|
Part I : Win rates by Matchup & Length of Game
Part II : Win rates for "Pro" players vs "Scrubs" by matchup & Length of Game
For my next big attack on the Playhem data, I want to dig into build orders to see if we can identify "allin" build orders and calculate their effectiveness, independent of the length of the game. But before I do that, which is going to require figuring out , I want to take a brief detour into the results Pro vs Scrub matchups, to help answer the question "How often does the better player win?".
The definition of "pro" used previously isn't really suitable, since I was using the same data set to both determine who was a pro and then analyze pro win rates. Oops. I'd really like to have a defintion of "pro" that's independent of the data. But, that's not exactly easy. So instead I'll do the next best thing; I'll split the data set in half, use one half to decide who's a Pro and who's a Scrub, and then use the second half for analysis. Games played on even days were used only to identify Pros, while games played on odd days were used to analyze matchups. In addition, I strengthened the definition of pro to "any player who playes 7 games in the Ro8 (but ignoring Ro4, Ro2, etc.). In essence this means that during the even day tournaments, you must make the Ro8 at least three times. There are 96 pros in the playhem data set. The Pros make up a similar share of the player pool at all rounds up until the Ro2 in the odd-day tournaments as they do in even-day tournaments.
Let's look at the rate at which pros lose to Scrubs in Pro-vs-Scrub matches, based on the race of the Pro. In this chart, the Pro is the first race listed, which is why there are entries for both PvZ and ZvP. There are 3176 games in the entire data set.
![[image loading]](http://farm8.staticflickr.com/7067/6804589646_e793255ce1_z.jpg)
Playhem: Rate of Pro Losses to Scrubs by niq77174, on Flickr
Consistent conventional wisdom, PvP is the most "random" of the matchups. A Protoss Pro is only a 3-1 favorite against a Protoss Scrub, while a Terran Pro is an 8-1 favorite over a Terran Scrub. ZvZ is also somewhat random; 5.5-1 favorite vs an average 6.5-1 favorite across all matchups.
If you add up the losses to Scrubs across all matchups, then the overall Rate of losses to scrubs for each race looks like this:
Terran: 10.9%
Zerg: 15.7%
Protoss: 17.6%
Conversley, here is the rate at which Scrubs of a given race advance vs Pros:
Terran: 10.9%
Zerg: 16.4%
Protoss: 16.4%
Having seen these numbers, let me extend an olive branch to the Protoss. For while they have a great a-move army in the late game, they seem to face a number of other difficulties.- In mirror matches, PvP is the matchup where the weaker player is most likely to advance. By contrast, in TvT the weaker player is least likely to advance.
- In non-mirror matches, Protoss is overall the weakest race (both Z and T are above 50% vs P), while Terran is the strongest race.
The randomness of PvP (and to a lesser extent ZvZ) suggests that Zerg and Protoss players may want to avoid the mirror match by offracing in tournaments that are not race-locked. It may also explain the difficulties Protoss have had in tournaments. If it's more likely that weaker Protoss players advance, then they will be easy pickings in later rounds.
Why are PvP and ZvZ so random? Why is TvT so not random? Of the non-mirror matches, why is ZvP the most random? I have no idea what the answers to these questions are. Any conjectures?
|
Well I think ZvZ and PvP may have a higher randomness because in both cases build order choice can play a large role, and may simply decide the outcome.
Also in ZvZ, ling/bane wars are quite volatile, and while the player with better control will generally win, one slip up early on could cost you the game
in PvP one force field on the ramp could cost you the game.
I think the only real equivalent in TvT would be if hellions get into your mineral line early on...
In general though, I think PvP and ZvZ are just more volatile, so it is easier for non-pros to take pros out. This is one of the reasons that Nestea's ZvZ undefeated streak was so amazing.
|
pvp is a bad matchup as is some zvz so that makes sense
tvt is the best matchup in game and thus is the best stat wise here
|
On March 04 2012 12:06 orangesunglasses wrote:
pvp is a bad matchup as is some zvz so that makes sense
tvt is the best matchup in game and thus is the best stat wise here
your bias is showing a wee bit.
|
On March 04 2012 12:18 sc14s wrote:Show nested quote +On March 04 2012 12:06 orangesunglasses wrote:
pvp is a bad matchup as is some zvz so that makes sense
tvt is the best matchup in game and thus is the best stat wise here your bias is showing a wee bit.
I'm Protoss and I agree. PvP is god-awfully bad and has an unshakable coinflip feel to it. It's like there's a cycle of builds that blind counter other builds, and you just roll whatever one you're most comfortable playing. Lategame PvP is downright stupid, what with mass Colossi/MS butting heads. Complete a-move.
|
On March 04 2012 12:02 j3cht wrote:
Well I think ZvZ and PvP may have a higher randomness because in both cases build order choice can play a large role, and may simply decide the outcome.
Also in ZvZ, ling/bane wars are quite volatile, and while the player with better control will generally win, one slip up early on could cost you the game
in PvP one force field on the ramp could cost you the game.
I think the only real equivalent in TvT would be if hellions get into your mineral line early on...
In general though, I think PvP and ZvZ are just more volatile, so it is easier for non-pros to take pros out. This is one of the reasons that Nestea's ZvZ undefeated streak was so amazing.
This is pretty much my thoughts on the matter. PvP and ZvZ are extremely volatile matchups with a very high focus on your opening and extremely minor mistakes. PvZ/ZvP games are very often decided by what you do in the midgame in my opinion, with a huge focus on unit composition and the protoss' ability to either secure a third effectively without falling hugely behind, or end the game right there, making it easier in that matchup to beat pros as well. In essence, the more decisions, engagements, and game knowledge a particular game requires in any matchup, the greater the chance is the "better" player will win, while the fewer that the outcome is dependent on the greater the chance the "worse" player will get lucky and come out ahead.
Edit: Interestingly, this also gives somewhat of a vindication to my belief that if you learn how to execute build orders proficiently into the midgame, you can get a 15% winrate against just about anyone. The few who can micro their way out of straight out of straight build order losses against people who can do this are truly exceptional.
|
I think we're seeing these ratios because not all of the MU's are understood. Players like NesTea have the best ZvZ in the world because of their UNDERSTANDING of the matchup. I think that the ratio is simply caused by the fact that at this stage in the game, the people who are in GM and high high masters RIGHT NOW are there because of excellent mechanics and an understanding of what playstyles have been explored SO FAR in the game.
I think in a few years we'll be seeing these ratios go way down (especially in ZvZ, I personally feel like this MU isn't nearly as random as people say it is). The only MU I think could stay volatile is PvP simply because of the build order rock paper cissors we seem to have.
Also, IMO ZvP is the most volatile non-mirror MU because very often it boils down to protoss hitting a specific timing that zerg needs to scout extremely well to defend adequately. The ammount of different timings is quite large so I think we'r seeing good players die to all-ins they've never seen (let's say a zerg not expecting DT's after counting a certain number of chronos on the forge and thinking warpgate pressure). Or protoss players being caught off guard by an inferior zerg player cancelling his 3rd hatch and baneling busting. Mainly the all-ins haven't all been fleshed out and identified by everyone 
|
glorious tvt 
nice data, thanks.
|
I did want to add that my theory for why TvT is so non-volatile can be summed up in one word:
Tanks.
|
On March 04 2012 12:18 sc14s wrote:Show nested quote +On March 04 2012 12:06 orangesunglasses wrote:
pvp is a bad matchup as is some zvz so that makes sense
tvt is the best matchup in game and thus is the best stat wise here your bias is showing a wee bit.
in an MLG interview Huk said pvp is still the worst matchup in the game.
And from a skill wise perspectve (ignoring 'boredom') I'd agree that TvT is the best matchup, followed by tvz.
|
The data is unsurprising but It still is great that you have confirmed what everyone had already suspected.
|
On March 04 2012 12:51 xrapture wrote:in an MLG interview Huk said pvp is still the worst matchup in the game.
And from a skill wise perspectve (ignoring 'boredom') I'd agree that TvT is the best matchup, followed by tvz.
Certainly from a top player's perspective PvP is the worst. In playhem there's a serious skill gap, and the better player still loses 1 in 4. When there is less of a skill gap, the situation gets much worse. I would suspect that once you get to HuK's level, you have close to no control over your PvP win rate against anyone capable of making it into Code A.
If there's a GM Protoss who wants to post their win rates in each matchup that'd be useful .
|
On March 04 2012 13:13 ZeroTalent wrote:Show nested quote +On March 04 2012 12:51 xrapture wrote:in an MLG interview Huk said pvp is still the worst matchup in the game.
And from a skill wise perspectve (ignoring 'boredom') I'd agree that TvT is the best matchup, followed by tvz. Certainly from a top player's perspective PvP is the worst. In playhem there's a serious skill gap, and the better player still loses 1 in 4. When there is less of a skill gap, the situation gets much worse. I would suspect that once you get to HuK's level, you have close to no control over your PvP win rate against anyone capable of making it into Code A. If there's a GM Protoss who wants to post their win rates in each matchup that'd be useful .
Well, even as a master's Terran I find it nearly impossible to lose to a diamond Terran. Is it the same way for master P and Z's with their mirrors?
|
Fucking love this chart.
Shows the real problems with PvP and ZvZ, and the fact that they're extremely volatile matchups.
Thanks for posting!
|
yea god i hate PvP with a passion. sometimes i want to switch races, but i just enjoy playing toss too much. but then i get like 3 PvP's in a row and want to kill myself.
|
I believe this picture explains how mirrow matchups work
+ Show Spoiler +
The defensive building is normally tied to the teching build for Terran and Zerg. Without bunkers or sunkens, ZvZ and TvT would be very similiar to PvP. But because cannons are tied to the forge, we have many build order wins/losses. Starcraft BW got around this problem by having the shield battery, a weaker defensive building tied to the teching building - the gateway. You need to get a defender's advantage from somewhere.
ZvZ really is a knife fight. Even if you're better with a knife you're still on a razor's edge. Plus, the level of execution you need is very high, so among lower level players, it's essentially a luck based matchup.
|
On March 04 2012 13:23 xrapture wrote:Show nested quote +On March 04 2012 13:13 ZeroTalent wrote:On March 04 2012 12:51 xrapture wrote:in an MLG interview Huk said pvp is still the worst matchup in the game.
And from a skill wise perspectve (ignoring 'boredom') I'd agree that TvT is the best matchup, followed by tvz. Certainly from a top player's perspective PvP is the worst. In playhem there's a serious skill gap, and the better player still loses 1 in 4. When there is less of a skill gap, the situation gets much worse. I would suspect that once you get to HuK's level, you have close to no control over your PvP win rate against anyone capable of making it into Code A. If there's a GM Protoss who wants to post their win rates in each matchup that'd be useful . Well, even as a master's Terran I find it nearly impossible to lose to a diamond Terran. Is it the same way for master P and Z's with their mirrors?
Same for me. I played on my friend's diamond league account for 2 days and won every single TvT. When I offrace as Protoss I can sometimes beat master league players in PvP but at the same time can lose to platinum league players aswell, the matchup seems so random when I play it, sure I'm not really a protoss player and don't understand the matchup too well, but in TvT I always feel like the better player wins, in PvP I don't have that same feeling at all.
edit: The numbers are actually really good for all matchups except PvP, when the better players wins around 8-9 out of 10 matches on average then the game is extremely skill-based, and luck is not that much of a factor like some people (namely idra) make it out to be. Sure there are all-ins and there is some luck factor, but when the better players wins 9 out of 10 games I think that you can't really argue that sc2 is very luck-based.
edit2: In most sports upsets happen and there is a chance the underdog wins, that's a good thing, as long as it is based on skill and there only is a small luck factor, which is exactly the case in SC2. It's mainly skill but there is a slight chance for an upset with a little bit of luck. You see that in european football all the time, where top teams sometimes lose to "worse" teams, it doesn't happen very often, but just like in SC2 it happens sometimes and makes everything more interesting. It might not even be luck anyway, just a top player or pro player having a bad game and the "scrub" playing extraordinarily well.
|
Personally I don't like how Playhem is formatted. There are so many best of 1's in the beginning rounds that, if you're not absolutely on your game in those early rounds, you're gonna get knocked out with no chance to redeem yourself. The statics are interesting, but not all that surprising to me.
However, I do hope to see some changes in the PvP and ZvZ match ups and some point to make them less volatile @_@
|
I am really interested in how these numbers pan out when cross referenced with how effective cheese/all-ins are. Although I have no stats to back it up, I always felt they seemed more effective that they should be. Still, these stats do confirm that feeling I get during PvP is totally natural and the match up is sort of a weird mess.
|
On March 07 2012 01:07 ChaosTerran wrote:
edit: The numbers are actually really good for all matchups except PvP, when the better players wins around 8-9 out of 10 matches on average then the game is extremely skill-based, and luck is not that much of a factor like some people (namely idra) make it out to be. Sure there are all-ins and there is some luck factor, but when the better players wins 9 out of 10 games I think that you can't really argue that sc2 is very luck-based.
edit2: In most sports upsets happen and there is a chance the underdog wins, that's a good thing, as long as it is based on skill and there only is a small luck factor, which is exactly the case in SC2. It's mainly skill but there is a slight chance for an upset with a little bit of luck. You see that in european football all the time, where top teams sometimes lose to "worse" teams, it doesn't happen very often, but just like in SC2 it happens sometimes and makes everything more interesting. It might not even be luck anyway, just a top player or pro player having a bad game and the "scrub" playing extraordinarily well.
I think this is right. Though, for the most we're talking about pro level players versus very serious casual players or up-and-coming pros here. It would be like taking the MLB All-Star team and having them play a AAA allstar team. Sure, the AAA team would win a few times (just look at the World Baseball Classic) but it's so rare that it's close to dumb luck.
I may try to re-run these numbers using more buckets for players than "pros" and "scrubs" to see what happens when one player is only a slight favorite. Because being an 8-1 favorite is actually a very uninteresting game from a spectators perspective. Even in the NBA, the best team finishes the season at something like 65-17, which is a 4-1 favorite or so. The '97 Bulls were a 7-1 favorite. Great NFL teams can go 15-1 or 14-2, but that's rare; in the typical season the best team is 13-3 or 12-4.
|
Love this chart, great data !
Could I suggest for one of you future analysis, you correlate the %win of a race in a given matchup with the research of a ceratin upgrade or creation of a certain building (notably, zealot charge, blink, termal lance, ghost academy, greater spire, etc).
Having, for example, the %winrate of a P in PvT with Charge researched and the %winrate without Charge researched could help understand the 'potential' late game imbalances in this matchup (this is by no mean a balance whine as the matchup seems more or less balanced over the whole course of a game).
|
Pretty nice analysis!
I think it all comes down to the defenders advantadge, which is basically nonexistent in both PvP and ZvZ, while it is very strong in TvT.
Here is hoping that Blizz is trying to improve the early game consistency of both P and Z in HotS, for the overall improvement of the game. Things like the no high-ground warp-in and the chronoboost-cannon might help with P, but with Z I have no idea if they're doing anything at all.
|
On March 07 2012 04:04 ZeroTalent wrote:Show nested quote +On March 07 2012 01:07 ChaosTerran wrote:
edit: The numbers are actually really good for all matchups except PvP, when the better players wins around 8-9 out of 10 matches on average then the game is extremely skill-based, and luck is not that much of a factor like some people (namely idra) make it out to be. Sure there are all-ins and there is some luck factor, but when the better players wins 9 out of 10 games I think that you can't really argue that sc2 is very luck-based.
edit2: In most sports upsets happen and there is a chance the underdog wins, that's a good thing, as long as it is based on skill and there only is a small luck factor, which is exactly the case in SC2. It's mainly skill but there is a slight chance for an upset with a little bit of luck. You see that in european football all the time, where top teams sometimes lose to "worse" teams, it doesn't happen very often, but just like in SC2 it happens sometimes and makes everything more interesting. It might not even be luck anyway, just a top player or pro player having a bad game and the "scrub" playing extraordinarily well. I think this is right. Though, for the most we're talking about pro level players versus very serious casual players or up-and-coming pros here. It would be like taking the MLB All-Star team and having them play a AAA allstar team. Sure, the AAA team would win a few times (just look at the World Baseball Classic) but it's so rare that it's close to dumb luck. I may try to re-run these numbers using more buckets for players than "pros" and "scrubs" to see what happens when one player is only a slight favorite. Because being an 8-1 favorite is actually a very uninteresting game from a spectators perspective. Even in the NBA, the best team finishes the season at something like 65-17, which is a 4-1 favorite or so. The '97 Bulls were a 7-1 favorite. Great NFL teams can go 15-1 or 14-2, but that's rare; in the typical season the best team is 13-3 or 12-4.
Excellent point.
What definition of "pro" did you actually use when calculating these stats? I mean there are alot of players I could think of where I wouldn't know how to categorize them and if you rank low- mid grandmasters players and silver league players as "scrubs" then there will always be some sort of inaccuracy.
I think the best way for you to sort of re-run these numbers is to look at the leagues, sure leagues don't mean alot, but it is less arbitrary then "scrub", "pro", "semi-pro". You could look at how often it happens that a diamond player beats a master league player or a master league player beats a pro and if there are any pros who lose to some players that aren't high master, etc..
that would be interesting aswell.
edit: Here is what I would do:
Pros
Grandmaster
Master
Diamond
Platinum + Gold
Silver + Bronze
this is how I would rank people, it's still not accurate, but if you really plan on re-running these numbers I think this would be a better system to use.
|
Lol this is great, showing the invalidities of each race and how the design needs to be fixed. It's clear that terran's design is sound, whereas someone like me, who doesn't even play protoss has a chance of beating SKMC. The game is still infantile I know, but Blizzard only cares about game balance, not design balance.
|
On March 07 2012 06:40 ChaosTerran wrote:
What definition of "pro" did you actually use when calculating these stats? I mean there are alot of players I could think of where I wouldn't know how to categorize them and if you rank low- mid grandmasters players and silver league players as "scrubs" then there will always be some sort of inaccuracy..
The OP says: "In addition, I strengthened the definition of pro to "any player who playes 7 games in the Ro8 (but ignoring Ro4, Ro2, etc.). In essence this means that during the even day tournaments, you must make the Ro8 at least three times. "
Seems like a perfectly good, measurable, objective criteria given his data set. It seems to go a long ways to identifying how much of a coin flip SC2 is which is what the post is about. Its not about some philosophical argument about the essence of "pro" or "scrub".
Anyway, great OP, well done, and very interesting.
|
On March 07 2012 06:47 BushidoSnipr wrote:
Lol this is great, showing the invalidities of each race and how the design needs to be fixed. It's clear that terran's design is sound, whereas someone like me, who doesn't even play protoss has a chance of beating SKMC. The game is still infantile I know, but Blizzard only cares about game balance, not design balance.
Honestly, it seems the data shows SC2 is in pretty darn good shape. PvP needs help, but the way people trash SC2, it all should have been much worse.
|
On March 07 2012 07:15 Smackzilla wrote:Show nested quote +On March 07 2012 06:47 BushidoSnipr wrote:
Lol this is great, showing the invalidities of each race and how the design needs to be fixed. It's clear that terran's design is sound, whereas someone like me, who doesn't even play protoss has a chance of beating SKMC. The game is still infantile I know, but Blizzard only cares about game balance, not design balance. Honestly, it seems the data shows SC2 is in pretty darn good shape. PvP needs help, but the way people trash SC2, it all should have been much worse.
looks pretty bad to me, even the 10% in tvt is probably worse than you would see in any broodwar matchup maybe excluding zvz. I hope this will improve with heart of the swarm
|
On March 07 2012 07:21 nttea wrote:Show nested quote +On March 07 2012 07:15 Smackzilla wrote:On March 07 2012 06:47 BushidoSnipr wrote:
Lol this is great, showing the invalidities of each race and how the design needs to be fixed. It's clear that terran's design is sound, whereas someone like me, who doesn't even play protoss has a chance of beating SKMC. The game is still infantile I know, but Blizzard only cares about game balance, not design balance. Honestly, it seems the data shows SC2 is in pretty darn good shape. PvP needs help, but the way people trash SC2, it all should have been much worse. looks pretty bad to me, even the 10% in tvt is probably worse than you would see in any broodwar matchup maybe excluding zvz. I hope this will improve with heart of the swarm
10% is not very good. 5% would be better for all the match ups, since we are talking about players of drasticly different skill levels. I would like to see the more skilled player win 19 out of 20 games. Blizzard seems to be on that track, with providing more tools to deal with early game pressure. We will have to see when the beta hits and people get to mess with those new units.
|
Still, it is strange that some players can have insane winratios in PvP and ZvZ. Are they lucky? Or do they know more?
|
So your definition of pro is based on how many playhem tournament one plays? This seems like an extremely flawed analysis...
|
Can you also provide the number of data you have for each matchup? Permutations dictate that mirror matchups happen at most half as often as non-mirrors. This will change the error bars quite a bit to the probabilities.
|
On March 07 2012 07:21 nttea wrote:Show nested quote +On March 07 2012 07:15 Smackzilla wrote:On March 07 2012 06:47 BushidoSnipr wrote:
Lol this is great, showing the invalidities of each race and how the design needs to be fixed. It's clear that terran's design is sound, whereas someone like me, who doesn't even play protoss has a chance of beating SKMC. The game is still infantile I know, but Blizzard only cares about game balance, not design balance. Honestly, it seems the data shows SC2 is in pretty darn good shape. PvP needs help, but the way people trash SC2, it all should have been much worse. looks pretty bad to me, even the 10% in tvt is probably worse than you would see in any broodwar matchup maybe excluding zvz. I hope this will improve with heart of the swarm
A 9-1 favorite is bad? Anyway, this is borderline trolling without real data. Keep in mind, we're talking about a daily open where we compare guys who hit quarterfinals a few times vs. those who don't. It's not Flash vs. a good amateur.
|
On March 07 2012 07:35 ThE_OsToJiY wrote:
So your definition of pro is based on how many playhem tournament one plays? This seems like an extremely flawed analysis...
Oh good grief. He simply has a data set where he's comparing players with consistent success vs. those who enter but are not consistently successful. He's not trying to tell you what a "pro" is.
|
Well, now we at least have a solid statistical verification that pvp and zvz are the most volatile matchups.
|
On March 07 2012 07:31 Plansix wrote:Show nested quote +On March 07 2012 07:21 nttea wrote:On March 07 2012 07:15 Smackzilla wrote:On March 07 2012 06:47 BushidoSnipr wrote:
Lol this is great, showing the invalidities of each race and how the design needs to be fixed. It's clear that terran's design is sound, whereas someone like me, who doesn't even play protoss has a chance of beating SKMC. The game is still infantile I know, but Blizzard only cares about game balance, not design balance. Honestly, it seems the data shows SC2 is in pretty darn good shape. PvP needs help, but the way people trash SC2, it all should have been much worse. looks pretty bad to me, even the 10% in tvt is probably worse than you would see in any broodwar matchup maybe excluding zvz. I hope this will improve with heart of the swarm 10% is not very good. 5% would be better for all the match ups, since we are talking about players of drasticly different skill levels. I would like to see the more skilled player win 19 out of 20 games. Blizzard seems to be on that track, with providing more tools to deal with early game pressure. We will have to see when the beta hits and people get to mess with those new units.
I don't get the "drastically" different skills" comment. "Scrub" can include players who has made top 16 often but have only broke quarters once. Look at the actual criteria used to bin the players. Heck, you could have DRG play in playhem once, win the whole thing, never play again, and he's a "scrub".
At least that's how I'm interpretting the OP. Did I miss something?
|
|
|
On March 07 2012 07:59 Smackzilla wrote:Show nested quote +On March 07 2012 07:35 ThE_OsToJiY wrote:
So your definition of pro is based on how many playhem tournament one plays? This seems like an extremely flawed analysis... Oh good grief. He simply has a data set where he's comparing players with consistent success vs. those who enter but are not consistently successful. He's not trying to tell you what a "pro" is.
Yea...he's only defining "pro" within the context of his dataset. He has to define it to establish the criteria that he is measuring. Show me a study where the researcher CAN avoid defining their terms. The important thing is that he is clear in his definition. And yes, given that his data is all based on Playhem, of course he is going to define "pro" in relation to the data he is analyzing. If you want to dismiss statistical studies, go for it...I'm always skeptical of statistics because of these built in assumptions. But don't fault the OP for being clear and self-reflexive about his assumptions--about his criteria for "pro."
|
Zerg and toss units are fast thus giving less reaction time to the opponent. Tvt is well balanced in skill cause the units dont bum rush you.
|
On March 07 2012 08:33 agahamsorr0w wrote:
Zerg and toss units are fast thus giving less reaction time to the opponent. Tvt is well balanced in skill cause the units dont bum rush you.
Terran has the most allins and the most effective ones too.....I still love how people think zerg is all about rushing because its the exact opposite lol
|
On March 07 2012 08:40 BushidoSnipr wrote:Show nested quote +On March 07 2012 08:33 agahamsorr0w wrote:
Zerg and toss units are fast thus giving less reaction time to the opponent. Tvt is well balanced in skill cause the units dont bum rush you. Terran has the most allins and the most effective ones too.....I still love how people think zerg is all about rushing because its the exact opposite lol
I think he means that the units like the zergling (and just morph banelings right outside his base) are fast as in they can cross the map quickly so there isn't alot of time to react. In PvP the rush distance is even lower thanks to warpgates.
|
This is amazing data, your graphs line up with my general understanding of the pro scene too, especially this most recent Pros: Rate of losses to scrubs by Matchup. ZvP has always been the most volatile non-mirror matchup, Terran seems to be able to control terrain much better than Z and P therefore seemingly allowing more strategy and less randomness.
When SC2 released I knew we'd soon enough have enough game data under our belt to actually come out with data like this. With 21k games on Playhem you're certainly doing that very well!
Just beautiful information, I look forward to the next number-crunch installment!
|
What I find remarkable is that right after PvP we have all three zerg matchups. As a zerg player myself, I'd say that is because zerg is most susceptible to cheese, and there's always the possibility of protoss or terran coming back from a disadvantage with their cost effective death balls.
|
On March 11 2012 19:51 Scorch wrote:
What I find remarkable is that right after PvP we have all three zerg matchups. As a zerg player myself, I'd say that is because zerg is most susceptible to cheese, and there's always the possibility of protoss or terran coming back from a disadvantage with their cost effective death balls.
wut terran deathball, what's that ?
|
I love the stats. The ratios aren't even that bad compared to real sports. Even PvP.
|
On March 11 2012 20:55 zezamer wrote:Show nested quote +On March 11 2012 19:51 Scorch wrote:
What I find remarkable is that right after PvP we have all three zerg matchups. As a zerg player myself, I'd say that is because zerg is most susceptible to cheese, and there's always the possibility of protoss or terran coming back from a disadvantage with their cost effective death balls. wut terran deathball, what's that ?
Mech
|
Nice analysis, maybe you could use TLPD to get a definition of "pro" players?
|
On March 04 2012 12:25 Shiori wrote:Show nested quote +On March 04 2012 12:18 sc14s wrote:On March 04 2012 12:06 orangesunglasses wrote:
pvp is a bad matchup as is some zvz so that makes sense
tvt is the best matchup in game and thus is the best stat wise here your bias is showing a wee bit. I'm Protoss and I agree. PvP is god-awfully bad and has an unshakable coinflip feel to it. It's like there's a cycle of builds that blind counter other builds, and you just roll whatever one you're most comfortable playing. Lategame PvP is downright stupid, what with mass Colossi/MS butting heads. Complete a-move.
I'm random, but PvP is one of my favorite matchups of all of the matchups, not just mirror matches. It's not that PvP is a coinflip, because it's not. It is however the most unforgiving all the matchups. Even a single mistake of losing a single stalker at the wrong time can mean death, and as an observer it is nailbitingly intense to watch. As a player, achieving a victory in PvP is difficult and challenging, sometimes aggravating, but extremely gratifying. Knowing that you played your best game and made as few mistakes a possible to be able to out-protoss another protoss is satisfying.
|
Keep up the good work, I love seeing stats like this. I think it's interesting that the three matchups with the lowest % are all vs Terran. Is the notion of Terran being a race full of surprise moves that can topple a superior player wrong?
|
Interesting stuff...
From terran PoV it seems like when playing against a better player, terran is the worst race. It's kinda funny since players all-in/cheese a lot in these situations. That said, it looks like terran all-ins/cheeses are the weakest out of all races, which many people wouldn't agree on.
Maybe it has something to do with the execution? The more micro-intensive your all-in is, the harder it is to pull off against someone better than you. Even the earliest viable terran all-ins such as bunker rush take quite a lot of control to work.
Or maybe the way the pros are measured is just flawed.
|
I read this chart as:
"If you're Terran and not pro, you're pretty f*cked."
|
On March 04 2012 12:18 sc14s wrote:Show nested quote +On March 04 2012 12:06 orangesunglasses wrote:
pvp is a bad matchup as is some zvz so that makes sense
tvt is the best matchup in game and thus is the best stat wise here your bias is showing a wee bit.
He's right though? In TvT the better terran will win, usually ^_^
|
I would love it if playhem had an invite only tourney of somesort that was a series of bo3 at the end of each week with the final match being a bo5.
|
On March 11 2012 21:41 Fyrewolf wrote:Show nested quote +On March 04 2012 12:25 Shiori wrote:On March 04 2012 12:18 sc14s wrote:On March 04 2012 12:06 orangesunglasses wrote:
pvp is a bad matchup as is some zvz so that makes sense
tvt is the best matchup in game and thus is the best stat wise here your bias is showing a wee bit. I'm Protoss and I agree. PvP is god-awfully bad and has an unshakable coinflip feel to it. It's like there's a cycle of builds that blind counter other builds, and you just roll whatever one you're most comfortable playing. Lategame PvP is downright stupid, what with mass Colossi/MS butting heads. Complete a-move. I'm random, but PvP is one of my favorite matchups of all of the matchups, not just mirror matches. It's not that PvP is a coinflip, because it's not. It is however the most unforgiving all the matchups. Even a single mistake of losing a single stalker at the wrong time can mean death, and as an observer it is nailbitingly intense to watch. As a player, achieving a victory in PvP is difficult and challenging, sometimes aggravating, but extremely gratifying. Knowing that you played your best game and made as few mistakes a possible to be able to out-protoss another protoss is satisfying.
No, sorry. I don't know what level you play at, but its coinflippy as hell. Its just that its unforgiving on top of that again. You have to guess and insinuate what you opponent is doing. Try coming back from a robo build when you opponent is going phoenixes for instance. And even if you are somewhat prepared for it, you will have no idea if your opponent is planning an allin with those phoenixes, or has taken an expansion. Both of which you will need completely different answers for.
|
Very interesting stat. What I would find interesting is to see with which strategies a non-pro player has the highest probability to win against a pro (e.g. 4-Gate, Baneling Bust). I know that's hard to extract automatically, but maybe it can be done with some simplifications/heuristics. It would also be nice to see if games where a non-pro wins against a pro are shorter on average than other games.
|
Excludos, if you went robo you will at the last have an observer and a scouting advantage (by the time you see the phoenixes, so at the very least you should know exactly when his expansion goes down (if it does). That doesn't mean you'll necessarily survive the phoenixes, but scouting shouldn't be your problem!
|
On March 12 2012 00:37 alQahira wrote:
Excludos, if you went robo you will at the last have an observer and a scouting advantage (by the time you see the phoenixes, so at the very least you should know exactly when his expansion goes down (if it does). That doesn't mean you'll necessarily survive the phoenixes, but scouting shouldn't be your problem!
If he makes it as his natural, makes no canons, and doesn't make a robotics himself. Yes, you will spot the expansion going down.. And like you said, that still doesn't mean you'll even survive long enough for that to even happen.
|
Well, if he is going stargate plus robo or cannons, and you're not going DTs, then we're not talking about super high level play here, are we, which is the topic of discussion.
|
On March 04 2012 11:52 ZeroTalent wrote:
In non-mirror matches, Protoss is overall the weakest race (both Z and T are above 50% vs P), while Terran is the strongest race.
Sounds familiar.
BTW. It's not really randomness in mirrors. There is just less room for mistakes in PvP and ZvZ, thus minor factors can have great influence on the outcome of the game. Most of the time it's engagements and the fact that unsafe styles of play are relatively strong.
|
On March 12 2012 01:05 alQahira wrote:
Well, if he is going stargate plus robo or cannons, and you're not going DTs, then we're not talking about super high level play here, are we, which is the topic of discussion.
What? Phoenixes + immortal is pretty much standard by now.. Its so incredibly strong I personally have no idea how to react to it.
edit: I am talking about high lvl play here. If you stay on only stargate the entire game, you're either allin or you don't know how to play phoenixes. And canons is pretty much standard when expanding with a phoenix build.
|
On March 12 2012 01:09 Excludos wrote:Show nested quote +On March 12 2012 01:05 alQahira wrote:
Well, if he is going stargate plus robo or cannons, and you're not going DTs, then we're not talking about super high level play here, are we, which is the topic of discussion. What? Phoenixes + immortal is pretty much standard by now.. Its so incredibly strong I personally have no idea how to react to it.
After you get blink as a response to phenixes, the 2 base chargelot/archon (with 6+ gates) with blink stalkers to snipe optional colossi works very well, unless they make wall of buildings in time, but you can just expand then. It's pretty much hard counter.
|
On March 12 2012 01:10 -Kira wrote:Show nested quote +On March 12 2012 01:09 Excludos wrote:On March 12 2012 01:05 alQahira wrote:
Well, if he is going stargate plus robo or cannons, and you're not going DTs, then we're not talking about super high level play here, are we, which is the topic of discussion. What? Phoenixes + immortal is pretty much standard by now.. Its so incredibly strong I personally have no idea how to react to it. After you get blink as a response to phenixes, the 2 base chargelot/archon with blink stalkers to snipe optional colossi works very well, unless they make wall of buildings in time, but you can just expand then.
The key word you used was "2 base"..which is pretty much what we have been talking about in the last few posts.
|
On March 11 2012 19:51 Scorch wrote:
What I find remarkable is that right after PvP we have all three zerg matchups. As a zerg player myself, I'd say that is because zerg is most susceptible to cheese, and there's always the possibility of protoss or terran coming back from a disadvantage with their cost effective death balls.
What..? ZvT is the 3rd least volatile and PvZ is more volatile than ZvP. Not sure what you're looking at.
|
On March 04 2012 11:52 ZeroTalent wrote:Playhem: Rate of Pro Losses to Scrubs by niq77174, on Flickr Consistent conventional wisdom, PvP is the most "random" of the matchups. A Protoss Pro is only a 3-1 favorite against a Protoss Scrub, while a Terran Pro is an 8-1 favorite over a Terran Scrub. ZvZ is also somewhat random; 5.5-1 favorite vs an average 6.5-1 favorite across all matchups.
Can't anyone in the world just learn to do math correctly? The only proportion you got right was the one that was 25%, nice and easily cut into 4ths, holy crap. Math lesson time.
If someone has a 25% chance to win, their chance is equivalent to 1 in 4 parts, so there are 3 parts that they will lose, so 1:3 that they will win, get it? You got that one right.
If someone has an 18% chance to win, then they will win 1 in 5.5 parts. This does NOOOOT mean that their ratio to win is 1:5.5. There are 4.5 parts that they will lose, so the ratio is 1:4.5. An easier (read: less steps) process to doing this is just dividing the chance that they have to lose (82%) to the chance that they have to win (18%). This will give you 4.5, which is the proportion 1:4.5 of how often the scrub will win.
Jeez.
|
Awesome stats <3
Playhem stats is one of the best thing that has happened to TL forums
|
On March 12 2012 01:27 K3Nyy wrote:Show nested quote +On March 11 2012 19:51 Scorch wrote:
What I find remarkable is that right after PvP we have all three zerg matchups. As a zerg player myself, I'd say that is because zerg is most susceptible to cheese, and there's always the possibility of protoss or terran coming back from a disadvantage with their cost effective death balls. What..? ZvT is the 3rd least volatile and PvZ is more volatile than ZvP. Not sure what you're looking at.
Oh you're right, the matchups go in both directions. I just saw three matchups involving zerg at #2-4 and derped. Never mind.
|
I don't want to sound snippy, but there's been a disappointing trend in statistics topics on TL where people don't bother trying to estimate statistical significance when comparing means between groups. Without an analysis of significance here, it's actually impossible to know if any of the differences or trends you describe are real.
Please, at least attempt to estimate how significant your results are! I understand it's incredibly easy to just calculate averages and convince people that there's a meaningful result here, but you're not really doing anybody any favors by (unintentionally) misleading people in this way.
|
On March 04 2012 11:52 ZeroTalent wrote:
Why are PvP and ZvZ so random? Why is TvT so not random? Of the non-mirror matches, why is ZvP the most random? I have no idea what the answers to these questions are. Any conjectures?
First of all, thank you for these stats, they are very insightful!
I can´t say that much about P and Z since i play Terran, but it seems to me, that both Zerg and Protoss have several major disadvantages compared to terran, especially in the early/early-mid game.
1)
Terrans have marines 
Marines and marine/marauder are good against pretty much everything in the early game. They are cheap, quickly build and have a lot of dps. on top of that, they scale very good with upgrades (stim, combat shields, weapon/armor upgrades and medivac heal). So its very unlikely to die to something unexpected.
2) scans
If all scouting fails for a terran, he can almost always scan. That helps a lot to identify some all in/cheese.
3) addon switching
tech switches are very fast for a terran. All he has to do (in most cases) is to switch the addons of the buildings.
Need some tanks? Take tech lab on your your fax and start production immediatly! Need more marines? Take a reactor on the rax a.s.o....
4)M.U.L.E.S.
Even when your economy is behind your oponents, mules help you to come bach into the game. As long mules "are" energy on the orbital, you cannor kill them as long as you dont kill the orbital, which is quite hard in early game. Therefore, mules excuse mistakes of a terran player much more than chrono and especially larva inject. They allow you to play an all in with scvs and still not be completely all in. Again, this is an ability, that buys time for a terran in all match ups.
all in all, i think terrans have more/easier to use tools for a better player/pro to not die in the early and mid game. That helps to get to a phase of the game where the pro can benifit from his/her better mechanics/macro/micro/multitasking.
|
On March 12 2012 21:32 CaptainApe wrote:Show nested quote +On March 04 2012 11:52 ZeroTalent wrote:
Why are PvP and ZvZ so random? Why is TvT so not random? Of the non-mirror matches, why is ZvP the most random? I have no idea what the answers to these questions are. Any conjectures? First of all, thank you for these stats, they are very insightful! 3) addon switching tech switches are very fast for a terran. All he has to do (in most cases) is to switch the addons of the buildings. Need some tanks? Take tech lab on your your fax and start production immediatly! Need more marines? Take a reactor on the rax a.s.o....
No. Terrans have the worst tech switching. Zerg can mass produce as soon as their tech building is up and Protoss can warp in what they need at will. Also Terran upgrades are the most separated between infantry, mech and air, which also hampers techs switches.
|
This is what I've been telling people for soo long. A lot of pros play terran because we have the most reliable MUs. Where other races and their matchups have more luck involved.
Even tho it's clear that T pros have a harder time beating Z and P scrubs than P Z pros have beating T scrubs. Tells a lot about the game kekeke.
|
On March 12 2012 21:32 CaptainApe wrote:Show nested quote +On March 04 2012 11:52 ZeroTalent wrote:
Why are PvP and ZvZ so random? Why is TvT so not random? Of the non-mirror matches, why is ZvP the most random? I have no idea what the answers to these questions are. Any conjectures? First of all, thank you for these stats, they are very insightful! I can´t say that much about P and Z since i play Terran, but it seems to me, that both Zerg and Protoss have several major disadvantages compared to terran, especially in the early/early-mid game. 1) Terrans have marines Marines and marine/marauder are good against pretty much everything in the early game. They are cheap, quickly build and have a lot of dps. on top of that, they scale very good with upgrades (stim, combat shields, weapon/armor upgrades and medivac heal). So its very unlikely to die to something unexpected. 2) scans If all scouting fails for a terran, he can almost always scan. That helps a lot to identify some all in/cheese. 3) addon switching tech switches are very fast for a terran. All he has to do (in most cases) is to switch the addons of the buildings. Need some tanks? Take tech lab on your your fax and start production immediatly! Need more marines? Take a reactor on the rax a.s.o.... 4)M.U.L.E.S. Even when your economy is behind your oponents, mules help you to come bach into the game. As long mules "are" energy on the orbital, you cannor kill them as long as you dont kill the orbital, which is quite hard in early game. Therefore, mules excuse mistakes of a terran player much more than chrono and especially larva inject. They allow you to play an all in with scvs and still not be completely all in. Again, this is an ability, that buys time for a terran in all match ups. all in all, i think terrans have more/easier to use tools for a better player/pro to not die in the early and mid game. That helps to get to a phase of the game where the pro can benifit from his/her better mechanics/macro/micro/multitasking.
I don't think you understand the game beyond your bias.
1) That's true in some cases, but the poll shows that P and Z pros have an EASIER time playing against T scrubs than T have against P and Z scrubs meaning that marines are not only not OP (or if they are other factors make them the only thing keeping Terran alive) and that something else must be keeping alive P and Z early game against the in your eyes invincible marines soo giving marines to P and Z would not make their mirrors less random (Remove the warp and give zerg a few more years to focus on ZvZ(and yeah literally years because that MU is so hard early game))
2) I agree that keeps Terran alive in TvT, but it must not be the only reason in TvZ and TvP because SOMEHOW P and Z have it better without scan. Soo again better scouting would fix PvP, but in ZvZ you pretty much know what's coming with overlord spread again it's just such a hard MU people need a lot of time for it.
3) Defending early pressure this only counts if you 1/1/1 and that's not a good eco build soo that's just bullshit and as I've said 2 times now this must somehow Z and P are pretty okay without this against worse players.
4) It's a mechanic for gods sake they removed their effect on gold so stop complaining about mules. I hate delusional people who believe Mules are somehow an answer to everything. If protoss chronos enough to get a 8 probe lead (this usually happens about a 6-8 probe lead for protoss and building the orbital gives protoss a 2 probe lead anyway) terran and protoss are on even ground. A mule equals the mining of 4 probes and if done correctly there should be 1 mule per OC so they P and T should be equal. Soo when you see a terran all in with 12 SCVs and he has 8 SCVs left at home that means he sacrificed 50% of his economy and he knows that if you want to come out ahead you have to end up with a 5 probe advantage (assuming you have an equal army count at the end)
If anyone wants to disagree go ahead just don't do it in a stupid manner.
Edit: I want to clear this up I'm not saying T is UP I'm just saying that giving what Terran has to Protoss and Zerg won't fix their MUs. Remove warp and obviously buff T1 units because there is no warp (PvP fixed) give more time to zerg (ZvZ fixes itself eventually). I've got no idea how to fix PvZ tho.
|
I think a lot of this might have to do with how easily one race can side step the other races' defenders advantage.
Just a thought on one contributor to the three most Scrub favored match ups, and three most Pro favored match ups.
PVP
Warpates
ZVZ
Lings are fast and banelings can be morphed in anywhere
PVZ
Warpgate again, but forcefields also help a lot in removing the Zergs ability to reinforce even when he is in his own base
--------------------------
TvT
Tanks are great defense..
Tanks and Mech are slow...
PvT
Again tanks are great defense
ZvT
Again Tanks?
Seems all three of the least Scrub favored matches are against Terran
By having a stronger defenders advantage or one which you cannot reduce its effectiveness enough forces the game to go longer and forces more mistakes. Pros make less mistakes, in say PvP a loss will be caused by less mistakes than a TvT, the more mistakes you have to make the less likely that the pro will make that many first.
|
On March 25 2012 22:22 lolphind wrote:
PvT
Again tanks are great defense
Whaaaaaa. Since when are tanks used in PvT.
Here is something interesting
MLG winter arena
1st place MKP
2nd place DRG
3rd place Huk
MLG winter championship
1st place MKP
2nd place DRG
4th place Huk
ohh and Arena MKP 4-2 DRG Championship MKP 5-2. That's pretty consistent isn't it?
|
On March 04 2012 13:26 GhostFall wrote:I believe this picture explains how mirrow matchups work + Show Spoiler +The defensive building is normally tied to the teching build for Terran and Zerg. Without bunkers or sunkens, ZvZ and TvT would be very similiar to PvP. But because cannons are tied to the forge, we have many build order wins/losses. Starcraft BW got around this problem by having the shield battery, a weaker defensive building tied to the teching building - the gateway. You need to get a defender's advantage from somewhere. ZvZ really is a knife fight. Even if you're better with a knife you're still on a razor's edge. Plus, the level of execution you need is very high, so among lower level players, it's essentially a luck based matchup.
oh my god this image is amazing, thank you.
|
|
|
|
|
|
EPT![[image loading]](http://i.imgur.com/YWnCb.jpg)
