SPCC: Testrun of Torch 1 finished

My UHO-Top15 Ratinglist is the world's first engine-ratinglist, using UHO-openings, and the world's first ratinglist offering additionally Gamepair-statistics.

UHO-Top15 Ratinglist testrun of Torch 1 finished.

https://www.sp-cc.de

Also take a look at the EAS-Ratinglist, the world's first engine-ratinglist not measuring strength of engines but engines's style of play:
https://www.sp-cc.de/eas-ratinglist.htm


(Perhaps you have to clear your browsercache (press STRG+SHIFT+DEL) or reload the website))

Code:

      Program                    Elo    +    -  Games    Score   Av.Op. Draws

   1 Stockfish 230929 avx2    : 3827    4    4 15000    73.3%   3642   45.8%
   2 Stockfish 16 230630      : 3821    5    5 15000    72.6%   3643   46.4%
   3 Torch 1 popavx2          : 3786    4    4 15000    68.3%   3645   46.9%
   4 KomodoDragon 3.3 avx2    : 3759    4    4 15000    64.8%   3647   47.5%

Pretty cool Stefan, 15,000 games, low draw rate.

So the Torch guys (likely with full access to the source code of Lefler) improved Komodo with 27 elo.

Still 41 elo to bridge.

Well done Torch....

But, if you look at the Gamepair-rescoring of my results, Torch is still clearly weaker than Stockfish:

Code:

Torch 1
vs.                             :  games (    +,    =,   -),   (%) :   Diff,   SD, CFS (%)
    Stockfish 230929 avx2       :    500 (   32,  208, 260),  27.2 :   -186,    7,    0.0
    Stockfish 16 230630         :    500 (   31,  214, 255),  27.6 :   -157,    7,    0.0

Torch played 1000 gamepairs vs. Stockfish (500 vs. SF 16 and 500 vs. SF 230929), but won only 63. Stockfish won 515...
The Gamepair-score of Torch vs. both Stockfish is only 27.4%.

https://www.sp-cc.de/files/uho_top15_gamepair.txt

Oh la la....

Admin wrote:

Oh la la....

On the other hand: First release and Torch is already "the best of the rest" (perhaps Lc0 could be still a little better, than Torch). Definitly an impressive start! Not only a new engine, but a new team. So, IMHO Torch 1 is real impressive.

Admin wrote:

Code:

      Program                    Elo    +    -  Games    Score   Av.Op. Draws

   1 Stockfish 230929 avx2    : 3827    4    4 15000    73.3%   3642   45.8%
   2 Stockfish 16 230630      : 3821    5    5 15000    72.6%   3643   46.4%
   3 Torch 1 popavx2          : 3786    4    4 15000    68.3%   3645   46.9%
   4 KomodoDragon 3.3 avx2    : 3759    4    4 15000    64.8%   3647   47.5%

Pretty cool Stefan, 15,000 games, low draw rate.

So the Torch guys (likely with full access to the source code of Lefler) improved Komodo with 27 elo.

Still 41 elo to bridge.

Well done Torch....

according to them all of the code is original but take that how you will

Sure, they said they started from scratch, I believe them. That does not undermine what I said. In the end it's all about novelty.

Torch is now useable for analysis on chesscom:
https://www.chess.com/news/view/torch-engine-available-on-chesscom?alert=content

I tried it, and it worked for me, as a non-paying, registered user (no Premium or Diamond Membership or so). So, it is free right now.