We know that it's not the most ideal way of presenting the data, but it's not wrong at all, people just interpret it wrong.
We'll look into normalizing it, but that requires a lot more work.
What if you try subtracting the overall winrates to the hero winrates? That way you're left with numbers like (example):
Broodmother +5.5%
Meepo +1.5%
Void -3%
Silencer -7%
Compared to average
That's a lousy way of doing statistics, and that's actually wrong data. We prefer showing only correct numbers.
Bumping this because I think it's important. Currently, when you click on a hero's stat page, over half of the page area is taken up by "Best Versus" and "Worst Versus"....but these statistics are useless because they almost perfectly mirror the over all win% page. It is very easy to normalize these statistics, and would greatly improve the usefulness of this site. Why settle with over half of a page area being incorrect?
Lütfen yorum yapmak için giriş yapınız
If you go to any hero's matchup page (https://dotabuff.com/heroes/centaur-warrunner/matchups), it is too similar to the overall winrates. It tells us that pretty much all heroes are bad against Centuar, Drow, Necrolyte, and good against morph, meepo, AM, etc. This is useless data! We need to normalize it against the average. So we can see who actually fares best against Cent, COMPARED TO EVERYONE ELSE!
This leads to people like this guy writing an entire statistically incorrect essay: http://www.reddit.com/r/DotA2/comments/13m8ac/to_calm_a_stampede/