|
I completely agree here with what Pelican has added to the discussion, and his given conclusion. I also want to note that what we don't see in the raw data is also some of the omitted variable bias that can play a role in the numbers, themselves. I'm not just speaking to values we cannot see, but also values that are rather hard to quantify. Such as that of differences in the goals of individual players on ladder. For instance, some will use ladder as a means of practice, while others will use it for experimentation. Unless you can justly quantify how this would affect percentages, or control for that in the data between the races, I don't see how you could logically come up with an argument.
The data also does not, to what I have seen, take into account what the players are matched up with. Such that there is a possibility that zerg mainly face protoss instead of mirror or zvt, and thus the data is biased in that the win percentage may be speaking about the zvp match-up instead of Zerg on the whole.
The data would also require trying to control for the differences in the ability of the individual. I would be so bold as to say that some players are simply better than others, in areas such as mechanics, game sense, and decision making. How can you control for the differences of the individual in trying to analyze the abilities of the races?
In summation, the data is too clouded, or rather, not collected enough in order to properly account the statistical relevance. If you would be able to embark upon the statistics with a team of people, and try to control for variable bias, then you may have an argument. As it currently stands, there is no such possibility. I like the idea, but the execution is a little poor.
|
Isn't GM based on last season?
|
On April 26 2011 02:13 Durp wrote:Show nested quote +On April 26 2011 02:06 tehemperorer wrote:
Well, from a Protoss perspective, if you are in GM and you get a lot of PvP matches in one day, it seems you could really either skew your score or drop in points dramatically because the matchup is very volatile (coinflip BO loss people tend to say).
To be a top player you have to be dominant in every matchup, not just 2/3, and since PvP isn't figured out yet, top players might still go 50/50 in that matchup, whereas a Terran player can be really strong in their mirror and can win them most of the time? It seems that the Terran mirror is the best mirror because of the differing ways it can be played and still won. QFT. I think this explains a lot towards terran being ahead in terms of all races, since it has the most explored and least volatile mirror matchup. In TvT, moreso than any other mirror, the better player tends to win.
Bullshit. MC has the highest PvP winrate out there. He's the better player, like it or not.
|
I wouldn't say terran is dominating, they only have about 1% win rate higher than everyone else, and only about 20 points more than everyone else. Yes terrans at the top are doing better, but far from dominating.
|
On April 26 2011 04:48 Xirroh wrote:
What your research shows is that right now the best Terran's are able to reach and stay at the top of their ladder better then the top Protoss and Zerg. Essentially they are more Stable.
I think this is because the TvT match-up is the most stable and generally allows the better player to win, where as right now PvP is a guessing game, and ZvZ is always crazy.
If you look at BW players, the ones that stay at the top of the rankings always dominate their mirror match-up. The best BW ZvZ, TvT and PvP in the last few years have usually been Jaedong, Flash, and Bisu respectively. The correlation between staying at a high ranking and having a good mirror win rate is an often overlooked factor.
Right now it's hard for P and Z to have a high mirror win rate because the match-ups are not as stable as TvT.
Bullshit again. P and Z mirrors require a different kind of skill is all. Just like BW ZvZ. Just because some people suck at it doesn't mean it's not a good matchup.
|
On April 26 2011 06:41 Etrnity wrote: the omitted variable bias
Given that the causal realtionships in social sciences are very specualative and that results are often contradictive even in fields with a great amount of prior reseach and established theoretical framework, we can't have too much hope of creating a model of imbalance for starcraft 
How can we even begin to measure imbalance when there isn't even a agreed upon definition of what it is?
|
It's about the same. Nothing really crazy happening, I don't think you'll ever get 1/3 of all stats to each race, just because of different amounts of representation at different levels. It often times doesn't mean anything about balance or anything else, it usually just shows the amount of players at that level playing a certain race.
|
I know in the OP you only gave figures using grandmaster/top % of players, but I really feel like you cannot come to the kind of conclusion you have solely based on these statistics.
Would it not be fair to suggest that there are in fact more Terran players total than any other race currently competing in the ladder? I think it's also reasonable to say that the more players playing 1 specific race *could* lead inevitably to that race creating more top level players in the long term. I don't know whether this is actually true, but I think if it were plausible, you should find out approx the ratio of players playing with each race.
Just throwing it out there, there are so many variables that you need to think about if you want to be taken seriously on this matter.
|
I won't speak to balance or statistics because I don't have any qualification to do so. What I will say is that from personal experience, reading strategy forums and viewing tournaments and streams is that terran does have a larger amount of variety in every matchup from player to player and even sometimes a single player.
As it appears to me that many zergs and protoss play in a smaller range of variety in their builds and styles. The differences are much more subtle. In pvt I've seen some great mech play, bio play, sky terran play (banshees, raven into BC type play), along with scary all ins and greedy macro play. Even within these groupings of play styles their are many variations and possible opening and transitions. With mech for example there is ghost mech, tank heavy mech, thor heavy mech. With bio I typically see the variation as drops frequency, sometimes bio/early emp, bio with tanks or thors or PDD. In other words what the terran supports his bio with. These variations make can make all the difference in whether or not a particular response by the opponent will suffice.
Against zerg we see many of the same variations in play from terran, bio with tanks, drops, thors, some mech play. The variety of openings like banshee harass, hellion harass, 2 rax bunker rushes are very wide as well.
Almost pure marine has been shown to be viable at even the highest levels of play in every terran matchup. I've seen events where a single terran player has success with 3 totally different styles in 3 games of the same series.
I don't think anyone could argue that the flexibility and synergy between units that terran possesses is strong. It may really boil down to the fact that protoss and zerg players haven't yet mastered how to deal with every possibility terran can throw at them because the game is still young. (there are 2 expansions and many patches to come!) Perhaps these numbers will simply balance out with time, when top protoss and zerg players better understand the timings, weaknesses and responses to the wide variety of play terran offers and have the game down to a science. Maybe in the expansions there will be a single new upgrade or unit that unlocks worlds of possibilities to zerg and protoss. Imagine if hellions had spider mines, immortals had a range upgrade, or hydras could morph into lurkers. A single new upgrade to a unit in the expansion could start a revolution in matchups.
That's my guess anyway, it's just the variety of terran and the learning curve to dealing with it. With many changes coming to SC2 in the future it really doesn't matter if the game is balanced now. Does anyone really expect the game to be perfectly balanced at every stage of patching and through the release of expansions?
Also, I'd like to point out that I think ladder isn't necessarily worse to look at for statistics than tourneys(as some people in this thread have pointed out.) It is my guess that tournaments results are affected greatly by outside factors, for example can they hear the commentators or crowds? Are they jet lagged? Are they eating food they are not used to? Did they get a good nights sleep? Are they nervous? Did they play enough games today to warm up? Many of these things would not be a big factor to a player sitting in his room where he is comfortable, playing on ladder.
Also, OP, I would love to see the same statistics for the GM league as a whole (not just top 100 or 200) and divided by regions as well. If the same slight differences and averages hold true when divided by region and throughout GM league, I think it would be a stronger demonstration of any slight imbalances. Even though these kinds of variations are to be expected, we're in for a long haul of sc2 changes!
I hope I don't get flamed for this post, my intention is just to make some helpful observations. Overall the balance of the game is pretty damn good when considering how hard most player work to abuse anything they can in the game to win.
|
On April 26 2011 06:41 Etrnity wrote:
I completely agree here with what Pelican has added to the discussion, and his given conclusion. I also want to note that what we don't see in the raw data is also some of the omitted variable bias that can play a role in the numbers, themselves. I'm not just speaking to values we cannot see, but also values that are rather hard to quantify. Such as that of differences in the goals of individual players on ladder. For instance, some will use ladder as a means of practice, while others will use it for experimentation. Unless you can justly quantify how this would affect percentages, or control for that in the data between the races, I don't see how you could logically come up with an argument.
The data also does not, to what I have seen, take into account what the players are matched up with. Such that there is a possibility that zerg mainly face protoss instead of mirror or zvt, and thus the data is biased in that the win percentage may be speaking about the zvp match-up instead of Zerg on the whole.
The data would also require trying to control for the differences in the ability of the individual. I would be so bold as to say that some players are simply better than others, in areas such as mechanics, game sense, and decision making. How can you control for the differences of the individual in trying to analyze the abilities of the races?
In summation, the data is too clouded, or rather, not collected enough in order to properly account the statistical relevance. If you would be able to embark upon the statistics with a team of people, and try to control for variable bias, then you may have an argument. As it currently stands, there is no such possibility. I like the idea, but the execution is a little poor.
I don't think that Pelican's analysis is relevant; the ladder tries to normalize win-loss ratios. Let's make some assumptions:
1. Z really is underpowered
2. The GM league requires players to have a certain amount of skill at the game to get in
3. Overall skill distribution across the races is roughly equal.
The result is that every player in GM is going to have a similar amount of skill; something above the threshold. Now, if Z is underpowered, their "skill" is reduced; that is, a Z has to be better than a P or T to beat them. So, what do you get from that? You'd expect two things:
1. Less Z and more P and T in GM league, because it's easier for them to get in; they need less "skill."
2. Roughly equal win rates across the races, because P and T skill equals Z skill, after accounting for the imbalance (Z are better overall, but the imbalance makes them play worse relative to others)
Hey, guess what you see if you look at the GM league statistics?
|
Maps in the ladder favor terran obviously, which is why no tournament worth a damn uses only ladder maps.
|
I think it's really naieve to say that it's just a coincidence that Terran has always held the highest position in the ladder since beta (look how many of these threads have come up over the months -_-), and that that has nothing to do with some combination of their mechanics+defensibility+having more units than any other race.
|
how many here think that that top one random is Day9s smurf?
It has to be
|
On April 26 2011 02:17 Shaetan wrote:Show nested quote +On April 26 2011 02:06 buldermar wrote:On April 26 2011 02:02 Shaetan wrote:
You can't claim to do statistical analysis then just throw out numbers and claim that X is true, you have to do the analysis. Can you examplify? I thought I did a decent job at showig a trend based on every aspect of the top of the ladder I could think of. I even specifically pointed out that I'd gladly look into more if anyone got ideas. You need to show that the difference you see is statistically significant and not just due to chance. You said yourself when looking at a sample size of ~1600 win rates were much more even so it's possible decreasing your sample size artificially created the disparity. Also not sure why you wouldn't look at the full 200 or at least all GM league top 100.
Yea I'm gonna agree with this, as well as the poster above him.
Basically, your saying that race is not independent of [array of different sc2 statistics]. If race was independent there would be no relationship, and thus, a balanced game. The problem is that you cannot just claim this, you must prove it using a valid test. I'm not an expert, but i think a chi-squared test works here. So:
+ Show Spoiler +
I divided the points by 10 so it would look better, but it doesn't change the result. Yes Terran seems to have a higher general ladder ranking, but is actually losing more than expected and all without any significance.
+ Show Spoiler +
These are the actual numbers, and according to this test, the data is not statistically significant, doesn't not fit model and so we fail to reject the null hypothesis, which is that race is independent of win/loss, points ect.
Note that while p ~= 0 the chi squared is way to high for the df.
Not very confident in my work here, feel free to point out what i got wrong.
|
On April 26 2011 07:09 imareaver3 wrote:Show nested quote +On April 26 2011 06:41 Etrnity wrote:
I completely agree here with what Pelican has added to the discussion, and his given conclusion. I also want to note that what we don't see in the raw data is also some of the omitted variable bias that can play a role in the numbers, themselves. I'm not just speaking to values we cannot see, but also values that are rather hard to quantify. Such as that of differences in the goals of individual players on ladder. For instance, some will use ladder as a means of practice, while others will use it for experimentation. Unless you can justly quantify how this would affect percentages, or control for that in the data between the races, I don't see how you could logically come up with an argument.
The data also does not, to what I have seen, take into account what the players are matched up with. Such that there is a possibility that zerg mainly face protoss instead of mirror or zvt, and thus the data is biased in that the win percentage may be speaking about the zvp match-up instead of Zerg on the whole.
The data would also require trying to control for the differences in the ability of the individual. I would be so bold as to say that some players are simply better than others, in areas such as mechanics, game sense, and decision making. How can you control for the differences of the individual in trying to analyze the abilities of the races?
In summation, the data is too clouded, or rather, not collected enough in order to properly account the statistical relevance. If you would be able to embark upon the statistics with a team of people, and try to control for variable bias, then you may have an argument. As it currently stands, there is no such possibility. I like the idea, but the execution is a little poor. I don't think that Pelican's analysis is relevant; the ladder tries to normalize win-loss ratios. Let's make some assumptions: 1. Z really is underpowered 2. The GM league requires players to have a certain amount of skill at the game to get in 3. Overall skill distribution across the races is roughly equal. The result is that every player in GM is going to have a similar amount of skill; something above the threshold. Now, if Z is underpowered, their "skill" is reduced; that is, a Z has to be better than a P or T to beat them. So, what do you get from that? You'd expect two things: 1. Less Z and more P and T in GM league, because it's easier for them to get in; they need less "skill." 2. Roughly equal win rates across the races, because P and T skill equals Z skill, after accounting for the imbalance (Z are better overall, but the imbalance makes them play worse relative to others) Hey, guess what you see if you look at the GM league statistics?
It would be nice if you had the ability to read. His argument is to actually use statistical tests in order to determine if the numbers are significant. My argument is that we are assuming too much, even with using more advanced statistical tools. All you have done is provided assumptions as controls without justification (Broad based assumption that all GM players are of equal skill is horrific to begin with).
You're trying to counter my argument by saying that assuming = controlling, which is absolutely ludicrous.
|
the question here is not why there are so much terrans on GM... but who the hell is that mega baller top 20 random player because GOD random is so hard.
|
On April 26 2011 07:09 imareaver3 wrote:Show nested quote +On April 26 2011 06:41 Etrnity wrote:
I completely agree here with what Pelican has added to the discussion, and his given conclusion. I also want to note that what we don't see in the raw data is also some of the omitted variable bias that can play a role in the numbers, themselves. I'm not just speaking to values we cannot see, but also values that are rather hard to quantify. Such as that of differences in the goals of individual players on ladder. For instance, some will use ladder as a means of practice, while others will use it for experimentation. Unless you can justly quantify how this would affect percentages, or control for that in the data between the races, I don't see how you could logically come up with an argument.
The data also does not, to what I have seen, take into account what the players are matched up with. Such that there is a possibility that zerg mainly face protoss instead of mirror or zvt, and thus the data is biased in that the win percentage may be speaking about the zvp match-up instead of Zerg on the whole.
The data would also require trying to control for the differences in the ability of the individual. I would be so bold as to say that some players are simply better than others, in areas such as mechanics, game sense, and decision making. How can you control for the differences of the individual in trying to analyze the abilities of the races?
In summation, the data is too clouded, or rather, not collected enough in order to properly account the statistical relevance. If you would be able to embark upon the statistics with a team of people, and try to control for variable bias, then you may have an argument. As it currently stands, there is no such possibility. I like the idea, but the execution is a little poor. I don't think that Pelican's analysis is relevant; the ladder tries to normalize win-loss ratios. Let's make some assumptions: 1. Z really is underpowered 2. The GM league requires players to have a certain amount of skill at the game to get in 3. Overall skill distribution across the races is roughly equal. The result is that every player in GM is going to have a similar amount of skill; something above the threshold. Now, if Z is underpowered, their "skill" is reduced; that is, a Z has to be better than a P or T to beat them. So, what do you get from that? You'd expect two things: 1. Less Z and more P and T in GM league, because it's easier for them to get in; they need less "skill." 2. Roughly equal win rates across the races, because P and T skill equals Z skill, after accounting for the imbalance (Z are better overall, but the imbalance makes them play worse relative to others) Hey, guess what you see if you look at the GM league statistics?
I wasn't trying to draw any conclusions from the stats that I ran, I was just trying to illustrate that drawing conclusions from numbers and statistical data isn't as simple as most of the posts I see on these forums would have one believe. Actual analysis has to be done to see if there are any real differences between sets of data. I'm not claiming that X or Y race is under/overpowered, I'm just saying that IF it's so important to look at these data in an attempt to answer that question, it should be done correctly. Edit: As correctly as possible, given the limited (quantitatively and qualitatively) data that we have, as Etrnity suggests.
|
it was said on the first page
terran is just more resilient against cheese
|
The fact that Terran "dominates" the ladder doesnt give way to anything about race balance, race potential, etc.
It could simply be the fact that there are more terran users than protoss or zerg users.
|
On April 26 2011 09:15 JerKy wrote:
The fact that Terran "dominates" the ladder doesnt give way to anything about race balance, race potential, etc.
It could simply be the fact that there are more terran users than protoss or zerg users.
Read the original post? There are more protoss (38%) than terran (30%).
|
|
|
|
|
|
EPT![[image loading]](http://imgur.com/f3PEA.jpg)
![[image loading]](http://imgur.com/LXWQR.jpg)
