Stefan Pohl Computer Chess

Latest Website-News (2019/03/18): Testrun of Stockfish 190310 finished - huge progress: +10 Elo to Stockfish 190227. Now +31 Elo to Stockfish 10 (!) Next testrun: Xiphos 0.5. Result not before next Monday.

Long thinking-time tournament updated. First result of a 40xxx Net - Net 41476 is at the same Elo-level as Net 32890 (used in TCEC Superfinal 14). Now, I will try a new, smaller 50xxx Net: Net 50611.

My Drawkiller Openings Project is finished. Never before any openings-set gave such low draw-rates without crunching the scores of the engines towards 50%, but instead pushing the scores away from 50%. The Drawkiller Normal- and Tournament sets nearly halve the draw-rate, compared to FEOBOS or the Stockfish Framework 8-move openings. I would never have expected, that this was possible – the Drawkiller project is really a breakthrough into another dimension of Computerchess. Learn more about Drawkiller openings in the "Drawkiller openings"- section on this website and download them here

Stay tuned.

Stockfish testing

Playing conditions:

Hardware: i7-6700HQ 2.6GHz Notebook (Skylake CPU), Windows 10 64bit, 8GB RAM

Fritzmark: singlecore: 5.3 / 2521 (all engines running on one core, only), average meganodes/s displayed by LittleBlitzerGUI: Houdini: 2.6 mn/s, Stockfish: 2.2 mn/s, Komodo: 2.0 mn/s

GPU (used by LC Zero): Nvidia Cuda Gforce GTX 950M (4GB memory, grafics: 914 MHz, memory: 1.25 GHz). LC Zero CUDA calculates around 1020 rollouts/s in the starting position (measured with "go infinte"-command until depth 28) with Netsize 20x256

Leela-Ratio: 0.67 for my tournament (opponent engines running singlethread). What is it? look here

Hash: 512MB per engine

GUI: LittleBlitzerGUI (draw at 130 moves, resign at 400cp (for 4 moves))

Tablebases: None

Openings: HERT testset (by Thomas Zipproth) (download the file at the "Download & Links"-section or here)(I use a version of HERT, where the positions in the file are ordered in a different way - makes no difference for testing-results, dont be confused, when you download my gamebase-file and the game-sequence doesnt match with the sequence of your HERT-set...)

Ponder, Large Memory Pages & learning: Off

Thinking time: 180''+1000ms (= 3'+1'') per game/engine (average game-duration: around  7.5 minutes). One 5000 games-testrun takes about 7 days.The version-numbers of the Stockfish engines are the date of the latest patch, which was included in the Stockfish sourcecode, not the release-date of the engine-file, written backwards (year,month,day))(example: 170526 = May, 26, 2017). Since July, 2018 I use the abrok-compiles of Stockfish again (http://abrok.eu/stockfish), because they are now much faster than before - now only 1.3% slower than BrainFish-compiles. So, there is no reason anymore to not use these "official" development-compiles.

Download BrainFish (and the Cerebellum-Library): here

Each Stockfish-version plays 1000 games versus Komodo 12.3, Houdini 6, Fire 7.1, Ethereal 11.25, Komodo 12.3 MCTS. All engines are running with default-settings, except: Move Overhead is set to 500ms, if an engine allows to do so.

To avoid distortions in the Ordo Elo-calculation, from now, only 2x Stockfish (latest official release + the latest version) and 1x asmFish and 1x Brainfish are stored in the gamebase (all older engine-versions games will be deleted, every time, when a new version was tested). Stockfish, asmFish and BrainFish older Elo-results can still be seen in the Elo-diagrams below. BrainFish plays always with the latest Cerebellum-Library of course, because otherwise BrainFish = Stockfish.

Latest update: 2019/03/18: Stockfish 190310

(Ordo-calculation fixed to Stockfish 10 = 3508 Elo)

See the individual statistics of engine-results here

See the ORDO-rating of the archive-gamebase since 2019 here

Download the current gamebase here

Download the archive-gamebase since 2019 here

     Program                     Elo    +    -   Games   Score   Av.Op.  Draws

   1 Stockfish 190310 bmi2     : 3539    8    8  5000    76.4 %   3325   40.4 % (new)
   2 BrainFish 180728 bmi2     : 3533    8    8  5000    77.8 %   3298   39.0 %
   3 Stockfish 10 181129       : 3508    6    6  9400    76.4 %   3289   39.4 %
   4 Stockfish 9 180201        : 3452    7    7  7000    73.3 %   3263   43.0 %
   5 Houdini 6 pext            : 3427    4    4 15400    63.8 %   3316   47.9 %
   6 LF 0.3 32930/Houdini 6    : 3419   14   14  1200    58.5 %   3356   55.5 %
   7 Komodo 12 bmi2            : 3395    5    5 10000    59.4 %   3320   49.8 %
   8 Komodo 12.3 bmi2          : 3395    6    6  8400    58.1 %   3331   49.3 %
   9 Lc0 0.20.2 N:32930        : 3318   14   14  1200    45.1 %   3356   43.7 %
  10 Fire 7.1 popc             : 3282    4    4 15400    44.8 %   3325   49.6 %
  11 Komodo 12.3 MCTS          : 3262    5    5  9400    38.7 %   3352   45.1 %
  12 Ethereal 11.25 pext       : 3261    6    6  7400    35.8 %   3374   48.3 %
  13 Laser 1.7 bmi2            : 3204    7    7  6000    30.8 %   3356   45.6 %
  14 Fizbo 2 bmi2              : 3193    6    6  8000    33.5 %   3329   37.3 %
  15 Shredder 13 x64           : 3192    6    6  8000    28.5 %   3371   40.0 %
  16 Booot 6.3.1 popc          : 3181    7    7  6000    32.1 %   3323   42.9 %
  17 Xiphos 0.4.6 bmi2         : 3172    8    8  5000    27.8 %   3350   39.3 %
  18 Andscacs 0.95 popc        : 3152    8    8  5000    23.1 %   3375   35.3 %

LeelaFish AuxEngine-parameters: Threshold=20, Depth=14. AuxEngine is Houdini 6 (with Threads=1 and Hash=1024)

The version-numbers (180622 for example) of the engines are the date of the latest patch, which was included in the Stockfish sourcecode, not the release-date of the engine-file. Especially the asmFish-engines are often released much later!!

Below you find a diagram of the progress of Stockfish in my tests since the end of 2018. 

And below that diagram, the older diagrams.

You can save the diagrams (as a JPG-picture (in originial size)) on your PC with mouseclick (right button) and then choose "save image"...

The Elo-ratings of older Stockfish dev-versions in the Ordo-calculation can be a little different to the Elo-"dots" in the diagram, because the results/games of new Stockfish dev-versions - when getting part of the Ordo-calculation - can change the Elo-ratings of the opponent engines and that can change the Elo-ratings of older Stockfish dev-versions (in the Ordo-calculation / ratinglist, but not in the diagram, where all Elo-"dots" are the rating of one Stockfish dev-version at the moment, when the testrun of that Stockfish dev-version was finished).