Stockfish 19.2.2017: any progress finally?

Discussion of computer chess matches and engine tournaments.

Moderators: hgm, Rebel, chrisw

Jouni
Posts: 3286
Joined: Wed Mar 08, 2006 8:15 pm

Stockfish 19.2.2017: any progress finally?

Post by Jouni »

STILL NO PROGRESS FROM SF8! I tested brainfish compile (faster than http://chess.ultimaiq.net/stockfish.html one) with 2*200 games vs. Houdini 5 and Komodo 10.
60s+0,6s games: -1 ELO to SF8.
Jouni
Jouni
Posts: 3286
Joined: Wed Mar 08, 2006 8:15 pm

Re: Stockfish 19.2.2017: any progress finally?

Post by Jouni »

I repeated same test with 26.2.2017 version and same conditions. Result: -2 ELO to SF8. I am surprised, that Pohl tests shows same gain! I didn't use SALC book.
Jouni
voyagerOne
Posts: 154
Joined: Tue May 17, 2011 8:12 pm

Re: Stockfish 19.2.2017: any progress finally?

Post by voyagerOne »

You will see progress after the next couple of elo patches...
ernest
Posts: 2041
Joined: Wed Mar 08, 2006 8:30 pm

Re: Stockfish 19.2.2017: any progress finally?

Post by ernest »

voyagerOne wrote:You will see progress after the next couple of elo patches...
At some moment you will be right !... :D
User avatar
MikeB
Posts: 4889
Joined: Thu Mar 09, 2006 6:34 am
Location: Pen Argyl, Pennsylvania

Re: Stockfish 19.2.2017: any progress finally?

Post by MikeB »

Jouni wrote:I repeated same test with 26.2.2017 version and same conditions. Result: -2 ELO to SF8. I am surprised, that Pohl tests shows same gain! I didn't use SALC book.
I do not see what you are seeing at all. My runs show a healthy 10 ELO gain. Ran it again this afternoon , just to be sure.

Code: Select all

1500 rounds and 1500 games completed...
time control: 3000+50
Date: 03/04/17 : 18:23:35

1500 game(s) loaded
Rank Name                      Rating   Δ     +    -     #     Σ    Σ%     W    L    D   W%    =%   OppR 
---------------------------------------------------------------------------------------------------------
   1 Stockfish 040317 64 POPC   3105   0.0    4    4  1500  771.5  51.4  263  220 1017  17.5  67.8  3095 
   2 Stockfish 8 64 POPCNT      3095   9.7    4    4  1500  728.5  48.6  220  263 1017  14.7  67.8  3105 
---------------------------------------------------------------------------------------------------------
ResultSet-EloRating>los (likelihood of superiority)
                          St St
Stockfish 040317 64 POPC     98
Stockfish 8 64 POPCNT      1

Several factors could be involved - is your machine overloaded at times, what type of opening book , how are they being compiled or your results of just 200 games could be an outlier - and so could mine, but I have run tens of thousands of games since SF 8 release and I'm confident ELO gains are being made. It is not straight line, they continually approve simplifications that are not totally ELO neutral. On the other hand, simplification patches is similar to pruning a tree and set the stage for the next round of ELO gaining patches. The SF team definitely has a well developed and proven collaborative process, certainly one that I cannot argue against.