EPTGame Balance and The Uncertainty Principle
| Forum Index > SC2 General |
|
jdseemoreglass
United States3773 Posts
| ||
|
darmousseh
United States3437 Posts
| ||
|
jdseemoreglass
United States3773 Posts
On January 21 2011 03:00 darmousseh wrote: They can find the correlation between skill and balance by looking at results of random players, unfortunately none of these exist at the top level. Some math guy there probably knows this, but doesn't say anything because then their entire idea of balance would be destroyed. You are making the assumption that random players are equally skilled at all three races. Sometimes people are simply naturally better at one race, or simply have a better strategic understanding of one over another. | ||
|
dragonsuper
Liechtenstein222 Posts
| ||
|
iSTime
1579 Posts
But yay for attracting lay people and fabricating credibility with fancy science words! EDIT: On topic. New Blizzard project, project make SC3 balanced! Blizzard needs to adopt several sets of newborn triplets from some poor country, raise them on a private island where they will all be given the exact same amount of schooling and attention, and are forced to spend the same amount of time as one another playing video games and such. Then, each triplet will be assigned one Z one P and one T. Only with such an objective test, where all players have the same experiences (AND DNA!) can we truly know that they are equally skilled, and so if the Z players all do badly we can fix the balance. | ||
|
mierin
United States4943 Posts
On January 21 2011 03:13 PJA wrote: Unlike the uncertainty principle, the more accurately you can measure skill or balance, the more accurately you can measure the other one. But yay for attracting lay people and fabricating credibility with fancy science words! Seriously...I saw balance and quantum mechanics in the same sentence and facepalmed  | ||
|
nekuodah
England2409 Posts
| ||
|
storm8ring3r
Germany227 Posts
| ||
|
searcher
277 Posts
[Edit: Actually they do use the normal distribution as you can see from the equation they show in the video lol.] | ||
|
jdseemoreglass
United States3773 Posts
On January 21 2011 03:13 PJA wrote: Unlike the uncertainty principle, the more accurately you can measure skill or balance, the more accurately you can measure the other one. But yay for attracting lay people and fabricating credibility with fancy science words! You are missing the point that they cannot be accurately measured in relation to eachother. The more we assume that the game is balanced, the more inaccurate our measure of player skill becomes, and the more we assume player skill is equal, the more inaccurate the measure of balance becomes. | ||
|
D10
Brazil3409 Posts
| ||
|
jdseemoreglass
United States3773 Posts
On January 21 2011 03:16 mierin wrote: Seriously...I saw balance and quantum mechanics in the same sentence and facepalmed That is what an analogy is... Comparing two generally unrelated concepts to provide a quicker path to understanding a point or message. *facepalm | ||
|
iSTime
1579 Posts
On January 21 2011 03:17 jdseemoreglass wrote: You are missing the point that they cannot be accurately measured in relation to eachother. The more we assume that the game is balanced, the more inaccurate our measure of player skill becomes, and the more we assume player skill is equal, the more inaccurate the measure of balance becomes. No, I'm not missing the point. His point is that without first knowing the skill level of the players, we cannot accurately judge balance. This is the exact opposite of how the uncertainty principle behaves. Why does OP bring the uncertaintly principle, which is a massively flawed analogy, into his discussion at all? To garner false credibility from people who see fancy science words and go GOOD POST OP!!1!. EDIT: Furthermore, the OP thinks that blizzard determines skill level based solely on win rate. Blizzard actually determines an approximate skill level by using win rate combined with a bunch of statistical tests (mostly just some Bayesian interference stuff) which work with a few assumptions. | ||
|
Hittegods
Stockholm4641 Posts
| ||
|
searcher
277 Posts
On January 21 2011 03:19 D10 wrote: I agree, and thats why the map makers kept BW balanced, the community is way more attuned to what actually matters in a high level game than the developers broad perspective approach. Well as mentioned in the video Blizzard keeps track of community thoughts on balance. Also, they speak to top-level pros, and I would have to imagine they are more attuned than the community as a whole. | ||
|
jdseemoreglass
United States3773 Posts
On January 21 2011 03:20 PJA wrote: No, I'm not missing the point. Why does OP bring the uncertaintly principle, which is a massively flawed analogy, into his discussion at all? To garner false credibility from people who see fancy science words and go GOOD POST OP!!1!. http://www.teamliquid.net/forum/viewmessage.php?topic_id=186260#12 But good job guys on totally derailing a thread into oblivion by the first page by completely ignoring the point of the post and offering splitting hair criticisms. | ||
|
iSTime
1579 Posts
On January 21 2011 03:24 jdseemoreglass wrote: http://www.teamliquid.net/forum/viewmessage.php?topic_id=186260#12 But good job guys on totally derailing a thread into oblivion by the first page by completely ignoring the point of the post and offering splitting hair criticisms. Except anyone who actually knows what the uncertainty principle is is just going to say "why the fuck did you bring up the uncertainty principle, when all it does is obfuscate your point?" Is it so much to ask that people just say what they mean instead of constantly trying to create false credibility with bullshit jargon? EDIT: If anyone wants to read a real balance post with meaningful information, http://www.teamliquid.net/forum/viewmessage.php?topic_id=186103. Honestly, you're insulting me for derailing a thread when OP is posting exactly what was posted in this other thread, except minus any real information or insightful analysis. | ||
|
Drunken.Jedi
Germany446 Posts
On January 21 2011 03:24 jdseemoreglass wrote: http://www.teamliquid.net/forum/viewmessage.php?topic_id=186260#12 But good job guys on totally derailing a thread into oblivion by the first page by completely ignoring the point of the post and offering splitting hair criticisms. No one is derailing your thread, all people did was pointing out that the uncertainty principle is not applicable here, because the exact opposite is the case: the better game balance is known, the easier it is to derive player skill and vice versa. This is not hair splitting, this is pointing out that about half of your OP is factually wrong and confusing. That said, I agree with the second part of the OP, Blizzard seem to be oblivious to this problem. My personal theory is that they know full well that statistical approaches to determining balance are very flawed but they just pretend that they have some sort of way to figure out "adjusted win percentages" to make it seem that the game is more balanced than it actually is. | ||
|
Tiazi
Netherlands761 Posts
| ||
|
MoreFaSho
United States1427 Posts
| ||
|
Obstbaum
Switzerland224 Posts
On January 21 2011 02:55 jdseemoreglass wrote: The Heisenberg Uncertainty Principle is a well-known theory of quantum mechanics which states that it is not possible to precisely measure a pair of inequalities, such as the position and momentum of a particle. Because the pair of inequalities are dependent upon eachother, they cannot be simultaneously derived. The more precise that one property is measured, the less precise the other becomes. What does this have to do with Starcraft? The answer can be found in this video of Blizzard discussing game balance, and their methods and strategies for balancing the game. The pair of inequalities in this situation is "skill" and "game balance." Blizzard is failing to recognize that it is simply impossible to determine one mathematically without also knowing the other in advance. You cannot determine the skill of a player using winrate statistics, without FIRST knowing the degree of balance in the game being tested. Likewise, it is impossible to determine the balance of a game using winrate statistics with FIRST knowing the relative skill of the players. The two variables are mutually dependent upon eachother, making it impossible to determine one without the other. im amused about your comparison, but does not really match up to your opinion about blizz` analysis. Dont worry about the general gamebalance, blizzard`s got this easily, they are easily able to balance 99.9% of sc2. I guess you personally will never really encounter REAL inbalances. The problem lies with a fast changing, map specific metagame of the pros. This will have to be balanced with mappools (huge change coming up in gsl btw) and i think blizzard should look into which maps are on the ladder (Look at steps or blistering for example..horrible maps). Gamebreaking things have and will ofc be changed by blizz, when discovered (1 food roach, bugs +co). PS: sc2uncertainty principle: deltaskill * deltawinrate =~ dustinbrowderconstant... | ||
|
Rodeo
United States39 Posts
On January 21 2011 03:16 mierin wrote: Seriously...I saw balance and quantum mechanics in the same sentence and facepalmed The uncertainty principle means that the more certain you are what the principle means the less you actually understand it, right? ^_^ Balance and skill aren't complementary variables, so the analogy doesn't really work. I think the reason people are mad about this is that they don't like seeing "sciencey" concepts evoked when they don't apply. You could just as well have used the theory of relativity (which would probably be a better analogy since what really lacks here is a universal frame of refrence, though it would still be annoying to people who actually care what these things mean). As for the content, I have a big long post agreeing with the basic idea. | ||
|
BenKen
United States860 Posts
| ||
|
Excludos
Norway8140 Posts
Can anyone who has actually played on both EU, NA and Korea server please point your viewpoints about this? EDIT: did they just say that stimmed Marauders are borderline useless? EDIT2: And then the same guy stated 2 minutes later that they are going to look at stim that tend to be too powerful early on.. | ||
|
Muff2n
United Kingdom250 Posts
2. Maths is a very powerful tool. If statistics of 10000s of games shows a balance improvement, its a balance improvement. Much better than the opinions of players as it is free from bias and missunderstanding. 3. You can tell skill by relative comparison. You have a very large population of players of a chosen race and can see their win percentages. Assumptions have to be made between races but these are reasonable with such a large population. Blizzard's stated methodology of balance information is fine. It just seems that they don't realise how the maps are susceptable to balance issues. If a zerg can see a push and then make sunkens in time, thats a gigantic advantage compared to having to have sunks in preparation for a push that might come. Standardise the maps and then balance the races. | ||
|
Neo.NEt
United States785 Posts
So Blizzard's brilliant idea is flawed... but how can they fix it? They can't. What else are they supposed to do? | ||
|
Lennon
United Kingdom2275 Posts
| ||
|
clickrush
Switzerland3257 Posts
On January 21 2011 03:51 Excludos wrote: What annoys me most of all, and the asian guy stated this in the interview, is that blizzard look at korea as the "better" players, and thus balance the game thereafter. Now what kind of evidence do they have to support that idea? Sure there arent many international players in the GSL, but there is not a lot of korean players in the MLG either. Can anyone who has actually played on both EU, NA and Korea server please point your viewpoints about this? EDIT: did they just say that stimmed Marauders are borderline useless? EDIT2: And then the same guy stated 2 minutes later that they are going to look at stim that tend to be too powerful early on.. no. they said if they nerf stim on marauders then stimming marauders becomes even harmfull in some situations. | ||
|
Drunken.Jedi
Germany446 Posts
On January 21 2011 07:12 Muff2n wrote: 2. Maths is a very powerful tool. If statistics of 10000s of games shows a balance improvement, its a balance improvement. Much better than the opinions of players as it is free from bias and missunderstanding. 3. You can tell skill by relative comparison. You have a very large population of players of a chosen race and can see their win percentages. Assumptions have to be made between races but these are reasonable with such a large population. Maybe it's just my ignorance speaking, but I just don't see how this can be possible. If you want to measure balance, you have to consider the win rates for each race while adjusting for player skill, but to measure player skill you have to consider the win rates while adjusting for balance. You cannot do both so if you want to measure balance through win rates you have to make the assumption that player skill is equal at any MMR level, which is only a sensible assumption if the game is balanced, thus invalidating the whole approach. | ||
|
Subversion
South Africa3627 Posts
Blizzard has access to an enormous amount of stats, they have top-level players, and gosu mathematicians and formulas all trying to calculate this shit. And they're doing a pretty good job so far. I'm sorry but I don't think any random Tl.net poster, who doesn't have access to the things Blizzard does, and is most likely way less qualified, can teach them about balancing. | ||
|
mierin
United States4943 Posts
On January 21 2011 08:29 Drunken.Jedi wrote: Maybe it's just my ignorance speaking, but I just don't see how this can be possible. If you want to measure balance, you have to consider the win rates for each race while adjusting for player skill, but to measure player skill you have to consider the win rates while adjusting for balance. You cannot do both so if you want to measure balance through win rates you have to make the assumption that player skill is equal at any MMR level, which is only a sensible assumption if the game is balanced, thus invalidating the whole approach. Actual balance discussion can't come from "MMR" and ladder...it's pointless, it's like Blizzard looking at your average D+/C- bnet games and making changes about balance. The only real stuff that needs to be looked at is the highest of the high level play...like PL and individual leagues in BW, and GSL / top western tournaments (eh maybe not), and so on in SC2. | ||
|
Stratos_speAr
United States6959 Posts
On January 21 2011 02:55 jdseemoreglass wrote: The Heisenberg Uncertainty Principle is a well-known theory of quantum mechanics which states that it is not possible to precisely measure a pair of inequalities, such as the position and momentum of a particle. Because the pair of inequalities are dependent upon eachother, they cannot be simultaneously derived. The more precise that one property is measured, the less precise the other becomes. What does this have to do with Starcraft? The answer can be found in this video of Blizzard discussing game balance, and their methods and strategies for balancing the game. http://www.youtube.com/watch?v=9OYTt_8zYHI&feature=related The pair of inequalities in this situation is "skill" and "game balance." Blizzard is failing to recognize that it is simply impossible to determine one mathematically without also knowing the other in advance. You cannot determine the skill of a player using winrate statistics, without FIRST knowing the degree of balance in the game being tested. Likewise, it is impossible to determine the balance of a game using winrate statistics with FIRST knowing the relative skill of the players. The two variables are mutually dependent upon eachother, making it impossible to determine one without the other. Blizzard claims to understand this problem, and they claim to fix it by factoring skill into the equation. But how is the skill of the player known? Well, by their winrate statistics! They tried to solve the problem and ended up falling right back in! How Blizzard failed to see this common sense error before they began parading their fancy formulas on stage puzzles me. Perhaps they know only too well that their mathematical attempts at balance are impossible, but they hope not enough people recognize why. The simple truth of the matter is that skill level can never be determined by winrate without the assumption of game balance. Perhaps it is possible for players to recognize skill in other factors... Multi-tasking, decision making, micro, economy management, and so on. The problem is, most of these things cannot be scientifically measured. Blizzard doesn't even try. For this reason, the communities assessment of a player's skill, and therefore the balance of the game, will always have a more rational and accurate foundation than the one that Blizzard is using. You're idea is faulty because you only take into account pure statistics. However, we (and Blizzard) have the ability to watch people play and see how they play, so even if they lose, we can see if someone is good at a game but just lost due to something being ridiculously imbalanced or if they actually messed up or whatnot. Statistics can't tell you that. What annoys me most of all, and the asian guy stated this in the interview, is that blizzard look at korea as the "better" players, and thus balance the game thereafter. Now what kind of evidence do they have to support that idea? Sure there arent many international players in the GSL, but there is not a lot of korean players in the MLG either. Can anyone who has actually played on both EU, NA and Korea server please point your viewpoints about this? EDIT: did they just say that stimmed Marauders are borderline useless? EDIT2: And then the same guy stated 2 minutes later that they are going to look at stim that tend to be too powerful early on.. It's not that difficult to come to the conclusion that Koreans, in general, are better at Starcraft. Their entire culture idolizes it - professional Starcraft players are as popular as sports athletes. You say that you play Starcraft in the west, and many people will just give you a weird look. This promotes a lot more Koreans to play and get better at it compared to western players. | ||
|
creamfilling
United States4 Posts
That being said blizz seems to have done a pretty good job of balancing the game in a relatively short period of time. On balance I think most of the crazy imba early game strats have been fixed at this point. | ||
|
MicroJFox
United States38 Posts
Balance based on racial win rates of random players only. This completely factors out player skill (the same player is playing all races), leaving only racial balance. | ||
|
azzu
Germany141 Posts
On January 21 2011 03:21 Hittegods wrote: Skill could be measured from mirror matchups? Which will help Blizzard so much on finally balancing mirror matchups! ... Wait what?  If someone needs an explanation why this is so funny: mirror matchups of course are perfectly balanced. So when you let MVP play 1000 games against Jinro, you maybe know which one of these players is more skilled. But it doesn't matter, since mirror matchups are already balanced. How are you going to determine who's more skilled: Nestea or Jinro? Let them play 1000 TvTs and then say Jinro is more skilled? or 1000 ZvZs and then say Nestea has more skill? | ||
|
Treehead
999 Posts
On January 21 2011 03:00 darmousseh wrote: They can find the correlation between skill and balance by looking at results of random players, unfortunately none of these exist at the top level. Some math guy there probably knows this, but doesn't say anything because then their entire idea of balance would be destroyed. Statistically, you can find correlation. The hard part is determining causation. You seem to gloss over this point. Let's say we do some statistical analysis and find that in PvZs, Z tends to win more, and their win rate has been climbing since 1.1.3. What can we derive from that? Well, as it turns out, all that stats can tell us is that this is happening - you can't really calculate that it absoultely MUST be because of zergling's speed. That just isn't possible. Now, if zerglings with speed were cost effective against all protoss units with sufficient micro, regardless of tactics/strategy chosen by the protoss, then math/theorycraft can tell you that, provided you are able to generate an absolutely exhaustive list of tactics/strategies somehow. But really, you'd have to be able to teach a computer to play SC2 as well as a top level player, teach it to learn, teach it to vary up it strategies, and then have it play itself a million times. Even then, I think you'd find that the computer's decision making (e.g. what units to make in response to scouting a particular tactic/composition) became cyclical as time went on, since I have generally found there is at least one unit each race has available which is at least slightly cost effective against a particular tactic. Also, a slight quibble on semantics - this is more like statistics than it is like math. Mathematics is much more abstract, much less applicable and much, much cooler. I'm just saying, as someone with a graduate degree in math, that some mathematicians will bristle at being expected to know statistics and vice versa. | ||
|
jalstar
United States8198 Posts
The two variables are mutually dependent upon eachother, making it impossible to determine one without the other. Very wrong, these kinds of equations are still solvable. Example: x = (y + 4)/2 y = (x + 8)/2 You can use substitution or an iterative method to get x = 5.33 and y = 6.67. The Heisenberg Uncertainty Principle is a well-known theory of quantum mechanics which states that it is not possible to precisely measure a pair of inequalities, such as the position and momentum of a particle. Because the pair of inequalities are dependent upon eachother, they cannot be simultaneously derived. The more precise that one property is measured, the less precise the other becomes. That's nice but we're talking about equations, not inequalities, so this applies instead: http://en.wikipedia.org/wiki/Law_of_large_numbers | ||
|
red_b
United States1267 Posts
in any thread discussing balance on team liquid, there will be at least one person who insists that anyone who wants to discuss balance is only doing so in order that they may blame every single loss they have ever earned in starcraft on balance. | ||
|
sofawall
29 Posts
On January 21 2011 10:33 MicroJFox wrote: Here's a crazy idea: Balance based on racial win rates of random players only. This completely factors out player skill (the same player is playing all races), leaving only racial balance. People who play random aren't necessarily equally skilled at all races. | ||
|
tsuxiit
1305 Posts
Follow me to the learning rainbow, quick! OK so 10 players enter Blizzard SC2 Ladder. Every player starts with a "skill" rating of 50 out of 100. The first five, well, let's call them "placement matches," award one "skill" point each. From that point on players are awarded skill points in proportion to the skill points of their opponents. WHAT HAVE YOU DONE, TSUXIIT1?!? | ||
|
SharkSpider
Canada606 Posts
2. ? 3. Make post claiming quantum mechanics is related to balance. Seriously, can someone fill the 2 in for me, here? | ||
|
IPA
United States3206 Posts
| ||
|
ChickenLips
2912 Posts
On January 21 2011 11:32 sofawall wrote: People who play random aren't necessarily equally skilled at all races. but since you have a bajillion random players their playtime around all races should equal out. so if you have 10000 random players that have all spent comparable amounts of games with each race show distinguishable winrates with one race of one specific matchup you might have reason to believe something is wrong in that matchup. | ||
|
Azzur
Australia6260 Posts
On January 21 2011 02:55 jdseemoreglass wrote: The pair of inequalities in this situation is "skill" and "game balance." Blizzard is failing to recognize that it is simply impossible to determine one mathematically without also knowing the other in advance. You cannot determine the skill of a player using winrate statistics, without FIRST knowing the degree of balance in the game being tested. Likewise, it is impossible to determine the balance of a game using winrate statistics with FIRST knowing the relative skill of the players. The two variables are mutually dependent upon eachother, making it impossible to determine one without the other. You're making a lot of assumptions by saying that it's impossible. Well, in those series of videos, they actually posted up a mathematical formula which they used to remove skill from game balance. If you were serious about it, you would've looked at that equation and tried to find flaws with it rather than just speculate. The Uncertainty Principle is not a useful analogy at all. What Blizzard did acknowledge is that while their stats show a reasonable level of balance, there were a few issues: - The win % of the various races are different across leagues and regions. For example, protoss is very highly represented in ladder. - There appears to be some "time imbalance". There was the famous quote where the TvP matchup needs to be won by terran in the first 12 minutes. Thus, while the % percentages were roughly equal (making it appear balanced), but they felt that having a race stronger at a particular time is not ideal. | ||
|
Captain Peabody
United States3125 Posts
These words do not mean what you think they mean. Skill and balance are interrelated, meaning the more one is fine-tuned, the better the other can be. Now, the thing is, WITHOUT recourse to mathematics, relative skill can be determined within a certain margin of error. Now that that margin is established, we can now find balance to within a fairly reasonable margin. And now that we have done that, we can finetune our relative skill rankings to within an even better margin. And so on. Because the thing is, exact balance is impossible; and skill is highly malleable, and changes rapidly. Thus, given this, and given that the two are based on one another, it's true that EXACT CALCULATIONS can never be 100% accurate. Thus, if there were really nothing at all between 'perfectly balanced' and '100% imbalanced,' then you would be absolutely right. But the thing is, in the middle between these two extremes is where all reality lies. Because, balance CAN be known to a certain (very rough) level of precision simply by common sense and actual playing of the game; and so can player skill. However, once these two rough estimates are created, both can be used together with math to more and more carefully finetune these margins over time. Thus balance is born. You're making a very simplistic point without considering the actual meanings of the things you're talking about. It simply doesn't work that way. | ||
|
Tektos
Australia1321 Posts
On January 21 2011 11:49 SharkSpider wrote: 1. Fail at undertstanding quantum mechanics. 2. ? 3. Make post claiming quantum mechanics is related to balance. Seriously, can someone fill the 2 in for me, here? 2. Get hit in the head with a shovel. | ||
|
SharkSpider
Canada606 Posts
One way to go about talking balance is by making the assumption of relatively equal skills over a distributed group. If you assume independence between skill and race played in say, the top 3000 players, you can "rank" them, and then use a floating variance to get a deviation and increasing margin of error. If the deviation passes outside the error "cone" then there's a problem. Basically, take the porportion of palyers in the top X, then make a function as X goes to zero, and plot that on top of a function that's basically a cone of error emanating from very small error at a large sample to 100% error at 0. The only issue is that in SC2, there arent enough good players yet, and there may never be, given the nature of the game. Its still a good exercise to perform, though. This one will eliminate the "skill" issue by assuming that the player distribution at X should represent that at values of X as they get small. Typically this will give mixed results as you alter your study point, you're looking for a trend, though. Another method is to apply the fencer's dilemma to assume that over time, the top X (say 1000) of each race represent the top 3X (say 3000) players in the world. You then perform a similar cone analysis as above, but with reduced error, and you do it for multiple values of X. (You need a supercomputer or a cloud to perform this kind of analysis, anyways.) And then graph the "balance" results according to X. In SC2's case you'd assume they got worse over time and figure that out. Either way, statistical proof theory can be applied to this situation, but obviously it has a lot of error. | ||
|
TehForce
1072 Posts
| ||
|
philcorp
Canada32 Posts
On January 21 2011 11:49 SharkSpider wrote: 1. Fail at undertstanding quantum mechanics. 2. ? 3. Make post claiming quantum mechanics is related to balance. Seriously, can someone fill the 2 in for me, here? 2. Just computed commutator of skill and balance, got zero. OP negated. | ||
| ||
