|
On October 27 2010 11:13 domovoi wrote:Show nested quote +On October 27 2010 10:28 Tray wrote:
Come on man really? You being wrong about statistics is going to get me banned again.
You're right that we would expect evenly skilled players to beat one another 50% of the time. Of course. Where you're wrong is saying that you expect higher than race representaion at the top for zerg. Sadly, you don't even bother to try to explain why this is the case. You simply make the assertion "it's more probable skill is evenly distrubuted at the top." This is just wrong, and I will tell you why.
You're reverse extrapolating the data to say that everyone in the top 200 has about the same skill, therefore each race should have 33% of those top 200. But you're completely ignoring the fact that only 1/5 of players play Zerg. Take a simple example of 100 players. Doesn't matter their races. Lets divide up their skill into 5 groups and use the same ones as Blizzard divisons so there's 20 bronze, 20 silver, 20 gold, 20 plat, and 20 diamonds. If only 20% of those players play Zerg, how many would you expect to be in Diamond if the game was perfectly balanced? It's 4. 20% of 20. If 10 of the 20 diamond players were Zerg you would say that Zerg must be overpowered because everyone is the same skill, yet they represent 50% of all players in the top bracket.
I hope that helps clear things up for you. I don't get anymore polite than that so before you reply, you better make sure your stats are sound. Before they were not. I'm not using "statistics.
Then everything you say is invalid.
|
|
|
On October 27 2010 11:29 MementoMori wrote:Show nested quote +On October 27 2010 10:18 domovoi wrote:
In fact, it's far more likely that the skill distribution is fairly even at the top level, which would mean the 66th best Zerg is about the same skill level as the 66th best Terran and 66th best Protoss. At least, I can't think of any reason why that wouldn't be the case. This would lead to equal representation among the top 200.
It would actually be quite problematic if only 20% of the top 200 were Zerg. If that indicated good balance, it would imply the 40th-best Zerg was as skilled as the 80th-best Terran and Protoss. This doesn't sound likely to me. It actually is true man. Think of it like this... if 20% of people play zerg and skill level is equal across all races then those smaller number of zergs will be spread out somewhat evenly over all the more common players of the other two races. Lets say #1, 2, 3 are terran, protoss and zerg respectively. You can't expect 4,5, and 6 to also be terran, protoss and zerg because there are just less people playing zerg (as an example). If there are less people playing a race and skill distribution is even across all races, the 50th best zerg will be worse than the 50th best terran. Edit: I just want to say that what I said isn't true if you're trying to say that the best players switch races to what they think is the best. What I said is only true if the races truly have an even distribution of skill.
The reason you have 20% of people playing zerg in the total population, is because you have things like, 70% playing terran in the bottom of the bottom in the bronze leagues, because it IS all they played. They did their five placement matches with the game, and are now finished with it.
This large population of casual players who will play less than 50 games in their lives (and probably, due to the campaign, picked terran), is throwing off a big chunk of this number.
I feel like if you took a poll of ACTIVE players, and compared race distribution, you'd see a more even spread.
And you have all the bronze and silver and gold guys riding terran to try to ride the "imba" wave they read about on the blizz forums, and are just flavor of the month kids from WoW. I'd say give it til just after Christmas / new years, and they'll all slowly gravitate back towards their cookie cutter rpgs and leave us be
|
Then everything you say is invalid.
Thanks for the worthless comment. Some things simply can't be proven by statistics, especially with limited data; anyone who knows a modicum about the subject could tell you that.
|
On October 27 2010 14:38 domovoi wrote:Thanks for the worthless comment. Some things simply can't be proven by statistics, especially with limited data; anyone who knows a modicum about the subject could tell you that.
Thankfully this is not one of them. This is how Blizzard balanced the races. You can sit there and say you have to do it off of intuition all day, but it will never be correct.
Take it or leave it but clearly you're done trying to learn, so I'm done trying to teach.
|
On October 27 2010 09:33 Tray wrote:
I see there's still people that think balance = all races with equal representation.
This is incorrect. Balance is obtained by equal representation relative to total players who play that race. This assumes skill to be distributed evenly amongst players, which is the only assumption you can reasonably make.
Alternatively you can also look at race by race matchup win rates.
Based on the figures here, if backed up by Blizzard's numbers, Zerg is in for a nerf. 20% race represetation with ~30% in the top 200 after a week and a half? Unless a gigantic number of people changed to Zerg, the balance of power shifted heavily in the last patch.
I think you're really missing a lot of important issues here that really have to be accounted for.
What population is important? Is Bronze population important? Is only Diamond population important? If only Diamond population is important, what if one race has an easier time getting in to diamond?
Couldn't the actual dynamics of a matchup have an impact on the top 200? What if one race has 2 coin flip matchups while another has 3 matchups where luck rarely plays a part?
Do you assume the activity of all the races is even? Does a patch make certain races more active? Regarding this most recent patch, I was practicing my Z on the ladder and I had to go back to Terran because there was just too many ZvZs.
I believe there has been a pretty large population shift to zerg over the last few weeks.
http://sc2ranks.com/stats/league/all/1/all/
That also doesn't account for players that are playing zerg now but are listed as something else because they haven't played enough games yet.
|
On October 28 2010 00:47 RivalryRedux wrote:Show nested quote +On October 27 2010 09:33 Tray wrote:
I see there's still people that think balance = all races with equal representation.
This is incorrect. Balance is obtained by equal representation relative to total players who play that race. This assumes skill to be distributed evenly amongst players, which is the only assumption you can reasonably make.
Alternatively you can also look at race by race matchup win rates.
Based on the figures here, if backed up by Blizzard's numbers, Zerg is in for a nerf. 20% race represetation with ~30% in the top 200 after a week and a half? Unless a gigantic number of people changed to Zerg, the balance of power shifted heavily in the last patch. I think you're really missing a lot of important issues here that really have to be accounted for. What population is important? Is Bronze population important? Is only Diamond population important? If only Diamond population is important, what if one race has an easier time getting in to diamond? Couldn't the actual dynamics of a matchup have an impact on the top 200? What if one race has 2 coin flip matchups while another has 3 matchups where luck rarely plays a part? Do you assume the activity of all the races is even? Does a patch make certain races more active? Regarding this most recent patch, I was practicing my Z on the ladder and I had to go back to Terran because there was just too many ZvZs. I believe there has been a pretty large population shift to zerg over the last few weeks. http://sc2ranks.com/stats/league/all/1/all/That also doesn't account for players that are playing zerg now but are listed as something else because they haven't played enough games yet.
If the population shift matches the increase in representation at the top, then that could be an indicator of correct balance. I doubt that many people have changed, but we will see when Blizzard releases the numbers eventually.
Of course the matchups impact the representation. We call it "balance." If a race is winning more than 50% we expect them to be highly represented and thus "overpowered."
You can also make an argument that the balance changes shift who plays what race, but that's covered in the first paragraph with overall racial representation.
Leagues don't matter. Blizzard has stated that race representation is almost the same across all leagues. If you were trying to figure out balance though you would only want to look at the Top "x" players where x is the threshold where you have decided that skill has nearly peaked out and racial imbalances make up most of the win rates. Blizzard has stated this point is Plat/diamond for an appropriate amount of data to mull over.
So yeah you bring up some good points and some stuff we've covered. It's never as simple as just population to representation, but with our limited data, and without Blizzard's extensive data, that's as close as we can come as outside observers.
|
The big thing worth pointing out the statistical noise in the top 200 is about sqrt(200), or 7%. So even if you have 33% Zerg in the top 200, that could be consistent with anything from 40% Zerg (OMG so OP!) to 26% (much closer to racial distribution).
Small number statistics can cause all sorts of spurious conclusions.
|
On October 28 2010 01:42 Dragar wrote:
The big thing worth pointing out the statistical noise in the top 200 is about sqrt(200), or 7%. So even if you have 33% Zerg in the top 200, that could be consistent with anything from 40% Zerg (OMG so OP!) to 26% (much closer to racial distribution).
Small number statistics can cause all sorts of spurious conclusions.
Haha correct. I just find it funny because when I said this exact same thing when everyone was saying T was overpowered because of their representation I kept hearing the invalid and incorrect argument that, "there is no variability! We're looking at the ENTIRE population!"
|
Really nice job by ReSpOnSe - hope to see top 20 soon!
I wonder what the cutoff for GM league will be.. ;o
|
Yes, it's a fine line between getting enough data to make statistically robust conclusions, while only including the top-level players from which the data is really relevent.
|
If we assume that every top player maximizes his amount of wins, players would be evenly distributed between the three races on top level. I think we could indeed apply such a model to the balance discussion, so I'd aim for equal racial distribution in tournaments since the quarter-finals and in top 200 ladder games.
I also have a question. When my division is very strong, do I gain more points or less?
|
On October 28 2010 02:58 Perscienter wrote:
If we assume that every top player maximizes his amount of wins, players would be evenly distributed between the three races on top level. I think we could indeed apply such a model to the balance discussion, so I'd aim for equal racial distribution in tournaments since the quarter-finals and in top 200 ladder games.
I also have a question. When my division is very strong, do I gain more points or less?
I'm pretty sure division doesn't matter at all, because Blizzard stated so ^^
|
On October 28 2010 02:58 Perscienter wrote:
If we assume that every top player maximizes his amount of wins, players would be evenly distributed between the three races on top level. I think we could indeed apply such a model to the balance discussion, so I'd aim for equal racial distribution in tournaments since the quarter-finals and in top 200 ladder games.
I also have a question. When my division is very strong, do I gain more points or less?
Why do you think assuming maximizing wins that each race would be equally represented at the top? That's entirely false. It should match ONLY the % of players who play that race, making the assumption skill is evenly distributed. See my very simple example on the previous page if you're having a hard time wrapping your head around this concept.
And no your division doesn't matter at all.
|
It does matter and was stated on page 5 and 6.
|
On October 28 2010 03:09 Perscienter wrote:
It does matter and was stated on page 5 and 6.
You're confusing division weight when calculating the top 200 with a normalizer that in theory (it doesn't exist) would change the point totals when people of different divisions play one another. This doesn't happen.
It's a direct MMR/Ratings comparison between the two players regardless of their individual divisions. Everyone gets the same bonus pool regardless of when you are placed.
|
Of course the matchups impact the representation. We call it "balance." If a race is winning more than 50% we expect them to be highly represented and thus "overpowered."
You missed my point. I wasn't talking about the win %, I was talking about the way a matchup plays. By coin flip I meant a BO game where skill doesn't play as much of a role in the outcome. If TvP/PvT was determined by flipping a coin then it would be a 50/50 matchup. However, in order to separate yourself from other players and climb the ladder you would have to do so in your other 2 matchups. If on the other hand zerg had no coin flip matchups then top zergs would be able to separate themselves from the field in all 3 matchups. This would impact the way the ladder shakes out even though all of the matchups could be 50/50 with regards to win %.
You can also make an argument that the balance changes shift who plays what race, but that's covered in the first paragraph with overall racial representation.
That can create a contradiction to your assumption that skill levels are evenly distributed. If players choose to play a race they perceive is strong or that they will win with, then you also have to factor in how perception is different among different players. If top players have a different perception than others, their racial distribution may reflect this.
|
On October 28 2010 03:23 RivalryRedux wrote:Show nested quote +Of course the matchups impact the representation. We call it "balance." If a race is winning more than 50% we expect them to be highly represented and thus "overpowered." You missed my point. I wasn't talking about the win %, I was talking about the way a matchup plays. By coin flip I meant a BO game where skill doesn't play as much of a role in the outcome. If TvP/PvT was determined by flipping a coin then it would be a 50/50 matchup. However, in order to separate yourself from other players and climb the ladder you would have to do so in your other 2 matchups. If on the other hand zerg had no coin flip matchups then top zergs would be able to separate themselves from the field in all 3 matchups. This would impact the way the ladder shakes out even though all of the matchups could be 50/50 with regards to win %. Show nested quote +
You can also make an argument that the balance changes shift who plays what race, but that's covered in the first paragraph with overall racial representation.
That can create a contradiction to your assumption that skill levels are evenly distributed. If players choose to play a race they perceive is strong or that they will win with, then you also have to factor in how perception is different among different players. If top players have a different perception than others, their racial distribution may reflect this.
Okay I follow your first part now. I guess you're saying that you need to see race by race matchups to see if one matchup is skewing the results. Makes sense.
As to your second paragraph - That's why we assume skill to be evenly distributed. Blizzard's numbers "take skill into account," which is more accurate, but we can't do that because we don't have access to the matchmaking algorithm.
|
On October 28 2010 02:58 Perscienter wrote:
I also have a question. When my division is very strong, do I gain more points or less?
I made a thread on this a while ago and I think that if your in a power division you will accumulate less points than you would in a weaker division.
http://www.teamliquid.net/forum/viewmessage.php?topic_id=146009
When top 200's come out, if you thumb through them, you'll see how players in those power divisions tend to move up on the ladder relative to their total points and the opposite happens for players in weaker divisions.
When Excalibur_Z asked the guy at blizzcon he mentioned "and the skill of your division" when he was saying that you can't directly compare point among divisions. I wasn't entirely sure though if meant points or rank (everyone already knows rank).
|
On October 28 2010 03:34 RivalryRedux wrote:Show nested quote +On October 28 2010 02:58 Perscienter wrote:
I also have a question. When my division is very strong, do I gain more points or less? I made a thread on this a while ago and I think that if your in a power division you will accumulate less points than you would in a weaker division. http://www.teamliquid.net/forum/viewmessage.php?topic_id=146009When top 200's come out, if you thumb through them, you'll see how players in those power divisions tend to move up on the ladder relative to their total points and the opposite happens for players in weaker divisions. When Excalibur_Z asked the guy at blizzcon he mentioned "and the skill of your division" when he was saying that you can't directly compare point among divisions. I wasn't entirely sure though if meant points or rank (everyone already knows rank).
I read most of this, but I don't see why the answer isn't "Point changes are based on MMR, not rating." Since there's no way you can know what your opponents' MMRs are, I don't think you can disprove this possibility.
I still don't see how/why divisions should/would have weight. Who you play is not linked to your division at all so this implies that, if true, you are better off waiting to qualify for a week then playing your games after a reset. This would be a giant abusable bug if true.
|
|
|
|
|
|
EPT
