cool, I think we both agree then that 2 player maps are best for stats analysis. Thats good. Also a little note about the game you linked of Jaedong vs Great on Circuit Breakers. The game illustrates my point perfectly. JD went for early aggression, but scouts wrong direction with the overlord, then goes cross postion with his lings and finds an empty main. Then he loses the game to an econ build. What Im saying is, this thread brings up the (false) idea about coinflip build orders. However, on 4 player maps, the real coinflip (if I may be so bold to use the term) is the overlord scout direction.
Im going to give another example (hypothetical) of how its not silly to [concentrate] on specifics of the data. Consider PvT on Fighting Spirit. My theory is that Terran wins everytime he spawns behind the Protoss. Example P@1oclcock and T@7oclock. However, Protoss wins everytime he spawns behind the Terran. Example, P@7oclock and T@1oclock. Lets say I then look at every game and find it to be 90 percent true. However the winrate of PvT is 50%. Does this mean PvT is balanced on Fighting Spirit or does it mean Terran has unbeatable advantage in one spawn and Protoss has unbeatable advantage in another spawn?
Edit: Also I dont blame you at all for not bothering to record overlord scouting and spawns etc, indeed it is alot more work. The problem is I truely dont think the results mean anything without. The reason Im saying it is because I want everyone to understand better for future stats gathering. So yeah, if you are a person who is about to go through a load of vods and gather information, please separate the data primarily by map and concentrate on 2 player maps only. If you use 3 and 4 player maps, separate the data into spawn position.
On 2 player maps, "optimal" build order choices are determined almost entirely by OL scouting time and path. Like it or not, Starcraft is a game of calculated risk taking, so this scenario is not accurately representative of ZvZ in general, nor the effectiveness of the builds in question.
On July 05 2012 05:56 FT.aCt)Sony wrote: Test 9hatch/9pool/9ves/ol against 9pool/12pool/12hatch and see what the outcome is
That was my main ZvZ build when I played BW solely because it was a direct counter to the openings that could follow up.
That build is in the data as 9h and 10h (same thing but with extractor trick); it only shows up 4 times.
I'm not exactly sure what you are trying to say, but if what you are saying is that that build has a favorable matchup with all three of 9pool/12pool/12hatch then I think you are dead wrong.
Ha! After 15 mins searching I found the thread I commented on in 2008 where flag (is that the flag from the flag test in TSL??) did game theory to find how to open ZvZ, given certain winning probabilities. His problem was that he had to guess the winning probabilities for each matchup (opener vs opener), which of course made his result a lot less reliable. Now that you have some measured numbers, you should plug your statistics into his calculation, and you will mathematically calculate how often you should do each opening.
Another important thing is when the game was played. Games played near the start of a maps life cycle are using builds that are less refined than later on in the season. So its good if data is recorded in chronological order, and if results can be displayed over time.
On July 05 2012 05:56 FT.aCt)Sony wrote: Test 9hatch/9pool/9ves/ol against 9pool/12pool/12hatch and see what the outcome is
That was my main ZvZ build when I played BW solely because it was a direct counter to the openings that could follow up.
That build is in the data as 9h and 10h (same thing but with extractor trick); it only shows up 4 times.
I'm not exactly sure what you are trying to say, but if what you are saying is that that build has a favorable matchup with all three of 9pool/12pool/12hatch then I think you are dead wrong.
Be more than happy to load up BW and show you first hand if you want to test it. You'd be shocked with the results
On July 05 2012 05:56 FT.aCt)Sony wrote: Test 9hatch/9pool/9ves/ol against 9pool/12pool/12hatch and see what the outcome is
That was my main ZvZ build when I played BW solely because it was a direct counter to the openings that could follow up.
That build is in the data as 9h and 10h (same thing but with extractor trick); it only shows up 4 times.
I'm not exactly sure what you are trying to say, but if what you are saying is that that build has a favorable matchup with all three of 9pool/12pool/12hatch then I think you are dead wrong.
Be more than happy to load up BW and show you first hand if you want to test it. You'd be shocked with the results
Bo11, every game I 12h and you do your build, play on some non-crazy 2 player map? My one question is how good you are - I'm not going to be shocked at the results if you are like B or something because I am max D+ and haven't played in ages. If you are better than me would you be willing to play against a different zerg who is closer to your skill level instead of me?
On July 05 2012 05:56 FT.aCt)Sony wrote: Test 9hatch/9pool/9ves/ol against 9pool/12pool/12hatch and see what the outcome is
That was my main ZvZ build when I played BW solely because it was a direct counter to the openings that could follow up.
That build is in the data as 9h and 10h (same thing but with extractor trick); it only shows up 4 times.
I'm not exactly sure what you are trying to say, but if what you are saying is that that build has a favorable matchup with all three of 9pool/12pool/12hatch then I think you are dead wrong.
Be more than happy to load up BW and show you first hand if you want to test it. You'd be shocked with the results
Bo11, every game I 12h and you do your build, play on some non-crazy 2 player map? My one question is how good you are - I'm not going to be shocked at the results if you are like B or something because I am max D+ and haven't played in ages. If you are better than me would you be willing to play against a different zerg who is closer to your skill level instead of me?
I haven't touched BW since 2008 and left off with B rank when I stopped playing.
On July 05 2012 05:56 FT.aCt)Sony wrote: Test 9hatch/9pool/9ves/ol against 9pool/12pool/12hatch and see what the outcome is
That was my main ZvZ build when I played BW solely because it was a direct counter to the openings that could follow up.
That build is in the data as 9h and 10h (same thing but with extractor trick); it only shows up 4 times.
I'm not exactly sure what you are trying to say, but if what you are saying is that that build has a favorable matchup with all three of 9pool/12pool/12hatch then I think you are dead wrong.
Be more than happy to load up BW and show you first hand if you want to test it. You'd be shocked with the results
Bo11, every game I 12h and you do your build, play on some non-crazy 2 player map? My one question is how good you are - I'm not going to be shocked at the results if you are like B or something because I am max D+ and haven't played in ages. If you are better than me would you be willing to play against a different zerg who is closer to your skill level instead of me?
I haven't touched BW since 2008 and left off with B rank when I stopped playing.
I'm afraid all we will test then is much bw skills you lose in 4 years... :/
On July 05 2012 05:56 FT.aCt)Sony wrote: Test 9hatch/9pool/9ves/ol against 9pool/12pool/12hatch and see what the outcome is
That was my main ZvZ build when I played BW solely because it was a direct counter to the openings that could follow up.
That build is in the data as 9h and 10h (same thing but with extractor trick); it only shows up 4 times.
I'm not exactly sure what you are trying to say, but if what you are saying is that that build has a favorable matchup with all three of 9pool/12pool/12hatch then I think you are dead wrong.
Be more than happy to load up BW and show you first hand if you want to test it. You'd be shocked with the results
Bo11, every game I 12h and you do your build, play on some non-crazy 2 player map? My one question is how good you are - I'm not going to be shocked at the results if you are like B or something because I am max D+ and haven't played in ages. If you are better than me would you be willing to play against a different zerg who is closer to your skill level instead of me?
I haven't touched BW since 2008 and left off with B rank when I stopped playing.
I'm afraid all we will test then is much bw skills you lose in 4 years... :/
Glad you think that like a random elitist who isn't skilled or has little skill. Pretty sure if I've been playing SC2 it remains the same as long as I changed hotkeys over from what BW was. Only difference is the luck factor and the hard counter all-ins.
On July 05 2012 05:56 FT.aCt)Sony wrote: Test 9hatch/9pool/9ves/ol against 9pool/12pool/12hatch and see what the outcome is
That was my main ZvZ build when I played BW solely because it was a direct counter to the openings that could follow up.
That build is in the data as 9h and 10h (same thing but with extractor trick); it only shows up 4 times.
I'm not exactly sure what you are trying to say, but if what you are saying is that that build has a favorable matchup with all three of 9pool/12pool/12hatch then I think you are dead wrong.
Be more than happy to load up BW and show you first hand if you want to test it. You'd be shocked with the results
Bo11, every game I 12h and you do your build, play on some non-crazy 2 player map? My one question is how good you are - I'm not going to be shocked at the results if you are like B or something because I am max D+ and haven't played in ages. If you are better than me would you be willing to play against a different zerg who is closer to your skill level instead of me?
I haven't touched BW since 2008 and left off with B rank when I stopped playing.
I'm afraid all we will test then is much bw skills you lose in 4 years... :/
Glad you think that like a random elitist who isn't skilled or has little skill. Pretty sure if I've been playing SC2 it remains the same as long as I changed hotkeys over from what BW was. Only difference is the luck factor and the hard counter all-ins.
wut
No need for that tone, I meant no offence at all. I'll gladly admit that I sucked at sc:bw (solid D at icc) and suck equally at sc2 (barely diamond). I just thought that not playing a game for 4 years make you get a bit out of touch (specially at a high level like you), but I don't mind being proven wrong. Sorry about that.
Apart from the hotkeys and general mechenics, I think I remember pros that have taken brakes complain mainly that they lost their sense of timing when they got back. Maybe you got used to the sc2 timings, and would need some time to relearn the bw timings? I dont know, what do you think.
Anyways, it wouldn't be much useful for getting any reliable statistics. Would need a lot of different players for that.
On July 05 2012 05:56 FT.aCt)Sony wrote: Test 9hatch/9pool/9ves/ol against 9pool/12pool/12hatch and see what the outcome is
That was my main ZvZ build when I played BW solely because it was a direct counter to the openings that could follow up.
That build is in the data as 9h and 10h (same thing but with extractor trick); it only shows up 4 times.
I'm not exactly sure what you are trying to say, but if what you are saying is that that build has a favorable matchup with all three of 9pool/12pool/12hatch then I think you are dead wrong.
Be more than happy to load up BW and show you first hand if you want to test it. You'd be shocked with the results
Bo11, every game I 12h and you do your build, play on some non-crazy 2 player map? My one question is how good you are - I'm not going to be shocked at the results if you are like B or something because I am max D+ and haven't played in ages. If you are better than me would you be willing to play against a different zerg who is closer to your skill level instead of me?
I haven't touched BW since 2008 and left off with B rank when I stopped playing.
So yeah; you will probably curbstomp me regardless of builds. I am still interested in playing though; it will probably be fun.
Very nice post. I've considered doing this before, but it's so much work, so I'm impressed you went to the trouble.
The interesting thing to me about this is that it indicates that at the pro level, 9p crushes 12h (not a surprise), and is too close to call against every other build (definitely a surprise vs 12p!). So why don't pros just 9p every game?
EPT
I'll gladly admit that I sucked at sc:bw (solid D at icc) and suck equally at sc2 (barely diamond). I just thought that not playing a game for 4 years make you get a bit out of touch (specially at a high level like you), but I don't mind being proven wrong. Sorry about that.
