|
well after a day I was finally able to read your OP  , most of it at least (i'm not into statistics all that much, at least the nitty-gritty of it). I thought it was some pretty awesome conclusions, just b/c I love learning new things, but that's just me. Anyways, I'm really hoping that you don't get dissuaded by all the random nit-pickers who have some issues with your OP, I'm really looking forward to another fantastic statystical analysis as you do two things:
1) keep in the statistical parts and laymen parts in spoilers so it is optional read depending on how bold you are
2) you keep both the statistical parts and laymen parts in the OP, where most other OPs would have just kept either only the statistical part of the laymen part but not both.
Must have been a lot of effort to code everything and go back and catch most edits and everything, all-in-all a fantastic thread, would love to rate it a 5/5 if it were blog. Keep up the good work 
|
Good post but the word balance is just to hard to start a discussion just because it is mostly about each players opinion. Now if we think about balance in a game where there are 3 races which are zerg, terran, and protoss and each race with specific units and different types of maps this just takes the word balance to a new level. But it was an interesting post and fun to read
|
Incredible post. Good work indeed.
|
So if a matchup is imbalanced, a disproportionate amount of race X players will be promoted to higher leagues, rather than maintaining equal league-population/total-population ratios to the other races.
^ OP in one sentence. I have to agree with the person who said "an incredibly long post does not a great OP make". (or whatever it was). It's a cute idea, but it's pretty short sighted. You assume so many things, the first of which being that a single point of imbalance affects all leagues equally, which is just plain not true.
And really, your entire mythbusting section just makes you sound like you haven't thought any of this through.
And even still, what's the point? Your entire analysis is "If this matchup is imbalanced, the distributions will move this way". Truth is, imbalance is not the only thing that can move those population distributions. The easiest example is when a new popular strategy comes out and nobody has any clue how to deal with it. No patch was released. It's just people playing the game in different ways, and yet your analysis would conclude there's imbalance now when there wasn't before. And then in a few months, people figure out how to scout and respond, and the "imbalance" is gone.
The only thing this analysis can actually provide is a test to see if there -isn't- imbalance, and it can barely do that. If it sees imbalance, it could be purely an effect of the matchup not being understood on one side. If it doesn't see imbalance, that would be just because we haven't found the imbalance yet.
There's so many words, but a complete lack of a point. Even if this sort of analysis showed conclusively that an imbalance existed, then what? You would still need to look at individual games to see what it is about strategy X that is imbalanced. You can't just say, "Hey, PvZ is imbalanced. Let's give Archons +5 more damage vs. bio and call it a day".
It just sounds like a nice idea, but it has absolutely nowhere to go.
|
It seems like some people are missing that this post is not about the best way to balance the game, but rather is about the best way to statistically analyze the balance of the game. He is saying that pros are statistically unhelpful. If you're balancing based in reasons besides statistics, obviously pro's opinions become much more important.
|
On May 21 2011 18:46 Authweight wrote:
It seems like some people are missing that this post is not about the best way to balance the game, but rather is about the best way to statistically analyze the balance of the game. He is saying that pros are statistically unhelpful. If you're balancing based in reasons besides statistics, obviously pro's opinions become much more important.
1. He did not, in fact, say anything, on how one should, practically, analyze the balance statistically. He is not, specifically,suggesting what one should be doing with the following methods: analyzing win ratios over a career, win ratios over some fixed time, ladder points, number of race x in league y. He only presents 4 cases with arbitrary proportions of players.
2. He says that "pros are statistically unhelpful" but it is wrong, you can statistically balance the game using only pros, if and only if (usually called iff) you: a) want to balance the game primarily for the pros and/or b) the balance changes gives equal results over all leagues.
If the win ratios from 99 GM-players (where 33 of each race are randomly selected from 3 lists of GM-players according to race) if these win ratios differ by much then it might very well be a result of imbalance.
That we need to include all players down to 50% win ratio to perform a statistical analysis is a pretty big fallacy.
Is every opinion poll in the US conducted by interviewing half the population? No, but a larger sample gives a smaller confidence interval.
I am sure that there are better and more experienced statisticians than myself, I would welcome them to take the debate if they feel that it's worth it.
This article does however contains errors and shortcomings, and so many people seem to get it, objectively, wrong, that in the mean time I will try to point out the faults lest people will fall into the wrong mindset about statistics and game balance.
|
On May 21 2011 15:06 Samhax wrote:Show nested quote +On May 21 2011 14:15 ploy wrote:
roooofl, don't want to derail or anything, but there is no good AI poker bot except for perhaps heads up games. Other than that... not even close.
Also, you are oversimplifying chess/go. If you could map out an entire game of go/chess so easily, then the games would have already been "solved" by computers.... which computers are not close to doing at all as of right now.
Edit: On second thought, there are definitely no good AI poker bots. The only bots that can play heads up well are based purely on game theory, which means whatever the other player does means nothing to the bot. No AI involved, just game theory (simplified forms of heads up poker have been solved by game theory). I think computers in chess are pretty damn good, they can beat or do draws against the top players, But in Go it's impossible for a computer to match a top player. Go is a lot more complicated than chess.
I read a pretty good article about how it has nothing to do with how complicated Go is. The issue between Go and Chess is that Go pieces don't move. The article was guessing and not sure, but it essentially argued that because Chess pieces move, it is easier for the computer to change the positional structure of the board in such a way as to benefit itself. With Go, the position changes only in so far as you add other pieces to change the position.
Part of it though is of course the large size of the board. A 5x5 Go board has in fact been solved. A 7x7 and 9x9 have not though, so of course a 19x19 board hasn't been solved either.
The analogy doesn't work with Starcraft very much though since units are not equal between races. Although they may serve the same function, zerglings, marines and zealots are different enough between each other that they are not strictly comparable the way a black piece and a white piece can be compared in Go or Chess (they are the same piece).
Anyways, I liked the OP's post. The OP may have made certain mistakes in logic, I cannot say since my math background is very poor compared to some other people in the thread. I do have to say that I agree with the notion that listening to pros discuss balance is usually pointless. Someone compared this with peer reviewed science, but I think it should be easily pointed out that peer reviewed science articles are reviewed by multiple scientists, and then ideally other scientists will try to recreate the experiments to see if there are problems in the initial experiment, or unaccounted for variables.
Starcraft pros don't do this. A few pros sitting around a table or in an interview discussing balance hardly makes for a good discussion. Looking at Idra and Day9 on SotG made this super obvious. Idra gave a number of problems he sees with Zerg, and Day9 asked to look at specific games to analyze. For all I know, Idra may have been 100% correct, 100% wrong, or anywhere in between. But seeing as they did not sit down to discuss it for any long period of time looking at a bunch of games, it really doesn't matter. And even if they did, they would need to have more than just Idra and Day9 looking at it since two people is hardly enough to smooth out any biases.
I also want to point the obvious point against Random players (if it hasn't been already, I tried reading all of the posts but I might have missed it). Random players can be biased against races too. Just as an example, TLO's least favorite race to play was Protoss, and I believe his least favorite matchup was PvP (I don't remember). TLO might be slightly less biased than single race players, but there is no reason to assume that Random players are automatically unbiased. And anyways TLO switched back to Terran, but the analogy works with any Random player.
Anyways, even if OP was wrong in many places, I don't understand why people are coming in here and shitting on him. Pointing out his mistakes is fine sure, but why is this post any more useless than any other post? Even if every point OP made was wrong, at least he is trying to think critically about it. Point out the mistakes, but leave personal insults out of it.
|
|
|
|
|
|
EPT
