So as far as I understand the ranking in OGS is divised with the size of the board and and the type of time the game have.
If I look at my ranking , it make no sense to me. My Global ranking on normal real time board
is 14.0 Kyu. But all my ranking on individual board are much lower ( higher in number) then 14.0 Kyu ( 16.0K for 9x9 , 15 K for 13 x 13 and 15.1K for 19 x19 )
So I dont understand where my global ranking come from , cause it can not be a sort of average of my individual board.
Indeed it’s not an average. From what I remember each time you play a ranked game your overall rank is adjusted based on your opponents overall rank.
Your individual rank, the ones that match the game settings, 19x19, blitz etc are also adjusted based on the opponents overall rank (not their individual one).
overall rank updates each time you play raked game
9x9 rank updates only when you play 9x9 ranked game, so it answers the question “what rank I would be if I played only 9x9 on this account?”
Glicko is perfectly capable of doing that, but anoek’s investigations found that the overall rank was a better predictor for game outcome than the individual categories. Turns out more data for more useful than tailored data.
oh, that’s what I meant by “better than just the overall”,
that the overall is better system than individuals as is, but if we find one that incorporates them better, we’d probably switch to that
ofc I strikethroughed, cuz I doubt this community will easily find a “statistically sound” way of doing it that properly incorporates training/test splits in testing