Page 1 of 1

Adaptive Elo rating without centralized server

Posted: Mon Dec 21, 2020 5:50 pm
by x3oo
Dear OpenRa-Community
I hereby suggest to implement a decentralized Elo-rating algorithm for openra without the need to collect statistics on a server.

https://ryanmadden.net/posts/Adapting-Elo

The algorithm is implemented as described in the linked blog article. Then the calculated elo ratings for all played matches including players are saved LOCALLY instead on a server. So everyone is doing the Elorating only locally for themselves and when you join a server you can transmit local elo ratings and recalculate them. basically this is how its done in chess and it works decentralized too.

Re: Adaptive Elo rating without centralized server

Posted: Sat Jan 02, 2021 8:09 pm
by Matt
Something in that vain was done with http://oraladder.net/ see http://blog.pkh.me/p/28-building-a-comp ... penra.html for a making of. It uses TrueSkill which unlike Elo supports games with more than 2 players.