EPT"There are three kinds of lies: lies, damned lies, and statistics." - Benjamin Disraeli
Balance and Bonjwas: A Statistical Analysis - Page 4
| Forum Index > BW General |
|
SlayerS_BunkiE
Canada1707 Posts
"There are three kinds of lies: lies, damned lies, and statistics." - Benjamin Disraeli | ||
|
blueblimp
Canada297 Posts
Muta stacking is not particularly map-dependent, since it makes Zerg better on every map. It's also not strategy-dependent, as long as the strategy builds mutas at some point. It's not skill-dependent, since even D players can have better muta micro now than was possible for pro players before muta stacking. It's reasonable to argue that the popularization of Forge FE is comparable, except that Nal_rA was doing Forge FE before revolution day. Forge FE needed the right combination of maps and follow-ups to become game-changing. On the other hand, I've seen Yellow doing 2 hatch muta in an old VOD. The strategies where muta stacking is useful existed before muta stacking. It's not crazy to think that Yellow might have won a title if he just knew to group his mutalisks with an overlord. (As a random example I just pulled out of the VOD archive, check out this moment in a Boxer vs Yellow game. Imagine if those mutas were stacked.) On March 29 2012 14:55 oldgregg wrote: Ok probably not relevant and maybe no one cares but I just did some averages and for the top 10 peak and current ELO of ZvZ and PvP, ZvZ got higher scores for both, so PvP is more of a coinflip matchup! ZvZ peak average (for top 10): 2223.9 PvP peak avg: 2194.5 ZvZ current avg : 2125.3 PvP current avg: 2109.8 A progamer (Stork?) said in an interview that pro PvP is more of a coinflip than pro ZvZ. | ||
|
Savant
United States379 Posts
All this effort to prove that Terran *actually* has a 2-3% more chance of winning a particular game than some other race. I'd argue that in an important match, the many many other factors involved - skill level, practice, mindgames, map, mental condition, build order, etc. etc. more than muddle the stats to the point that racial imbalances don't matter when applied to an individual's league success. Flash wins because he can harness all factors to beat any player inferior to him no matter what and reach the finals consistently, and if he was Zerg or Protoss, he'd do the same thing. Bisu on the other hand, isn't dropping out of OSL's due to PvP's... | ||
|
Mortality
United States4790 Posts
Anyway, muta stacking had MUCH LESS OF AN IMMEDIATE IMPACT than netizens seem to think. It's a bad case of selective memory. Here's the game of an unknown Zerg DEMOLISHING NaDa who was 70% wins in the past 12 months TvZ against THE HARDEST IMAGINABLE opponents at the time: (see record here: http://www.teamliquid.net/tlpd/details.php?section=korean&type=players&id=147&part=games&vs=Z&league=standard&map=any&from_year=2004&from_month=4&from_day=27&to_year=2005&to_month=4&to_day=27&action=Update) That player's name was SAVIOR and he did it WITHOUT muta stacking. (see video here: http://www.teamliquid.net/tlpd/korean/games/2081_NaDa_vs_sAviOr/vod) Prior to Shark inventing the muta stacking trick, it was still possible to group your mutas, for example, by clicking them all over the same mineral patch. This form of "stacking" is what Savior uses to trap NaDa in his base and it's so effective he is able to expand twice AND tech hive before NaDa even moves out. It was 3 hatch muta into fast 4 base hive (later changed to 3 base hive) that "fixed" ZvT. The importance of muta stacking was that it allowed this style to remain effective to this day. Forge -> FE is actually VERY old. I first saw it done by {B_Blade]Annippy who was a progamer around 2001-2002. The build started entering the mainstream around 2004 or so and became the backbone of PvZ long before Bisu. In fact, Bisu did not even invent the "Bisu build." It was Daezang. Bisu perfected that style of play, but again, it wasn't really until about a year later that Protoss entered its height as this strategy was perfected. OLD SCHOOL ZvZ was the TRUE coinflip. In old school ZvZ palyers would go 9 pool (before OL) + gas, 12 pool + gas or 12 hatch. This created a rock-paper-scissors dynamic. If two players were of equal skill, the player with the BO advantage had an almost guaranteed victory, period. In modern ZvZ, 9 pool before OL is no longer used instead replaced by overpool (popularized by Savior) and 12 pool gas was changed into 12 pool -> 12 hatch gas. With these new builds, although BO advantages and disadvantages still exist, at the highest level it is possible to work around them with careful play and a little luck. basically, a 90% advantage got turned into more like a 60 or 70% advantage assuming equal skill. In PvP things have kind of gone the other way it seems. The issue is the reaver. As Starcraft history moves forward, build orders and midgame strategies always become more and more streamlined, so every scarab counts for more and every injury your reaver or your shuttle takes counts for more. | ||
|
Keone
United States812 Posts
On March 29 2012 15:48 Lightwip wrote: Mortality is right. It's a binomial distribution scenario. Given 2/3 winrate, you have a 74.1% chance of not getting less than 2 wins. Sorry to jump in with the statistics again... Let's clear this all up: Ways to win: WW, WLW, LWW Calculating A: (2/3)*(2/3) + (2/3)*(1/3)*(2/3) + (1/3)*(2/3)*(2/3) = 4/9 + 4/27 + 4/27 = 20/27 = 74.1% Calculating B: 1-A = 25.9% As far as I know, Lightwip & Mortality are right. | ||
|
Grend
1600 Posts
| ||
|
blueblimp
Canada297 Posts
On March 30 2012 03:49 Mortality wrote: OLD SCHOOL ZvZ was the TRUE coinflip. In old school ZvZ palyers would go 9 pool (before OL) + gas, 12 pool + gas or 12 hatch. This created a rock-paper-scissors dynamic. If two players were of equal skill, the player with the BO advantage had an almost guaranteed victory, period. In modern ZvZ, 9 pool before OL is no longer used instead replaced by overpool (popularized by Savior) and 12 pool gas was changed into 12 pool -> 12 hatch gas. With these new builds, although BO advantages and disadvantages still exist, at the highest level it is possible to work around them with careful play and a little luck. basically, a 90% advantage got turned into more like a 60 or 70% advantage assuming equal skill. Thanks for your interesting post. This paragraph is mistaken about modern ZvZ builds though. 9 pool gas (with the pool before overlord, and the gas usually for speed), 12 pool gas expand, and 12 hatch remain the most common builds in modern ZvZ. There are some other builds (such as overpool gas, overgas) but they are much less common. To ensure that I'm not crazy, I went through the 5 most recent ZvZs and checked builds:
The only unusual build in these 5 was Shine's overpool gas, and the game illustrates why overpool gas is rare, as his zerglings are too late to punish Hoejja's 12hatch. Edit: Regarding Forge FE, I thought it was pretty well known that Bisu didn't invent the Bisu build (and of course the Bisu build is only one particular follow-up to Forge FE). If you have a link to a VOD where the muta stacking trick was first used in a promatch, I'd be very interested. I thought the first time it was shown on TV was in Hwasin vs July, since Hwasin looks pretty surprised in that game. Given that July and Shark were both on MBC at the time, it seems plausible to me that Shark could've taught July the trick. | ||
|
VeNoM HaZ Skill
United States1528 Posts
On March 30 2012 03:49 Mortality wrote: The problem with using muta stacking as a cut-off is that it took a while before everybody started using the skill and even longer for everybody to master it. Muta stacking was invented in 2005 during the height of Terran supremacy. Anyway, muta stacking had MUCH LESS OF AN IMMEDIATE IMPACT than netizens seem to think. It's a bad case of selective memory. Here's the game of an unknown Zerg DEMOLISHING NaDa who was 70% wins in the past 12 months TvZ against THE HARDEST IMAGINABLE opponents at the time: (see record here: http://www.teamliquid.net/tlpd/details.php?section=korean&type=players&id=147&part=games&vs=Z&league=standard&map=any&from_year=2004&from_month=4&from_day=27&to_year=2005&to_month=4&to_day=27&action=Update) That player's name was SAVIOR and he did it WITHOUT muta stacking. (see video here: http://www.teamliquid.net/tlpd/korean/games/2081_NaDa_vs_sAviOr/vod) Prior to Shark inventing the muta stacking trick, it was still possible to group your mutas, for example, by clicking them all over the same mineral patch. This form of "stacking" is what Savior uses to trap NaDa in his base and it's so effective he is able to expand twice AND tech hive before NaDa even moves out. It was 3 hatch muta into fast 4 base hive (later changed to 3 base hive) that "fixed" ZvT. The importance of muta stacking was that it allowed this style to remain effective to this day. Forge -> FE is actually VERY old. I first saw it done by {B_Blade]Annippy who was a progamer around 2001-2002. The build started entering the mainstream around 2004 or so and became the backbone of PvZ long before Bisu. In fact, Bisu did not even invent the "Bisu build." It was Daezang. Bisu perfected that style of play, but again, it wasn't really until about a year later that Protoss entered its height as this strategy was perfected. OLD SCHOOL ZvZ was the TRUE coinflip. In old school ZvZ palyers would go 9 pool (before OL) + gas, 12 pool + gas or 12 hatch. This created a rock-paper-scissors dynamic. If two players were of equal skill, the player with the BO advantage had an almost guaranteed victory, period. In modern ZvZ, 9 pool before OL is no longer used instead replaced by overpool (popularized by Savior) and 12 pool gas was changed into 12 pool -> 12 hatch gas. With these new builds, although BO advantages and disadvantages still exist, at the highest level it is possible to work around them with careful play and a little luck. basically, a 90% advantage got turned into more like a 60 or 70% advantage assuming equal skill. In PvP things have kind of gone the other way it seems. The issue is the reaver. As Starcraft history moves forward, build orders and midgame strategies always become more and more streamlined, so every scarab counts for more and every injury your reaver or your shuttle takes counts for more. Just as as a note: very few builds in BW are attributed to the right person, SK Terran is one of the only ones that come to mind. And even though SoulKey pioneered it, NaDa is the one who took it to the point where it became a core style in TvZ. Which brings up the huge point of attributing the person who did perfect it. A lot of the times, a player can win with a certain style for the wrong reasons, mainly the whatthefuckbuildisthisi'veneverseenthisbeforewhatthefuckdoidohowdoicounteritwowilostwhatthefuck?!?!?!? factor, more commonly known as the Surprise Factor. The player who can take a build like that and make it perfectly viable, even when everyone knows it's coming is just as much, if not more, important to the process of making a build. | ||
|
koreasilver
9109 Posts
The main difference though, is that people have started to learn how to make the most out of build disadvantages a lot more now, mostly by studying how Jaedong used to pull out the most ridiculous wins during his long ZvZ supremacy. | ||
|
blueblimp
Canada297 Posts
On March 30 2012 03:49 Mortality wrote: Here's the game of an unknown Zerg DEMOLISHING NaDa who was 70% wins in the past 12 months TvZ against THE HARDEST IMAGINABLE opponents at the time: (see record here: http://www.teamliquid.net/tlpd/details.php?section=korean&type=players&id=147&part=games&vs=Z&league=standard&map=any&from_year=2004&from_month=4&from_day=27&to_year=2005&to_month=4&to_day=27&action=Update) That player's name was SAVIOR and he did it WITHOUT muta stacking. (see video here: http://www.teamliquid.net/tlpd/korean/games/2081_NaDa_vs_sAviOr/vod) Prior to Shark inventing the muta stacking trick, it was still possible to group your mutas, for example, by clicking them all over the same mineral patch. This form of "stacking" is what Savior uses to trap NaDa in his base and it's so effective he is able to expand twice AND tech hive before NaDa even moves out. It was 3 hatch muta into fast 4 base hive (later changed to 3 base hive) that "fixed" ZvT. The importance of muta stacking was that it allowed this style to remain effective to this day. I feel this game actually supports the importance of muta stacking. If Savior had known the muta stacking trick, he would've had a decent chance of ending this game with mutalisks (since Nada built only a single turret), without any change to his strategy. That's why muta stacking is so important: it makes every mutalisk build better, on every map, in every situation. Strategic innovations only apply if you actually use the strategy. | ||
|
Mindcrime
United States6899 Posts
On March 30 2012 09:01 blueblimp wrote: If you have a link to a VOD where the muta stacking trick was first used in a promatch, I'd be very interested. I thought the first time it was shown on TV was in Hwasin vs July, since Hwasin looks pretty surprised in that game. Given that July and Shark were both on MBC at the time, it seems plausible to me that Shark could've taught July the trick. No game showcases magic box stacked mutas better than that one, but it's not the first televised appearance of the trick. TL became aware of the trick a couple months earlier: http://www.teamliquid.net/forum/viewmessage.php?topic_id=40490 I have no idea when it was was first used in a televised game though. | ||
|
ajmbek
Italy460 Posts
good job man | ||
|
Mortality
United States4790 Posts
On March 30 2012 09:30 blueblimp wrote: I feel this game actually supports the importance of muta stacking. If Savior had known the muta stacking trick, he would've had a decent chance of ending this game with mutalisks (since Nada built only a single turret), without any change to his strategy. That's why muta stacking is so important: it makes every mutalisk build better, on every map, in every situation. Strategic innovations only apply if you actually use the strategy. It seems I am wrong about ZvZ, although I should add that when I said "12 pool gas" I don't mean "12 pool gas expo" but just "12 pool gas." That's how ZvZ used to be played. Lair would follow before a second hatch would be made. Something else that I did not point out is the benefit of modern muta micro to ZvZ (not just stacking, but in particular the use of "patrol"). This is why if you did not have the economic advantage you had to at the very least severely damage your opponent with lings or else you had no chance of victory. However, I disagree with your post above. Yes muta stacking would have made Savior even more imbalanced, but the point is that: 1) It is commonly thought among netizens that mutas were useless in ZvT prior to stacking, which is simply not true. In fact 3 hatch spire was a common build, although the particular variant I remember as being typically used was something like expo -> lair -> 3rd hatch -> spire, which was popularized for Lost Temple since mutas are crucial for defending your cliff from a drop. Savior's style revolutionized timings and gave Zerg a cohesive mid game plan that lead to a strong late game. Even without muta stacking. 2) Muta stacking was not the real turning point for ZvT. It was Savior's style. The real problem with ZvT was the timing issues. Trying to get such a fast hive would mean giving up the initiative, but persisting in lair tech would ultimately mean a weaker late game. Savior gained control of the timings, which when you get down to it, is the one common defining feature behind every revolutionary player since mid/late 2002 and onwards, certainly including TBLS. 3) Muta stacking was not perfected the moment it was introduced. It definitely created a skill gap between elite players and non-elite, as was made most painfully obvious by July's thorough dismantling of In_Dove in WCG Korea in 2006, but it took a full year before this skill gap started becoming obvious and it wasn't until much later (in particular Jaedong) that muta stacking showed its full advantage, which played a significant role in re-popularizing 2 hatch lair, which had been made completely obsolete by Savior. 4) In the long run it did indeed become necessary for Zerg's to continue with his style, but that turning point did not occur until sometime after Savior had already switched from 4 gas hive to 3 gas hive that it became a necessary component for success, because until then the main advance had been the streamlining of the timing inherent to the build. Edit: 5) At the end of the day what's relevant when choosing a cut-off date is that changes that happen that day are instantaneous and universal, so the 1.08 patch really is the only logical cut-off date. Muta stacking was really a much more gradual change than people remember. We didn't go from "lololol mutas are irrelevant" to Jaedong over night. Mutas were always relevant and the true impact of muta stacking took at least a year, arguably 2 years before it was truly felt. So no. I mean, I disagree with OP jumping to such conclusions based only on the rough numbers because it does not adequately describe the situation on the SC scene, and I do agree that muta stacking, Savior's style, and other innovations make the balance issues of Boxer era Starcraft completely irrelevant. But you can't just choose muta stacking as the cut-off point and say that's an accurate representation of balance either. There's no reason why that date stands out as being more significant than, for example, 3/3/2007. Prior to Bisu Savior was winning 80% of his ZvP games in brutally one-sided fashion. But again, it took about a year before the full impact of the Bisu build was realized... and what followed from there? Protoss domination in 2008. But then we could just as easily choose Jaedong's 4 hatch hydra/muta build as a turning point. Or July's 3 hatch spire -> 5 hatch hydra that followed about 6 months later. And again, it took about a full year for these advances to shift the balance. And then what occurred? SWARM SEASON! So OP did choose the best cut-off date, but what I've been saying this entire time is that Starcraft balance is controlled (mainly) by the metagame. And yes, he's not without a point that Terran has historically had the overall metagame edge for the longest, but that edge came from innovation, not imbalance. | ||
TheYango
United States47024 Posts
Games that have been played more recently are more indicative of balance than games that have been played further in the past. There are a multitude of factors to this--player improvement, discovery of certain core mechanics, more balanced map design, etc. The proper way to approximate this is not to toss out games at an arbitrary cutoff, but rather to re-calculate a time-weighted win percentage. Rather than using raw historical win % numbers, it would make more sense to determine a weighting scheme that gives higher value to recently played games. | ||
|
GhostOwl
766 Posts
On March 30 2012 03:35 Savant wrote: Lightwip made this thread just to cover his back every time he complains when Fla...I mean Terran wins an SL and every time Bis...I mean a Protoss doesn't. Your argument is that Terrans are statiscally favored towards the end of starleagues by nature of their imbalance, but look at recent results....its the ZERGS who are crowding the top finishes at starleagues, and almost all of them did it with victories in ZvZ's. Not to mention many many of them were flukes and not considered top tier players (Great vs Hydra anyone?). The ONLY terran players who've accomplished anything of note in (recent) individuals are Flash and Fantasy. All this effort to prove that Terran *actually* has a 2-3% more chance of winning a particular game than some other race. I'd argue that in an important match, the many many other factors involved - skill level, practice, mindgames, map, mental condition, build order, etc. etc. more than muddle the stats to the point that racial imbalances don't matter when applied to an individual's league success. Flash wins because he can harness all factors to beat any player inferior to him no matter what and reach the finals consistently, and if he was Zerg or Protoss, he'd do the same thing. Bisu on the other hand, isn't dropping out of OSL's due to PvP's... This is so wrong on so many levels.... Yes, Z did win a lot of starleagues but there are many factors in a tournament(luck, opponent level, maps, etc) If Flash played Z or P, he would NOT have the amount of dominance he has today. Terran, at S class level, is the strongest race. | ||
|
blueblimp
Canada297 Posts
That said, I still support muta stacking as an arbitrary cutoff. The reason is that there have been many times in Brood War's history when strategic innovations (and map pool for that matter) have changed match-up win rates. While Savior was one of the most significant, there have been many changes afterwards too. On the other hand, I can't think of any significant micro trick that was discovered after muta stacking. (Maybe Valkyrie micro? That's minor, anyway.) Since all-time match-up imbalances are less than 5%, then even if muta stacking only makes a slight difference to win rates, the lack of that technique would still be a significant disadvantage to early Zergs. | ||
|
GhostOwl
766 Posts
@Mortality: You do realize swarm season was 50/50 ZvT right? The fact that the period the Zergs did the best vs Terran was relatively even doesnt set off any alarms about match imbalance? | ||
|
moopie
12605 Posts
On March 30 2012 11:09 GhostOwl wrote: What is this season/past season's race winrates? http://www.teamliquid.net/tlpd/korean/team-leagues click on league -> see statistics. | ||
|
shaftofpleasure
Korea (North)1375 Posts
On March 30 2012 10:12 ajmbek wrote: 99% of you guys speaking vs the original post are really good at understading sc, but have apsolutly no idea of statistic good job man lol Why not divide the history of the scene into parts and not take the history of the game as a whole? divide it into 4s. 2000-2003 2004-2006 2007-2010 2011-present? or base the division on evolution/quality/balance-ish of maps? Started following back in 2007 but when I watched the older games, maps before were horrendous. | ||
|
Infinite Lurker
United States24 Posts
1. Examine what makes a unit "cost-effective", based on range, DPS, health, harassment potential and cost. Then determine what the most cost-effective units are. 2. Examine compositions. Which race's units have the best synergy and what are the risks of each unit composition. 3. Build times for units and buildings, transitioning between tech tiers, and the ability to exploit certain timings. For example, whichever race has the earliest exploitable timing windows will clearly have an advantage. 4. Spellcasters and support units (assume jaedong/boxer level micro) 5. Special abilities (i.e. repair, lifting, building warp-in, regeneration) 6. Responsiveness. The combination of a race's ability to be versatile with the each unit (i.e. which has the fewest hard and soft counters) and the speed with which a race can adapt their army composition. 7. Ability to exploit maps 8. Best music I'm sure there are many more features (SCV IMBA) to consider but I feel like it's impossible to address all of these coherently, and I think that if somebody could, they would find that different races win in each category. As a protoss though, vultures are the only thing i consider when i think of balance. If reavers were that fast I would drop the issue. | ||
| ||
