David Kim's Current Balance Thoughts - Page 10

On January 10 2014 15:29 geokilla wrote:
Just watched Taeja play on ladder against barcode Protoss.. 1 rax into triple CC on Frost and still manages to defends Oracles..

On January 9 2014 20:52 PST iamcaustic wrote:
I've always found these balance stats from Blizzard to be rather dubious, for a few reasons.

First, the idea that one can filter out such a high-variable concept as skill with a single calculation (http://i.imgur.com/JOoZz.png) is bizarre, especially since Blizzard has officially stated that the calculation does not factor in how a player won their game, which is what skill is all about. It'd make more sense if Blizzard was referring to factoring out (or at least weighting) games won/lost between players with a high MMR disparity as some people have theorized, but without some explanation from Blizzard on this we're left to scratch our heads.

Second, even if we assume "factoring out skill" means weighting or eliminating MMR discrepancies, that does very little to demonstrate overall balance in the current ladder setup, as the numbers will simply gravitate to 50% regardless; there's no real way to see if better players are having their MMR weighed down by racial imbalances with this data alone.

You would need to also look at racial distribution in the leagues, coupled with this current balance data, to see what's going on. If there was truly racial balance, you would expect to see each race populating the different leagues respective of their usage base (give or take some percentage points to accommodate the fact that the best players won't be playing each race at a perfect 33% representation) on top of relatively stable win rates.

e.g. if we were to assume the player base was:

Terran - 75,000 people (25%)
Protoss - 120,000 people (40%)
Zerg - 105,000 people (35%)
Total - 300,000 people (100%)

Then we would expect to see something close to the top 2% of each race in Masters (again, give or take a bit):

Terran - 1,500 people (25%, 0.5% of total player base)
Protoss - 2,400 people (40%, 0.8% of total player base)
Zerg - 2,100 people (35%, 0.7% of total player base)
Total - 6000 (100%, 2% of total player base)

This is, unfortunately, not the data the community is seeing. Terran in particular is under-represented in the higher leagues by more than just a couple percentage points, while bloated in the lower leagues. This leaves the impression that Terran MMR is deflated relative to the other races. Add on top that these players are then still on the losing end in the balance stats for PvT, and it starts to become small wonder that the majority of the Terran player base finds PvT to be unfair.

I say all this without even dipping toes into the concept of quality of design. The idea of a race having a one-button solution to such a wide variety of strategies and tactics, for example, is absurd (though I am glad that the recent balance test map is looking at addressing Photon Overcharge from a balance standpoint, even if it doesn't address the absurdity of its design). Suffice to say, people are frustrated, and this blog provides little relief from players' day-to-day experiences with the game.

Speaking of experiences with the game, gameplay comes first, then balance. It doesn't matter how statistically balanced a game is if people are unable to enjoy it. SC2 isn't in a gutter by any means (let's be honest, it's still the best modern RTS on the market), but there are some pretty obvious gameplay issues that deserve some attention.

On January 10 2014 15:58 Bagi wrote:
How do you separate player skill and race strength to begin with?

On January 10 2014 13:45 RampancyTW wrote:
I'll repeat this as much as necessary to as many people as necessary:

The ladder matches you up to win 50% of your games, right? Great!

So it thus follows that your matchups will all approximate 50%, right? ...Not even remotely.

Most people have strong and weak matchups, so they may have a 50% overall, but a 60% TvZ, a 50% TvT, and a 40% TvP. So even if they win 50% of their games, they won't win 50% of their games in every matchup.

This actually makes it EXTREMELY easy to tell if there is an overall racial imbalance, because what you'll find is that even though there is a 50% winrate for the total populations, the matchup winrates will be more or less than 50% for the non-mirrors, as a population. Protoss appears to be (extremely slightly) overpowered vs. T at the moment, having just above 50% PvT winrates across region, skill level, and at the pro level. But it's nothing like the huge imbalance it's been claimed to be.

TL;DR: The matchmaking system's desire to put you at 50% winrate makes it extremely easy to spot racial imbalances across a population, because a Terran population-wide problem in TvP would require an increase in TvZ winrates to keep the population at 50%.

On January 10 2014 16:07 pure.Wasted wrote:


By that definition, coin flips are balanced. That is a useless definition.

Instead of arguing over semantics and terms, how about discussing things that actually matter, like the way the matchups are designed and why that can make for a predictable game that does not reward skill.