szr: Thank you for your love and support of QLRanks :) We've had such a great time bringing this project to the community and in celebration of hitting 1 Million Duels tracked we're really excited to present you with... A new design, bug fixes and many new features!!! 
sirax
0x5f3759df
Fattywhale
The fairest way to calculate performance is based upon victory and the balance of teams.This doesn't explain why he gets 200 ELO points for that single match.
Memento_Mori
annihiluzz
Because the algo is adjusting rapidly for a player who has never lost.
"Some random data" can't invalidate the hypothesis but it can invalidate reasoning behind it in this case. (as far as you don't take unrealistic assumptions like private matches being such huge factor in this data.)
When I wrote about diversity I meant chance that diversity is more or equal than this.I know what a p value is. Also, I didn't claim that you didn't.
Not probability of exactly these numbers - it would be silly.You did test whether or not the results (or more extreme ones) are likely to emerge when the chances are exactly 50-50.
Ok, I admit that my approach wasn't optimal in this case.Huh? I didn't criticize your approach. I used the same one and I don't know a much better one for the given data. There's really not much to test, since we know hardly anything about the data.
We do not think that this could simply be explained by a preference of better players for the red teams because (a) we only included results from contests involving the top playersand this just doesn't make any sense to me. ;>
Drexciyian
I don't think there's a single top player out of the top 200.
dashimself
Q: Does the TrueSkill ranking system reward individual players in a team game?
A: The only information the TrueSkill ranking system will process is:
Which team won?
Who were the members of the participating teams?
You forgot one thing: it has knowledge about people's current rating.Yes, rating, not skill. And that rating can be completely wrong indication of skill because of exactly what I described, so the system is working based on incorrect information.
If a good player has bad team, it knows it and won't punish him too much (like -10 Elo)The whole problem is that system doesn't know it, because his ELO is low because of constant losing with bad teams. He is treated the same as any other noob player, because for the system he has the same low ELO rating as everyone else in his team.
So as you see, even though player's skill will fluctuate endlessly just because of the teams he's on, he'll start "orbiting" around some number after enough games.Stability of rating doesn't mean correctness.
No, because a good player put against a bunch of noobs will win games, easy.That is, assuming, he is actually put against a bunch of noobs sometimes, which he might not, because QL has shuffle.
If you think a good player will never win a single game, you're just wrong.Of course he would win some games, but if he selects (or is selected by shuffle or any other external factors) to join teams that are unlikely to win, then his win/loss ratio will be biased and his computed rating will be lower than his real rating.
It's not about how many times you lose and win. It's about who you lose and win against.In qlranks, will you get positive or negative rating change depends on whether you win or lose, period. In team games, a single player does not determine whether team wins or loses. If he plays with noob teammates, and his contribution is enormous, but just a little bit not enough to cross that "win/loss" threshold, then his rating will decline, even though player with less contribution, but with higher skilled teammates would get rating increased vs same opposite team.
But you can compare players and practice shows it works good after you gather enough data.It works only for some (most) players. But qlranks does not gather all the data required to adequately rate players in team games, because it uses just plain ELO system, which is only designed for duels.
my rating before the game: 1696
damiah's rating before the game: 1715
but what in the fuck is the reasoning behind someone who performs worse (which isnt even supposed to matter) gains more points than someone who obviously did much better in that game?
![shaftwhores only by [EXE]dann](/files/avatars/2007/51852-shaftwhore.gif "shaftwhores only by [EXE]dann"){kind=small}
