Anomaly of Stockfish eval depending on phases of the game

Discussion of chess software programming and technical issues.

Moderators: hgm, Dann Corbit, Harvey Williamson

User avatar
cdani
Posts: 2204
Joined: Sat Jan 18, 2014 10:24 am
Location: Andorra

Re: Anomaly of Stockfish eval depending on phases of the gam

Post by cdani »

Thanks!
I will try some raw corrections in the evaluation function. I will tell any results.
Zenmastur
Posts: 919
Joined: Sat May 31, 2014 8:28 am

Re: Anomaly of Stockfish eval depending on phases of the gam

Post by Zenmastur »

Laskos wrote:
Zenmastur wrote:
I don't think move number is the best way to measure this effect. I would think that total material left on the board would be a better metric. These graphs with and with out passers might be instructive as well. In fact I would think breaking them up by general pawn formations may tell an interesting story.

Regards,

Forrest
Theoretically it's doable, but I spent 1-2 hours on this with move number only for all these datapoints on a database of 40,000 games. With computing material, it would take an order or two of magnitude longer. Maybe it's better to have a map between move number and average material computed on a much smaller database (statistically it is sound, as dispersion of material for a fixed move is not that high).
Something that would be interesting and also doable would be to compare SF6 with SF7 and current development version.

Regards,

Forrest
Only 2 defining forces have ever offered to die for you.....Jesus Christ and the American Soldier. One died for your soul, the other for your freedom.
petero2
Posts: 680
Joined: Mon Apr 19, 2010 7:07 pm
Location: Sweden
Full name: Peter Osterlund

Re: Anomaly of Stockfish eval depending on phases of the gam

Post by petero2 »

Laskos wrote:For now Texel (1.06beta). A bit more straight than Stockfish, but still not quite constants.
Interesting stuff!

I compared to earlier investigations by you:

Expected performance and eval of Komodo 8 and SF 6
Expected performance and eval of Komodo 8 and SF 6

Do you have any theory on what is causing the difference?
User avatar
Laskos
Posts: 10948
Joined: Wed Jul 26, 2006 10:21 pm
Full name: Kai Laskos

Re: Anomaly of Stockfish eval depending on phases of the gam

Post by Laskos »

petero2 wrote:
Laskos wrote:For now Texel (1.06beta). A bit more straight than Stockfish, but still not quite constants.
Interesting stuff!

I compared to earlier investigations by you:

Expected performance and eval of Komodo 8 and SF 6
Expected performance and eval of Komodo 8 and SF 6

Do you have any theory on what is causing the difference?
I guess time control wildly differing. Here I built a 40,000 games database of 1.5''+0.015'' ultra-fast games. There I had much longer games. Also, the number of games back then was much smaller, I don't even know how I got something meaningful in the first attempt with 200 games. I would go with current results, although the time control is unrealistic.