K-Factor and the SSDF

[Sean Evans]

If FIDE changes the K-Factor, I am wondering what advantages and disadvantages this would give chess programs at the SSDF, who will presumably follow along with FIDE!

Polish GM Bartlomiej Macieja is insisting that the decision to increase the K-factor in rating calculations is not just necessary and good in the current tournament situation, it is in fact irrevocable and can only be legally changed by the body that passed it.

http://www.fide.com/component/content/a ... e-k-factor

[Vinvin]

SSDF (as other computer rating lists) doesn't use K-factor because computer rating lists use performance to calculate ratings. K-factor is usefull when the strenght of a player changes (when human improve or regress).

If FIDE changes the K-Factor, I am wondering what advantages and disadvantages this would give chess programs at the SSDF, who will presumably follow along with FIDE!

Polish GM Bartlomiej Macieja is insisting that the decision to increase the K-factor in rating calculations is not just necessary and good in the current tournament situation, it is in fact irrevocable and can only be legally changed by the body that passed it.

http://www.fide.com/component/content/a ... e-k-factor

[Uri Blass]

SSDF (as other computer rating lists) doesn't use K-factor because computer rating lists use performance to calculate ratings. K-factor is usefull when the strenght of a player changes (when human improve or regress).

I am not sure if you are right for the ssdf.

computers use book learning in the ssdf games so the strength of the players can be changed in the ssdf games.

Uri

[Laskos]

SSDF (as other computer rating lists) doesn't use K-factor because computer rating lists use performance to calculate ratings. K-factor is usefull when the strenght of a player changes (when human improve or regress).

I am not sure if you are right for the ssdf.

computers use book learning in the ssdf games so the strength of the players can be changed in the ssdf games.

Uri

No, I think SSDF uses something like Elostat, not K-factor. Asymptotically they are the same, if players do not evolve or devolve rapidly. The K-factor diminishes the variability of the rating, if I remember well, as it overlaps several dozens of games game by game. I am curious about another thing, about inflationary/deflationary ratings of FIDE and SSDF. Here they may differ greatly, as it depends on new entrants and retirements. In FIDE the new entrants are usually weaker young players, which are advancing, and the older, maybe higher rated players are retiring. In SSDF most new entrants are strong from the beginning, and leftovers are weaker engines on a weak hardware. Therefore inflationary/deflationary nature of ratings may differ strongly.

Kai

[MattieShoes]

Given all of the historical games, one can tweak rating systems retroactively and see what happens. Sonas did this with ChessMetrics. There's also an interesting Chessbase article from him about the shortcomings of FIDE rating system.

K factor doesn't affect SSDF I don't think. Changes to the FIDE rating system might affect how the SSDF ratings compare to FIDE ratings, but given the issues with rating systems over time, this is a problem regardless.

[Laskos]

Given all of the historical games, one can tweak rating systems retroactively and see what happens. Sonas did this with ChessMetrics. There's also an interesting Chessbase article from him about the shortcomings of FIDE rating system.

K factor doesn't affect SSDF I don't think. Changes to the FIDE rating system might affect how the SSDF ratings compare to FIDE ratings, but given the issues with rating systems over time, this is a problem regardless.

Exactly, the new K-factors are not so relevant when comparing FIDE ratings to SSDF, because they anyway use different rating schemes, and are hard to compare for any K-factor (AFAIK SSDF uses something like Elostat with no K-factors). More of a concern to me are general inflationary/deflationary trends. I am aware of the Sonas work, he even proposed a new rating scheme, if I remember well. Would be interesting to repeat his work for the computer ratings (Elostat and Bayeselo).

Kai