|
On May 20 2011 23:25 Warble wrote:
Blizzard probably has its own in-house statisticians who spend 8 hours a day on this problem while I do it on the bus. The difference is that Blizzard isn't sharing their findings with the community. No they don't. Did you ever read any book on game design? Balancing is a very simple process. It's a brute force repetition of trial and error. Nothing more than that. It's pure intuition. Pure guess work. There's zero science. There's zero math. Blizzard has no statisticians at all.
+ Show Spoiler +- First of all, to even start arguing, we need to decide what "balance" is. Seems silly, but actually people have very contradictory opinion about this
- Even if the game has 1/3 of each race at every level. It doesn't mean it's "balanced". Balance is much more than that.
- ZvZ has a 50% win ratio. But often times the winner is the one who got a lucky coin flip in the build order battle. And not the most skilled player. Does that fit your definition of the word "balanced"? Opinions will vary.
- SC2, like every other game in history, does NOT use objective math to "calculate" balance. There is no formula where you put in variables and find out how much damage a stalker should do
- SC2, like any other game. Is balanced through brute force. You put a random value, test it, if it seems imba, you change it out of pure intuition. There's zero science in this.
- It is mathematically impossible to achieve perfect mathematical balance with this brute force approach.
- Only solution would be remake a game from scratch. With balance in mind since start. And calculate balance (once we define what that is) before designing the game. Then make the game around this balanced model. ie.: completely remake sc2
- Of course blizzard will never do this. So we will never have perfect balance. We can only hope for "balanced enough" (like many would argue bw is)
- Realistically, considering the brute force approach. Our best bet is have as many balance-test iterations as possible. The easiest way to do this is to balance through designing maps (which is about ~50% of what balances the game), instead of patching. Just like BW is being rebalanced by kespa mappers after blizzard stopped patching
|
On May 21 2011 05:28 PJA wrote:Show nested quote +On May 20 2011 15:00 avilo wrote:
The model everyone uses for imbalance is...guess what?
Their personal bias and opinion. Notice my use of italics for emphasis. So basically your argument is, "Everyone uses a broken model for imbalance, therefor fuck you for trying to present a more logical model." That is such mind-boggling poor logic. OP's model is flawed in many ways, but your counterargument is even worse.
I was pointing out that there is no model and you cannot model it. Everyone has their own perception of what the balance of the game is based on their own bias and experiences. It's not something you can quantify with numbers or statistics.
You can't try and model something like this with numbers or statistics. Someone might think a unit is massively OP, and someone else might think it's just fine. That's their opinions. They didn't need a complicated formula to come to their conclusion, and people's opinions and conclusions about balance are ever changing as they learn more about the game.
So yah...you missed the point too...
|
Thank you for being such a baller. I'd love to see the analysis that you didn't post, please do it!
|
I agree with what most people said above.. nice post, especially it being your first.
|
I normally don't like these types of posts-but as I read on I quickly agreed with every point. This is probably one of my favourite reads on TL so far. Good job man!
|
Not sure what significance lower leagues have with regard to imbalance. And stats are utterly meaningless when it comes to balance. Blizzard doesn't balance their game around percentages but instead try to balance things that break the game.
Look at the thor change for example. Was there a lot of people using thors in TvP before? no Yet they nerf the thor because of an abusive strategy that was used very rarely.
edit: same with 4 gate and so on.
|
On May 21 2011 06:02 avilo wrote:Show nested quote +On May 21 2011 05:28 PJA wrote:On May 20 2011 15:00 avilo wrote:
The model everyone uses for imbalance is...guess what?
Their personal bias and opinion. Notice my use of italics for emphasis. So basically your argument is, "Everyone uses a broken model for imbalance, therefor fuck you for trying to present a more logical model." That is such mind-boggling poor logic. OP's model is flawed in many ways, but your counterargument is even worse. I was pointing out that there is no model and you cannot model it. Everyone has their own perception of what the balance of the game is based on their own bias and experiences. It's not something you can quantify with numbers or statistics. You can't try and model something like this with numbers or statistics. Someone might think a unit is massively OP, and someone else might think it's just fine. That's their opinions. They didn't need a complicated formula to come to their conclusion, and people's opinions and conclusions about balance are ever changing as they learn more about the game. So yah...you missed the point too...
Just because people have a massively flawed understanding of game theory doesn't mean they are correct.
EDIT:
To give an example: Take the game of Go. Black goes first, but white gets some number of points at the beginning of the game for free, this is called komi. The game is clearly imbalanced if komi is 0, since with perfect play black will always win. With 50 komi it is clearly imbalanced, since with perfect play white will always win.
Pros used to believe 5 komi made the game balanced. This was their opinion, but as pros became stronger they now believe that 7 or 8 komi is actually balanced. Again, komi is more or less set by current professional opinion, though I'm sure statistics from recent games are taken into account.
Anyway, despite any professional player's opinion, there is actually some value for komi at which the game of Go is perfectly balanced. Even if it were set to that value, though, I'm sure plenty of pros would feel that the game isn't balanced.
You might argue that Go is different because it is a finite deterministic game, but that is actually not relevant. If you'd like to read up on game theory you can do that yourself, though.
|
On May 21 2011 06:30 PJA wrote:Show nested quote +On May 21 2011 06:02 avilo wrote:On May 21 2011 05:28 PJA wrote:On May 20 2011 15:00 avilo wrote:
The model everyone uses for imbalance is...guess what?
Their personal bias and opinion. Notice my use of italics for emphasis. So basically your argument is, "Everyone uses a broken model for imbalance, therefor fuck you for trying to present a more logical model." That is such mind-boggling poor logic. OP's model is flawed in many ways, but your counterargument is even worse. I was pointing out that there is no model and you cannot model it. Everyone has their own perception of what the balance of the game is based on their own bias and experiences. It's not something you can quantify with numbers or statistics. You can't try and model something like this with numbers or statistics. Someone might think a unit is massively OP, and someone else might think it's just fine. That's their opinions. They didn't need a complicated formula to come to their conclusion, and people's opinions and conclusions about balance are ever changing as they learn more about the game. So yah...you missed the point too... Just because people have a massively flawed understanding of game theory doesn't mean they are correct. EDIT: To give an example: Take the game of Go. Black goes first, but white gets some number of points at the beginning of the game for free, this is called komi. The game is clearly imbalanced if komi is 0, since with perfect play black will always win. With 50 komi it is clearly imbalanced, since with perfect play white will always win. Pros used to believe 5 komi made the game balanced. This was their opinion, but as pros became stronger they now believe that 7 or 8 komi is actually balanced. Again, komi is more or less set by current professional opinion, though I'm sure statistics from recent games are taken into account. Anyway, despite any professional player's opinion, there is actually some value for komi at which the game of Go is perfectly balanced. Even if it were set to that value, though, I'm sure plenty of pros would feel that the game isn't balanced. You might argue that Go is different because it is a finite deterministic game, but that is actually not relevant. If you'd like to read up on game theory you can do that yourself, though.
Simple and precise. And this illustrates how modeling works. Chess, despite more possible games than atoms in the UNIVERSE, is easily modeled and cheated to simplicity by Computers. Programmers first tried brute force and failed epically. Then they "taught" computers candidate move theory. Then working with GMs they have finally taught it sufficient positional understanding that the top computers dominate most top GMs handily.
Go on the other hand cannot be simplified usefully at our understanding of the game. So computers barely beat mediocre amateurs.
Starcraft sits at the Go level, look at competitive AIs in BW and SC2, if they don't cheat, they can't possibly handle the data in a relevant manner. Humans however can ignore pieces selectively on the fly and assess "what matters," allowing us to create models computation fails to resolve.
Statistics are perfectly useful in appropriate settings, aka trying to determine the gestalt of the MU imbalances. Peer review is the method for fine tuning. This is what the OP supports, this is what is done.
Cheers.
|
this is topic abouth imbalace ye ? if it is i have few things to say... 
first there is imbalance around unites that their conters dont work...and some abylitys help to be unconterible...
simply exemples : colloss you cant conter em with imortals...(maby void ray work but not in game..)
mass marines with stim pack... : there is no unites to conter em in bettle expect colloss and tunder storm that is ability and banglings...fungel... problem is that that is only splash dps can stop it...
simply fact is that mass dps colloss banlgings stim pack MMM you cant conter without AOE unites or abilitys...i have 1000 what unite (compositions) dont conter em but i dont have time to write..
problem is in gameplay 2... i look about 5 min ago liquid tayler great player great skill everything but he played agenst Terran and he didnt build colloss and hts.. but becous of presure he builed gateway unites...
what hepened he menage to great macro micro pull out (kill) few medivac but simply without any skill and A MOVE MMM is able to kill Great player army like nothing....you dont need skill to a move and win whatewer your oponent is..
gameplay mestace is that unites and abilitys that killed skill and micro is STIM PACK banglings and colloss and FF...but if they nerf this abilitys and unites DPS maby gameplay will not be broken as it is now... sry for my ENG....thx for reading.. 
|
On May 21 2011 06:45 Sleight wrote:
Chess, despite more possible games than atoms in the UNIVERSE, is easily modeled and cheated to simplicity by Computers. Programmers first tried brute force and failed epically. Then they "taught" computers candidate move theory. Then working with GMs they have finally taught it sufficient positional understanding that the top computers dominate most top GMs handily. What are you talking about, computers still use brute force to beat GMs with varying levels of candidate moves to try to reduce time. Candidate moves for computers is a tradeoff that makes the computer play worse but is faster to calculate. So they try to balance how much brute forcing to use for slower computers.
The only reason Go is harder for computers is because there are more possible moves to analyze through brute force.[/quote]
On May 21 2011 06:45 Sleight wrote:
Starcraft sits at the Go level, look at competitive AIs in BW and SC2, if they don't cheat, they can't possibly handle the data in a relevant manner. Humans however can ignore pieces selectively on the fly and assess "what matters," allowing us to create models computation fails to resolve. One thing has nothing to do with the other, existing SC AI only lose to humans because they suck and no one with the right skills ever tried to seriously tackle that problem. Making a SC AI that beats humans is technologically trivial. The only hard problem would be reverse engineering the game engine to realize what effect each move will have before executing it (in chess you already know that if you place your peon in range of another peon, you lose your peon, in SC that depends on the inside of the game engine). Once that's done, there's very very few possible moves to analyze. It would be a very easy game to brute force.
|
On May 21 2011 07:05 VIB wrote:
One thing has nothing to do with the other, existing SC AI only lose to humans because they suck and no one with the right skills ever tried to seriously tackle that problem. Making a SC AI that beats humans is technologically trivial. The only hard problem would be reverse engineering the game engine to realize what effect each move will have before executing it (in chess you already know that if you place your peon in range of another peon, you lose your peon, in SC that depends on the inside of the game engine). Once that's done, there's very very few possible moves to analyze. It would be a very easy game to brute force.
You don't need access to the game engine to do this. You only need to know the game rules, which are easy enough to determine.
I like that OP tries to find a balance discussion for the ladder at steady-state, rather than just talking about the game. There's an inherent problem in this though, because it assumes everyone on the ladder has reached their level of incompetency and is no longer improving. In actuality, players are constantly improving and doing so at possibly different rates.
If Player A and Player B play 100 games with a 50% win ratio, only in a theoretical world could you say with any certainty that they are evenly skilled. The order that they won games is important. If Player A wins games 1-90 with one strategy and player B discovers a new strategy and wins games 91-100... who is more skilled? The player with 90% win rate or the player who will now win every game until the other improves?
Why does this effect your statistical summary? Skill gain does not occur evenly for each player, and cannot be predicted by any conceivable ladder-based metric. There is no state that would allow an analysis of the type you're proposing. It's an interesting idea, but I don't think it's ultimately useful for balance discussions.
|
On May 21 2011 07:32 scorch- wrote:
You don't need access to the game engine to do this. You only need to know the game rules, which are easy enough to determine. Yes you do, no it's not easy to determine. Bugs and other peculiarities in the engine are not technically possible to determine from an outsider point of view. Reverse engineering is required.
|
Show nested quote +On May 21 2011 07:05 VIB wrote:On May 21 2011 06:45 Sleight wrote:
Chess, despite more possible games than atoms in the UNIVERSE, is easily modeled and cheated to simplicity by Computers. Programmers first tried brute force and failed epically. Then they "taught" computers candidate move theory. Then working with GMs they have finally taught it sufficient positional understanding that the top computers dominate most top GMs handily. What are you talking about, computers still use brute force to beat GMs with varying levels of candidate moves to try to reduce time. Candidate moves for computers is a tradeoff that makes the computer play worse but is faster to calculate. So they try to balance how much brute forcing to use for slower computers. The only reason Go is harder for computers is because there are more possible moves to analyze through brute force. Show nested quote +On May 21 2011 06:45 Sleight wrote:
Starcraft sits at the Go level, look at competitive AIs in BW and SC2, if they don't cheat, they can't possibly handle the data in a relevant manner. Humans however can ignore pieces selectively on the fly and assess "what matters," allowing us to create models computation fails to resolve. One thing has nothing to do with the other, existing SC AI only lose to humans because they suck and no one with the right skills ever tried to seriously tackle that problem. Making a SC AI that beats humans is technologically trivial. The only hard problem would be reverse engineering the game engine to realize what effect each move will have before executing it (in chess you already know that if you place your peon in range of another peon, you lose your peon, in SC that depends on the inside of the game engine). Once that's done, there's very very few possible moves to analyze. It would be a very easy game to brute force.
I don't understand how there are very very few possible moves to analyze... Even in a simple engagement of, say, 1 zealot 2 stalkers vs 1 zealot 2 stalkers there are so many different variations. If you include possible locations to have your scouting probe for each player and all the possible paths you could be taking it gets even worse.
EDIT: In case I was not clear, the location of the scouting probes is important because they sometimes come into play in these situations.
|
On May 21 2011 07:36 VIB wrote:Show nested quote +On May 21 2011 07:32 scorch- wrote:
You don't need access to the game engine to do this. You only need to know the game rules, which are easy enough to determine. Yes you do, no it's not easy to determine. Bugs and other peculiarities in the engine are not technically possible to determine from an outsider point of view. Reverse engineering is required.
Human beings play the game. An AI plays the game. The AI performs the exact same actions a human being does when a human plays the game, only instead of actually moving a mouse, it just tells the computer it moved a mouse. Did humans "reverse engineer" starcraft to be able to play it? No they played it, learned the rules, and got better at it. Stop making shit up.
|
On May 20 2011 14:10 usa11220 wrote:
best first post ever?
Best first reply to best first post. Wait has this been done?
This is def my favorite OP since starting to visit this website I think. At least when it comes to any kind of strategy or balance discussion.
|
The biggest problem I think this model has is that it takes too broad of an approach to what "balance" is. You are modeling the statistics around trying to show if matchups are imbalanced. According to your article as far as I understand it we could have a statistical situation of balance even if each race has very abusive strategies for certain maps and match ups.
I do not think blizzard "balances" the game at this large a scale, nor is it necessarily useful to. What blizzard considers balance is outcomes being highly dependent on player skill and dynamic play. They do not want a situation where one strategy is always the best against x race, no matter what that race does even if statistically we could consider that balanced. They have been known to make changes based on replay packs sent in by pros showing certain abusive timings for instance. This is what any kind of statistical analysis should be focused on not a broad race vs. race discussion. A race is a set of all possible units and strategies that race is able to use. In any particular game specific parts of that set are used in quantities that vary with time. That is why even a mirror match up can be "imbalanced", if both races arrive at a place where they must play out the game the exact same way. The player with more skill will undoubtedly win but there is no variation in game play. Of course some strategies will be more useful than others but the goal of blizzard is to have the widest array of viable strategies well at the same time keeping player skill the predominant factor in determining outcomes and you are completely ignoring the first part.
In this sense it is much more useful for a pro player to say to blizzard, "this strategy on this map is abusive it limits my options too severely and/or it does not allow the more skilled player to win as often as they should" and then provide examples, than it is to give a statistical analysis of a ton of different strategies across a large number of maps. By ignoring the fact that blizzard wants to have varied game play in starcraft you are not correctly defining what balance is.
TLDR; Race vs. Race balance is not important, strategy vs. strategy balance is; and blizzard has two goals with balance to favor the player with more skill AND to have varied and dynamic gameplay which this analysis does not even consider
|
TLDR; Race vs. Race balance is not important, strategy vs. strategy balance is; and blizzard has two goals with balance to favor the player with more skill AND to have varied and dynamic gameplay which this analysis does not even consider
Man problem is that gameplay dising is broken... idra said the truth whan he said that low skill ppl can win agenst high skilled ppl...
i whatch liquid tayler in game where he faced MMM (stimed) and he only pull out gateway unites and he losed EASY...
so no mether how much skill you have its not inportant THIS GAME BECOMED A GAME OF UPGREADS NOT SKILL...END...
no mether how much micro you have stim pack colloss banglings Force fealds kill your skill...Game LIMITE skill of great PRO players like IDRA and other ppl...that is BIG PROBLEM...
The my conclusion that unites That does ULtra mass DPS they are killing gameplay and skill..
|
I appreciate the work you put into this, but really don't agree with alot of it
|
On May 20 2011 15:00 avilo wrote:Just because someone made an incredibly long post does not make it mega awesome or even remotely accurate. The most obvious thing completely wrong with the "post" is to not look at pros for balance. In any RTS game or game you always look at the top level for balance because these are the people playing the game at the highest level and are actively trying to "break the game." Really all the OP is saying (but ironically not doing in many of his examples in the OP) is: "don't be biased with your balance judgements." Nothing new...and there's no need to go into "intricate mathematics" or math at all for any of this...the OP is overcomplicating things, and likely has not enough experience to legitimately comment on balance or imbalance in the first place. The most qualified people to talk about balance are the pro players and people high up on ladder that are playing the game everyday versus other good players. But 99% of these players are trying to practice and improve themselves and not even worry about balance in the first place, though everyone QQ sometimes. imo OP is just trying to re-invent the wheel on balance discussions aka having a discussion about how to discuss things lol...there's about one of these posts per month or so that pop up with some guy that thinks he's mega smart and mystical with "the maths" -_- there's just so many things wrong in the OP and ironically "biased." Do we really need another thread discussing how we should be discussing things and hordes of low post count people going, "wow you're so smart and amazing."  Nice effort sure...but i think a bit misplaced. Also, the entire premise of the thread doesn't work because there is no definitive model for imbalance. The model everyone uses for imbalance is...guess what? Their personal bias and opinion. Notice my use of italics for emphasis.
Funny coming from the biggest balance whiner around... only about their own race, of course.
|
On May 21 2011 05:56 VIB wrote:
No they don't. Did you ever read any book on game design? Balancing is a very simple process. It's a brute force repetition of trial and error. Nothing more than that. It's pure intuition. Pure guess work. There's zero science. There's zero math. Blizzard has no statisticians at all.
To be fair, Blizzard did hire some statisticians to develop the new ladder system, though I don't know if they are still around.
I agree with your sentiment though. I don't think posts like these are that helpful. Balance is about perception and gut instinct, not about statistics. It's about game design -- losses should feel "fair". From an entertainment perspective, the breakdown of win percentages is less important than whether or not losses feel like the game has conspired against you. As someone else mentioned, mirror matches can feel incredibly unfair at times, which hints at balance problems even though the win percentage itself is exactly balanced.
tldr: Balance is a subjective opinion and is reached by consensus, not by statistics.
|
|
|
|
|
|
EPT
