Game's a lot more balanced than Forum QQ would have you believe from the looks of things. But then again I knew that before I saw these stats.
I wonder what it is about Zergs in Korea vs the rest of the world. The overall Master's/GM win rate for that race in particular is much different than the others.
On September 23 2011 07:03 Vindicare605 wrote: Game's a lot more balanced than Forum QQ would have you believe from the looks of things. But then again I knew that before I saw these stats.
I wonder what it is about Zergs in Korea vs the rest of the world. The overall Master's/GM win rate for that race in particular is much different than the others.
tournament results back up the qq more than they back up those statistics
On September 23 2011 04:59 Wuster wrote: I think you could argue that the bottom of Code S really isn't that different from the middle/top of Code A. And the balance in Code A has steadily gotten better, while before it was the opposite, with Code S having the better ratio and Code A being Terran dominant. So it make sense that Code S would eventually become Terran dominant, just like looking at Code A today gives hope that Code S will again diversify.
Well I'm not sure the "diversity" in code A will last long, we don't even know if there will be a protoss from code A in the up/down.
Hence me saying hopeful, we know already that there are only going to be 2 or 3 terrans advancing to up/downs this season. Protoss has their own issues with Zerg, true, but that points more to Protoss being UP than Terran being OP (I mean there's no T in PvZ so what else can you conclude if P sucks in PvT and PvZ in GSL anyways).
Funny enough all 4 Protoss in the Ro16 are facing Zergs.
On September 23 2011 05:10 sitromit wrote:
On September 23 2011 04:59 Wuster wrote:
On September 23 2011 04:42 sitromit wrote:
On September 23 2011 04:36 Namu wrote:
On September 23 2011 04:34 sitromit wrote:
On September 23 2011 03:35 willz22912 wrote: I'd just like to chime in for everyone who is trying to use GSL as a sort of balance discussion topic. While GSL is generally accepted as the highest level SC2 tournament in the world, with the best players, it can't be used exclusively on it's own without support as some unquestionable fact.
The fact is that people seem to forget that the GSL is in itself not perfect as a tournament or as a structure. Do we not all forget the open seasons and the difference it is in terms of skill level today, it took a long time for certain "average" players who everyone didn't think belonged in code S to fall down with the up/down format.
Remember how the ro32 is set up for both Code A and S. Before in Code S, the bottom two in groups were always sent into the up and downs with 1 person from code A coming up, with the top 2 of those 3 players making it back into Code S. Now the turnover is better for code A -> S but still the same for Code S. There is still only 8 people out of a 32 man tournament that could potentially lose Code S every season. Then there's code A Ro32 which is just a bo3 and if you lose you are knocked into code B which everyone knows is insanely difficulty to get back into code A.
My final point regarding the GSL is the group structure in ro32 code S. While people cannot handle multiple bo3s because of time constraints, this leaves us with the much worse outcome in terms of player potential, 2 bo1s and a 3rd potentially for a tiebreaker. Will you not agree that a single bo1 is not enough of a factor to determine who is the better player? Nerves aside in a tournament setting, this also increases the chances of cheese being seen in people's strategies since it's only a bo1 and if you cheese well, you generally have a high chance of just straight up winning which is why cheese is so popular. And I think most people would generally agree that a cheese game is 1) not fun for viewers, 2) not indicative of a players true skill. Remember IMmvp in GSL March? He got baneling busted twice in a row by July in an "upset" and fell to up and downs, and then lost in up and downs to two protosses and fell into code A. Now look at him, 3 time GSL champion with another high chance of winning a 4th title. Do people not slump and do weird results happen in the group stages. Yes and yes. Don't keep using GSL as a stand alone #s game since even out of 20 Terrans, at most 8 can go into up and downs every season, so getting the GSL to be racially balanced will take a long time regardless of game balance because of the tournament format.
This makes absolutely no sense and here's why. If everything were as random as you say, ie Bo1s not a good indicator who is better, by skill or by power of having picked the right race, all races would fare equally well. It's random after all, a matter of chance. How is it that Terran always seems to get lucky? Why has Terran representation in the GSL consistently gone higher and higher every season?
No, what he is saying is that even if the game becomes balanced now after 1.4 or subsequent patches, it will take more than one season, probably like 3 for the terran number to go down to the appropriate ammount. He's saying that the number of races in GSL will lag behind the actual current balance.
We already know that. But there's nothing in the balance changes that actually significantly effects Terran. So we're stuck with this situation for the foreseeable future, and for a good while afterwards, even if Terran eventually gets balanced.
Blue flame hellion and Barrack build time weren't significant nerfs? GSL is still playing with the old patch, let's wait and see before we complain that nothing changed for Terran.
If you want to see the balance shifting, you have to look at Code A, and of the qualifiers, only 2 were Terran who faced a TvT to get in. Code A is much more diverse overall as well. While, as stated before Code S is set up to keep people in it (Artosis mentioned that having consistent players in Code S lets Gom give them a surrogate salary).
I think you could argue that the bottom of Code S really isn't that different from the middle/top of Code A. And the balance in Code A has steadily gotten better, while before it was the opposite, with Code S having the better ratio and Code A being Terran dominant. So it make sense that Code S would eventually become Terran dominant, just like looking at Code A today gives hope that Code S will again diversify.
Just look at the list of players who have gone from Code S to Code B, I'll skip the Protoss because so many people will insist they fell out due to imbalance. Maka, Leenock, Byun, Kyrix, TheWind, Boxer (if he's not Code A quality what does that mean to everyone who placed behind him at MLG and NASL?), Jinro (a former MLG champion), ect. Code S and A don't have a huge skill divide except maybe at the very top (MVP cruising through Code A when he got knocked down), so it's not all doom and gloom.
BF Hellion nerf only really effects TvT. They're worse against Marines now, but they're still just as effective vs. Zerglings, and it only takes a +1 upgrade for them to two-shot Drones like before.
And 5 in game seconds is going to effect what exactly? 11/11? Come on... Terrans in GSL out-macroing Zerg and trading armies with incredible cost efficiency until the Zerg can't keep up? 5 seconds is going to change that?
You'd be surprised, 5 seconds was enough to kill Reapers after all. Getting +1 for your BFH sounds trivial, but that requires an armory, meaning to get the same 2-shotting on drones you are investing a lot, also delaying things quite a bit.
Think of the time/resource investement especially if your plan is Bio instead of Mech. I would certainly expect much fewer BFH drops to harass while massing up a huge bio-ball ala MGL Anaheim.
The reason reapers died off isn't because of any 5 second timing change. It's because reaper speed now requires a factory and you require a supply depot before you can build a barracks...
That was a month later and the final nail in the coffin. But reaper openings were very rare after the 5 second went in, since the timings for reaper openings were already very slim.
On September 23 2011 05:40 jdseemoreglass wrote: When it comes to "measuring" balance, Blizzard has some of the most ass-backwards logic I've ever heard in my life.
Every game that's played on battlenet is a function of two dependent variables: skill and balance. If you don't know one, it is mathematically impossible to calculate the other. This is an undeniable common sense fact.
If you don't know what the balance is, you can't calculate a player's relative "skill." After you've falsely and incorrectly assumed you know what a players skill is, you can't use that number to somehow assess the balance of the game. It makes no fucking sense!
Am I the only one who sees this? Let me break it down to kindergarten level for you...
Let's say Idra and MC play 10 games together. MC wins 6, Idra wins 4.
Now... did MC win more because of balance or because of skill? YOU CAN'T KNOW THAT SIMPLY FROM THE WIN/LOSS STATISTICS! No matter how many millions of games were played, it would be absolutely impossible to know if MC was winning more because of skill or balance. You cannot calculate EITHER with win ratios. And so long as they are both dependent and unknown variables, you won't ever, ever, be able to figure them out independently.
Is Blizzard really this stupid, or do they assume everyone else is?
I know this will probably be buried under piles of "ZOMG BLIZZARD IS BLIND AND WE'RE ALL SMARTER!" but there is a way to determine the value of 2+ unknown variables. It's called systems of equations. Here is a link:
In this case, Blizzard can take some variables to measure, like average skill, skill deviation, and racial balance. You combine these with knowns, like race, winner, and rating before game. The end equation USUALLY looks something like this:
Ax+By+Cz=Q
The capital letters are the numbers you know, and the small letters are the variables. You do this over a long period of time and you end up with a multitude of numbers for your knowns. Since this involves behavioral unpredictability and an imperfect system, you're going to get a bunch of equations that can't be solved by systems of equations. This is where things like linear algebra and statistics comes in with a least squares approach to determine the best fit approximations for each variable. The beauty of this approach isn't just that you can determine things like overall balance, but you can also track balance across skill levels and see learning curves of each race, or population/skill shifts as they occur.
In short, please shut up about stuff you don't know. The second you think something is "common sense," stop yourself and do a little research into the subject to make sure it really is common sense.
It is impossible to solve for two variables in a single equation. It is possible if you have multiple independent equations. And we don't.
Even though the games played are all different, the resulting equation is the same no matter how many times you run it.
ie. Skill + Balance = Result.
It doesn't matter how many players, how many games, at some point you are going to have to make an assumption about balance or player skill. There is nothing in statistics or any other kind of mathematics that will be able to solve two dependent variables in a single equation. So long as the ONLY metric being used is win ratio, then Blizzard only has a single equation.
In short, please don't tell me to shut up. What I'm talking about is common sense, and basic mathematics. Blizzard puts a big complicated equation on the board and people's eyes glaze over, but at the end of the day it is nothing more than two variables in a single equation.
On September 23 2011 05:52 ChriseC wrote: there are some things that i dont understand
ladder system tries to keep you to 50% win/lose ratio, so isnt it representitive at all?
AHHHHHHHHHHHHHH
There really should be a link that is required viewing for every new TL member to the Blizzcon in which they explain their way of taking away that. Oh, and the squillions of other posts that ask exactly the same question.
Basically, Blizzard has a fucking huge equation to remove the matchmaking system from the equation (allegedly), among other things. So NO, the 50% win/lose ratio SHOULDN'T have any impact.
Without the formula to examine, there's no fucking reason to believe that.
You can find the formula, look for Blizzcon 2010 then search for part 3(?) of the panel section, AFAIK, then pause when they show the formula and take it down. Nope, it meant literally nothing to me either.
At 2.50.
It looks (to me) like a Bayesian-esque sort of analysis.
The basic process appears iterative; they start with a prior distribution (probably flat) and produce a posterior distribution by feeding it the result of a game, by using the win-percentage probability distributions for each player, for each race. Note that this doesn't need to know any true sense of skill, it just needs to know the outcome of the match and the predicted win percentages for those players - the latter of which are obviously accurate as it uses them for match-making.
The process is repeated over and over, adding the results of all the thousand (millions?) of matches taking place on Battlenet, with the previous posterior distribution becoming the new prior distribution each time the process is run. Eventually one hopes the distribution converges on a distribution from which the win percentages can then be simply extracted, as a simple function of league level.
I will have to think about this some more, but this makes sense and I think should work.
I don't think they are trying to 'scare' anyone away, as some other posters seem to imply. I think they don't want to have to wheel their math-guy out to give a series of long, dry seminars to explain Bayesian analysis. Personally, given how the match-maker seems to work pretty well, I'm inclined to take their word for this process working to the level of accuracy (+/-5%) that they claim.
it makes some sense...not the equation but the simple example that they gave....but it doesn't seem liek it would work in practice
they said they don't take into account how the player wins....perhaps the million zerg player have cheesed their way to victory.....that doesn't make the matchup balanced or say much about the skills of the player. 2 it doesn't take into account that some players are better against certain races or racial playstyles than others, meaning that there isn't a static "skill" that they can look at. example: idra is great at the long game, so if you play against him and play a macro game and lose, assuming races are balanced, that means hes better, however if you rush him early and he loses that doesn't mean that you are a better player, just better against that long view style. 3. it doesn't take into account pure build order losses, i don't really see how it could.
you can't take into account all the variables that affect a player and put them into a formula, its not possible. it also doesn't take into account that most of the pro's don't actually ladder for practice they ladder to refine a build or get a new build down.
You are partially correct. Cheese wont be take into consideration here. This winning percentage estimation system is only result based.
But I do believe they have another system to solve the strategy imbalance. In all, they have all the data. They can do whatever they want. And data dont lie. Yet here majority of the members just make conclusion based on their instinct.
id argue that they are making conclusions based on actual results
Well, u got to trust the 12k+ paid Ph.D's from Blizzard cuz they are suppose to be much better in mathematics than the community member here.
Ppl always thought they are smarter, but 99.9% of them cant figure out why they just cant have a A+ on their statistic class
On September 23 2011 07:03 Vindicare605 wrote: Game's a lot more balanced than Forum QQ would have you believe from the looks of things. But then again I knew that before I saw these stats.
I wonder what it is about Zergs in Korea vs the rest of the world. The overall Master's/GM win rate for that race in particular is much different than the others.
They don't use infestors as much, that's probably the main reason.
On September 23 2011 05:40 jdseemoreglass wrote: When it comes to "measuring" balance, Blizzard has some of the most ass-backwards logic I've ever heard in my life.
Every game that's played on battlenet is a function of two dependent variables: skill and balance. If you don't know one, it is mathematically impossible to calculate the other. This is an undeniable common sense fact.
If you don't know what the balance is, you can't calculate a player's relative "skill." After you've falsely and incorrectly assumed you know what a players skill is, you can't use that number to somehow assess the balance of the game. It makes no fucking sense!
Am I the only one who sees this? Let me break it down to kindergarten level for you...
Let's say Idra and MC play 10 games together. MC wins 6, Idra wins 4.
Now... did MC win more because of balance or because of skill? YOU CAN'T KNOW THAT SIMPLY FROM THE WIN/LOSS STATISTICS! No matter how many millions of games were played, it would be absolutely impossible to know if MC was winning more because of skill or balance. You cannot calculate EITHER with win ratios. And so long as they are both dependent and unknown variables, you won't ever, ever, be able to figure them out independently.
Is Blizzard really this stupid, or do they assume everyone else is?
I know this will probably be buried under piles of "ZOMG BLIZZARD IS BLIND AND WE'RE ALL SMARTER!" but there is a way to determine the value of 2+ unknown variables. It's called systems of equations. Here is a link:
In this case, Blizzard can take some variables to measure, like average skill, skill deviation, and racial balance. You combine these with knowns, like race, winner, and rating before game. The end equation USUALLY looks something like this:
Ax+By+Cz=Q
The capital letters are the numbers you know, and the small letters are the variables. You do this over a long period of time and you end up with a multitude of numbers for your knowns. Since this involves behavioral unpredictability and an imperfect system, you're going to get a bunch of equations that can't be solved by systems of equations. This is where things like linear algebra and statistics comes in with a least squares approach to determine the best fit approximations for each variable. The beauty of this approach isn't just that you can determine things like overall balance, but you can also track balance across skill levels and see learning curves of each race, or population/skill shifts as they occur.
In short, please shut up about stuff you don't know. The second you think something is "common sense," stop yourself and do a little research into the subject to make sure it really is common sense.
It is impossible to solve for two variables in a single equation. It is possible if you have multiple independent equations. And we don't.
Even though the games played are all different, the resulting equation is the same no matter how many times you run it.
ie. Skill + Balance = Result.
It doesn't matter how many players, how many games, at some point you are going to have to make an assumption about balance or player skill. There is nothing in statistics or any other kind of mathematics that will be able to solve two dependent variables in a single equation. So long as the ONLY metric being used is win ratio, then Blizzard only has a single equation.
In short, please don't tell me to shut up. What I'm talking about is common sense, and basic mathematics. Blizzard puts a big complicated equation on the board and people's eyes glaze over, but at the end of the day it is nothing more than two variables in a single equation.
have to say, you are not understanding their equation/method at all.
balance is not a variable. your equation is far more wrong.
On September 23 2011 05:52 ChriseC wrote: there are some things that i dont understand
ladder system tries to keep you to 50% win/lose ratio, so isnt it representitive at all?
AHHHHHHHHHHHHHH
There really should be a link that is required viewing for every new TL member to the Blizzcon in which they explain their way of taking away that. Oh, and the squillions of other posts that ask exactly the same question.
Basically, Blizzard has a fucking huge equation to remove the matchmaking system from the equation (allegedly), among other things. So NO, the 50% win/lose ratio SHOULDN'T have any impact.
Without the formula to examine, there's no fucking reason to believe that.
You can find the formula, look for Blizzcon 2010 then search for part 3(?) of the panel section, AFAIK, then pause when they show the formula and take it down. Nope, it meant literally nothing to me either.
It looks (to me) like a Bayesian-esque sort of analysis.
The basic process appears iterative; they start with a prior distribution (probably flat) and produce a posterior distribution by feeding it the result of a game, by using the win-percentage probability distributions for each player, for each race. Note that this doesn't need to know any true sense of skill, it just needs to know the outcome of the match and the predicted win percentages for those players - the latter of which are obviously accurate as it uses them for match-making.
The process is repeated over and over, adding the results of all the thousand (millions?) of matches taking place on Battlenet, with the previous posterior distribution becoming the new prior distribution each time the process is run. Eventually one hopes the distribution converges on a distribution from which the win percentages can then be simply extracted, as a simple function of league level.
I will have to think about this some more, but this makes sense and I think should work.
I don't think they are trying to 'scare' anyone away, as some other posters seem to imply. I think they don't want to have to wheel their math-guy out to give a series of long, dry seminars to explain Bayesian analysis. Personally, given how the match-maker seems to work pretty well, I'm inclined to take their word for this process working to the level of accuracy (+/-5%) that they claim.
it makes some sense...not the equation but the simple example that they gave....but it doesn't seem liek it would work in practice
they said they don't take into account how the player wins....perhaps the million zerg player have cheesed their way to victory.....that doesn't make the matchup balanced or say much about the skills of the player. 2 it doesn't take into account that some players are better against certain races or racial playstyles than others, meaning that there isn't a static "skill" that they can look at. example: idra is great at the long game, so if you play against him and play a macro game and lose, assuming races are balanced, that means hes better, however if you rush him early and he loses that doesn't mean that you are a better player, just better against that long view style. 3. it doesn't take into account pure build order losses, i don't really see how it could.
you can't take into account all the variables that affect a player and put them into a formula, its not possible. it also doesn't take into account that most of the pro's don't actually ladder for practice they ladder to refine a build or get a new build down.
You are partially correct. Cheese wont be take into consideration here. This winning percentage estimation system is only result based.
But I do believe they have another system to solve the strategy imbalance. In all, they have all the data. They can do whatever they want. And data dont lie. Yet here majority of the members just make conclusion based on their instinct.
id argue that they are making conclusions based on actual results
Well, u got to trust the 12k+ paid Ph.D's from Blizzard cuz they are suppose to be much better in mathematics than the community member here.
Ppl always thought they are smarter, but 99.9% of them cant figure out why they just cant have a A+ on their statistic class
12k+ phd's and 6 years of development should have yielded a game that was almost completely balanced already
On September 23 2011 05:52 ChriseC wrote: there are some things that i dont understand
ladder system tries to keep you to 50% win/lose ratio, so isnt it representitive at all?
AHHHHHHHHHHHHHH
There really should be a link that is required viewing for every new TL member to the Blizzcon in which they explain their way of taking away that. Oh, and the squillions of other posts that ask exactly the same question.
Basically, Blizzard has a fucking huge equation to remove the matchmaking system from the equation (allegedly), among other things. So NO, the 50% win/lose ratio SHOULDN'T have any impact.
Without the formula to examine, there's no fucking reason to believe that.
You can find the formula, look for Blizzcon 2010 then search for part 3(?) of the panel section, AFAIK, then pause when they show the formula and take it down. Nope, it meant literally nothing to me either.
It looks (to me) like a Bayesian-esque sort of analysis.
The basic process appears iterative; they start with a prior distribution (probably flat) and produce a posterior distribution by feeding it the result of a game, by using the win-percentage probability distributions for each player, for each race. Note that this doesn't need to know any true sense of skill, it just needs to know the outcome of the match and the predicted win percentages for those players - the latter of which are obviously accurate as it uses them for match-making.
The process is repeated over and over, adding the results of all the thousand (millions?) of matches taking place on Battlenet, with the previous posterior distribution becoming the new prior distribution each time the process is run. Eventually one hopes the distribution converges on a distribution from which the win percentages can then be simply extracted, as a simple function of league level.
I will have to think about this some more, but this makes sense and I think should work.
I don't think they are trying to 'scare' anyone away, as some other posters seem to imply. I think they don't want to have to wheel their math-guy out to give a series of long, dry seminars to explain Bayesian analysis. Personally, given how the match-maker seems to work pretty well, I'm inclined to take their word for this process working to the level of accuracy (+/-5%) that they claim.
it makes some sense...not the equation but the simple example that they gave....but it doesn't seem liek it would work in practice
they said they don't take into account how the player wins....perhaps the million zerg player have cheesed their way to victory.....that doesn't make the matchup balanced or say much about the skills of the player. 2 it doesn't take into account that some players are better against certain races or racial playstyles than others, meaning that there isn't a static "skill" that they can look at. example: idra is great at the long game, so if you play against him and play a macro game and lose, assuming races are balanced, that means hes better, however if you rush him early and he loses that doesn't mean that you are a better player, just better against that long view style. 3. it doesn't take into account pure build order losses, i don't really see how it could.
you can't take into account all the variables that affect a player and put them into a formula, its not possible. it also doesn't take into account that most of the pro's don't actually ladder for practice they ladder to refine a build or get a new build down.
You are partially correct. Cheese wont be take into consideration here. This winning percentage estimation system is only result based.
But I do believe they have another system to solve the strategy imbalance. In all, they have all the data. They can do whatever they want. And data dont lie. Yet here majority of the members just make conclusion based on their instinct.
id argue that they are making conclusions based on actual results
Well, u got to trust the 12k+ paid Ph.D's from Blizzard cuz they are suppose to be much better in mathematics than the community member here.
Ppl always thought they are smarter, but 99.9% of them cant figure out why they just cant have a A+ on their statistic class
You are correct sir but I don't think they judge everything based on statistics.
On September 23 2011 07:03 Vindicare605 wrote: Game's a lot more balanced than Forum QQ would have you believe from the looks of things. But then again I knew that before I saw these stats.
I wonder what it is about Zergs in Korea vs the rest of the world. The overall Master's/GM win rate for that race in particular is much different than the others.
tournament results back up the qq more than they back up those statistics
You realize that Tournament results come from a ridiculously small pool of players right?
Blizzard's sample size in those stats is made up of literally MILLIONS of games played. You can't argue with that with results pooled from less than a couple hundred games played by less than 100 total players. There's just way too many variables to take into consideration and not nearly large enough of a sample size to draw any real conclusions from.
It does mean that the races are fairly balanced at a non-pro level. It's useful information in that case, if ever anyone decides that the game can't be changed at the highest level because it's already so hard to do [x] in lower leagues.
On September 23 2011 07:05 Bashion wrote: what scares me is that Blizz can use those Protoss numbers as an excuse to nerf the race even more.
What scares me are imagining the Z numbers in America/EU after the patch.
Acknowledging that Terrans received two nerfs one of them which cut Blue Flame damage by 50% which was the core of a strategy that skewed the win rate in favor of Terrans in TvZ, while Protoss have a longer Blink research time which was the core of a strategy of theirs, I don't think you're going to see the Zerg win rate drop nearly as much as you'd think.
On September 23 2011 07:13 Amaranthine wrote: September 13? this data is taken from one day of games? that's not enough information to claim the match-ups are balanced...
Montlhy statistics I could take more seriously, but only one specific day...?
That's when the data was compiled. They don't say when they began or stopped collecting the data they merely mean they compiled and analyzed it on that day.
You really think all of those numbers are based on one day only? You really have no faith in Blizzard to do a thorough job do you?
On September 23 2011 07:05 Bashion wrote: what scares me is that Blizz can use those Protoss numbers as an excuse to nerf the race even more.
What scares me are imagining the Z numbers in America/EU after the patch.
Acknowledging that Terrans received two nerfs one of them which cut Blue Flame damage by 50% which was the core of a strategy that skewed the win rate in favor of Terrans in TvZ, while Protoss have a longer Blink research time which was the core of a strategy of theirs, I don't think you're going to see the Zerg win rate drop nearly as much as you'd think.
That's true, I had forgotten blink was delayed heavily. I guess we'll see how it pans out; I like that Blizzard release the numbers even if (or maybe because? ) it always causes chaos at the TL forums.
EPT
) it always causes chaos at the TL forums.
