How it works
Gouki.com uses the ELO system to rank players in our tournaments. The Ultra Street Fighter IV rankings are now current as of the most recent FMIWT#4 tournament. (MKX and Smash4 are up-to-date too, though they only have one tourney each so far).
This is how ELO works (I copied all this off the internet....)
Chance of you winning = 1 / (1 + 10^((opponent rating - your rating) / 400))
If your equal to your opponents rating, you have a 50% chance of winning, if 200 points above your opponent you have a 76% chance of winning, and 400 points above your opponent give you a 91% chance of winning.
There is a value which varies based on the system your using called k-factor. The SC2 k-factor appears to be 32, which is a common k-factor value.
When you win, you receive (k-factor * odds of losing) points. So if a 1900 player wins a 1500 player then the 1900 player will receive 2 or 3 points, because (32 * 9%) is 2.9 - I don't know how the rounding works.
When you lose, you lose (k-factor * odds of winning) points. Thus the losing player in the above scenario would lose (32 * 9%) points - which is 2 or 3 points.
Contrast this to a fair game between 1500 vs 1500, where the winner receives 16 points, and the loser loses 16 points.
Note that this is not like an average. An average is harder to change the more games are played. If you lose your first 100 games, it will be very very difficult to ever get an 80% win average. If you lose your first 100 game in ELO, and then win several games in a row your score will rebound very quickly.
Read more here: http://en.wikipedia.org/wiki/Elo_rating_system
Gouki uses the same K value of 32 and starts new competitors out at 1000 points.
So anyway, this used to be done by hand, and as such was subject to data entry errors, it's now pulled directly from the challonge brackets via the challonge web API. This does lead to duplication as it is matching players based on the name the input when signing up. (e.g. "tg FnJimmy" is different from "FnJimmy" or "tg|DTB FnJimmy") I have tried to fix these on our end by averaging the results. I'm not sure if that's right and I'm not inclined to do the math to prove it one way or the other, so what I've done now is gone back through the challonge brackets and normalized the names a bit and then re-imported everything. This has the added benefit of removing any previous hand-entered errors that may have been left over, some of the win/loss records have changed, so you may want to check your ratings again to see where you are. Please let us know if you see any duplicates in the list still.
These rankings will be used for our upcoming ranked tournament series and invitational tournament.