According to statistician Jeff Sonas, lower rated players win more often than they're supposed to.
  • kephalos@discuss.tchncs.de
    1·
    1 year ago

    Not sure I understand everything, although I think I understand that an influx of new players can totally skew statistics.

    It’s worth noting that this is about ELO ratings, so lichess may have a similar problem (I think it might be the case for Blitz on Lichess to suffer from a similar problem), but some of the proposed fixes are FIDE ELO specific (increasing starting ELO).

    The problem seems to be that the ELO ratings aren’t accurate to estimate the correct probabilities for a match between a long-time chess player with higher ELO rating and a player from the “Queen’s gambit wave”.

    Now the authors seem to paint this as a problem with the new players being underrated, the ELO distribution to be skewed. I agree that this can be a skew, I wonder however if the solution should be to boost ELO ratings of lower-ranked players.

    • Overall the best fix would IMHO be to bring together higher-ELO with lower-ELO players in matches in order to allow the ELO distribution to move ELO points down from the upper end, so that the ELO numbers again match the winning-probabilities between two players. I guess there is hesitance to do that because it means the old-guys might lose rating points and people are naturally protective in this regard.
    • bumping ELO points would lead to an inflation in ELO rating overall, it does not fix the root cause.