EPT
losing to 
really means is weak to 
.Warning: this post suffers severely from tl;dr. Read the abstract to see if this interests you. If not, just move to the next thread.
Abstract: Balance is one of the favorite topics on this forum despite the fact that there is little critical thought about what it actually means and how it can be measured. This post will lay out why it is impossible to achieve a complete objective measure of balance. It will describe exactly what a player would see on average in terms of wins and losses as a result of imbalance (I guarantee this will surprise you). It will also discuss what I think might be the only "good" measure of balance, which has not been discussed openly by Blizzard as far as I have seen. Even then, it won't work that well. Note that I don't consider the role of maps here, which is enormously influential. Changing the map pool can completely rebalance the game of course, so assume every statement is made with "using the standard ladder rules and map pool" after it.
This whole post starts with a whine. Any whine will do. Oh, here's one now.
"I lose to
all the time. Blizzard loves and hates . I quit forever."Lovely. So here's my question: If Blizzard was actually insane and loved
and hated and really had no feelings whatsoever about money, what would happen to ? And for that matter, to ? Try to predict the outcome. I bet you can't.Before we dive in, some quick background. I will identify two types of imbalance, "racial" imbalance and "matchup" imbalance. Racial imbalance is an imbalance that affects *both* of a race's non-mirror matchups, i.e. making SCVs mine faster would allow
to beat both and more easily. Matchup imbalance is an imbalance that primarily affects *one* of the matchups, while leaving the other alone. For instance, making EMP's area of affect larger would allow to beat more easily, but it wouldn't really do anything to the v matchup. Obviously a lot of changes would fall in between, but we'll get to that.The whine that I described above specifically complains of a matchup imbalance, that Z loses to T, but doesn't complain about any of the other matchups. So we will assume that P versus either of the two races is balanced. And of course the mirror matches are balanced. If that was actually the case, what would happen?? Well, here's your answer: If T actually beat Z a little more than 50% of the time, Z players would have, on average, a little lower MMR than they should. That means they would, on average, be matched up against slightly weaker P players. This means Z would beat P slightly more than 50% of the time. So from the perspective of Z, T>Z>P. Easy peasy. Let's all complain to Blizzard... But not so fast. Because decent Z players are losing to slightly less skilled T players, those Z players have a lower MMR than they ought to. That means your average P player is facing more skilled Z players than he should. That, in turn, lowers his MMR slightly leading him to face slightly weaker T players. So from P's perspective, Z>P>T. To round it all off, we know that T us supposed to beat Z, but we just figured out that due to the magic of the ladder and MMR ratings, T now loses to P more than 50% of the time on average. So from T's perspective, P>T>Z. What a disaster! A single matchup imbalance causes everyone to lose, on average, by some non-50% amounts to every race except their own (in a mirror match). So the source of the imbalance is completely disguised. You cannot figure out by any means which race is the culprit. T being overpowered against Z causes T players to *lose* more games against P. Note that they deserve to lose these games.. they are inadvertently facing more skilled P players than they should be due to their extra "unfair" wins against Z. Did you predict that? Liar!
What happens in the case of a racial imbalance, where, say, Z is equally weaker than *both* T and P, which are balanced with respect to each other? This case is a little simpler, though no less infuriating. A Z player will win 50% of his/her games against each race and look for all the world to be perfectly balanced. Why you say? Well because the imbalance will lead him to face weaker T and P players on average. He will be lower in the ladder than he should be, but there's no way for him to know (except by the brilliant intuition that he should be higher in the ladder than he is because he is so pro). As long is the imbalance isn't so severe as to cripple the top of the ladder, it is invisible. Top players playing Z will be forced to be more innovative in the metagame to make up for this "unfair" advantage, but if the imbalance is relatively minor or confined only to certain situations: say, Z has a hard time responding when T and P do low-econ openings in close position maps that force Z to make extra units and delay their third, much of the time Z will overcome this disadvantage by either taking the risk and getting away with it or making up for the late third later. However, this balance issue (hypothetically, I'm not actually saying Z is underbalanced here, I just need an example) allows P and T players of lesser skill to defeat Z players a slightly higher percentage of the time. This imbalance would be entirely invisible statistically.
Now imagine a world in which there are *multiple* minor imbalances of *both* matchup and racial types, i.e. reality. Sorting out which race needs tweaking is hopeless. Nothing yet described publicly by Blizzard even approaches being able to do that. As I said at the beginning of the post, there are two ways I think it could be done. But first, one common idea that *wouldn't* work.
The logical impulse is to use random players. Sure, each random player may be better at one race than another, but on average, random players should give an idea of which race is best. In reality this doesn't work, and that's because being random changes the game. Now your opponent doesn't know what race you are, and that lack of information may confer a greater advantage in certain matchups than in others. In other words, this would only work if random players' races were announced at the start of the match. However, even if you were to fix that, it changes the game in another way. Random players at high levels of play are notorious for being bad at the late game in their weak matchups because they have less time to study the late game of each matchup in detail. That means that, on average, how random players do (or even dedicated multi-racers) will reflect more on race balance for players who are bad at the late game (and note also that their opponents may be gearing their builds to force a late game for this specific reason, further convoluting the situation). So random players and multi-racers can't be used to balance the game.
One thing you could is look at the top off-ladder competitive or professional play. This is probably the most sound method for achieving balance, though there are problems here as well. The first, which I mentioned above, is that a minor imbalance may just force more creativity from players of the "weaker" race. This will be especially true soon after the release of the game, i.e. now. This factor becomes less important as the game gets older and the strategies are tested further. The other major problem is the sample size. If you only look at GSL players your sample is too small to be significant in diagnosing all but glaring imbalances. If you include all professional and top amateur competitions you might have a better chance of getting a significant read on which matchups are imbalanced, but remember that the metagame changes relatively rapidly in the first years after release, so the data won't really mean anything at first.
So this brings me to the conclusion. I'll put it in bold. There is no way your personal experience can determine which matchups are balanced. You can't even do it with masses of data. Measuring balance is a huge problem for Blizzard, and my assumption is that they probably have a bunch of numerical systems, but that in the end it's a refined "squeaky wheel" model that trumps. (of course, only certain wheels get any attention, and they are probably deaf to whining)
A little about me. I am a diamond
, though not high enough diamond that it means anything. I am a professional measurer of things (i.e. scientist) so I have some idea what people face when they try to measure things that are hard to measure. There are assumptions in what I say here that are all based on what Blizzard has actually communicated to the community. It may be that the system is actually different, but whatever system they use, it will still suffer from this problem. I hope this post serves to help people realize what a complicated system this 3-race game is and what steps are probably necessary to actually bring it into some semblance of balance. Really the best work that can be done in achieving balance is innovating within each matchup. The faster the metagame is mapped out, the faster the statistics at the top levels of play will actually matter.Thanks for reading. Go Jinro.
