Surprisingly, for the defeat of the party charged rating!
So it should be?
http://ru.lichess.org/MQ1gvmUB(Удивительно, за поражение в партии начисляется рейтинг!
Так и должно быть?)
I've noticed this bug too, it's really weird. See
http://prntscr.com/9w2qcg. However, in that case, netslov2 might have played a game outside the tournament, which influenced his rating too.
Anyhow, this bug exists. And it shouldn't.
@1,
This is my rough calculation (found below). In (A) since bambam333 loses the game, his rating change is -0.23283. In (B) since saleger won his rating change is +0.16464.
The calculation in (B) agreed with Lichess because it was rounded-off to a whole number. But I am not sure how the loser in (A) got a positive rating change from Lichess calculation.
You may post this in Feedback Forum so developers will notice this as early as possible.
:: Glicko-2 rating system ::
A. Rating update of player "bambam333"
Old Rating::::::::::: 796.0
Old Rating Deviation: 100.0
Old Volatility::::::: 0.05
Opponent::::::::::::: saleger [R=1779.0, RD=81.0, V=0.05]
Result::::::::::::::: 0
New Rating::::::::::: 795.76717
New Rating Deviation: 100.31162
New Volatility::::::: 0.05000
Rating change:::::::: -0.23283
B. Rating update of player "saleger"
Old Rating::::::::::: 1779.0
Old Rating Deviation: 81.0
Old Volatility::::::: 0.05
Opponent::::::::::::: bambam333 [R=796.0, RD=100.0, V=0.05]
Result::::::::::::::: 1
New Rating::::::::::: 1779.16464
New Rating Deviation: 81.42774
New Volatility::::::: 0.05000
Rating change:::::::: 0.16464
RealKool,
I wrote (Forum->Lichess Feedback), but my post removed
800 is the minimum rating, it is "reset" to this just before a new calculation happens: bambam333's old rating 796 is replaced with 800 and then lowered with 1 for the loss. This lowering is not really doing anything though, as it is reset to 800 next calculation.
@4,
It is just fine as this is also a good place for discussions.
My calculation is based from this site,
http://www.glicko.net/glicko.htmlUsing the formula from the python program pyglicko2.
@5,
Good to know that. I read a doc here
http://www.glicko.net/glicko/glicko2.pdfbut cannot find this rating reseting.
Yeah, it's not defined in the rating system. I don't fully know all the reasons for the limit, but I don't see any improvements with unlimited; 800 feels far enough from starting rating 1500.
Here is the relevant code
github.com/ornicar/lila/tree/68ab2dce/modules/rating/src/mainThe directory java/glicko2 contains the system implementation from
github.com/goochjs/glicko2Especially this file that uses the limit
github.com/ornicar/lila/blob/68ab2dce/modules/rating/src/main/Perf.scala#L32-L33Which is called when ratings update after a game has finished
github.com/ornicar/lila/blob/68ab2dce/modules/round/src/main/PerfsUpdater.scala