SPCC: Testrun of Stockfish 150314 finished

[pohl4711]

Testrun of Stockfish 150314 finished.

Endless RoundRobin-tournament updated, too.

http://spcc.beepworld.de


(Perhaps you have to clear your browsercache or reload the website)

[beram]

Testrun of Stockfish 150314 finished.

Endless RoundRobin-tournament updated, too.

http://spcc.beepworld.de


(Perhaps you have to clear your browsercache or reload the website)

Clearly better result for SF150314 against Komodo 8
59,6% where it was 57,8% for SF6.
Win:loss ratio against K8 now 2:1 by a drawratio of 42,6%
So this SFdev scores approx 12Elo better, looking at games vs K8

Code: Select all

Individual statistics:

1 Stockfish 150314 x64  : 3244 7000 (+3326,=2734,-940), 67.0 %

Mars 3.35 x64             : 1000 (+519,=372,-109), 70.5 %
Critter 1.6a x64          : 1000 (+499,=387,-114), 69.3 %
Fire 4 x64                : 1000 (+490,=403,-107), 69.2 %
Equinox 3.3 x64           : 1000 (+538,=377,- 85), 72.7 %
Komodo 8 x64              : 1000 (+383,=426,-191), 59.6 %
Houdini 4 x64             : 1000 (+414,=374,-212), 60.1 %
Gull 3 x64                : 1000 (+483,=395,-122), 68.0 %

2 Stockfish 6 150128    : 3242 7000 (+3294,=2774,-932), 66.9 %

Mars 3.35 x64             : 1000 (+521,=384,- 95), 71.3 %
Critter 1.6a x64          : 1000 (+483,=386,-131), 67.6 %
Fire 4 x64                : 1000 (+485,=399,-116), 68.5 %
Equinox 3.3 x64           : 1000 (+527,=393,- 80), 72.3 %
Komodo 8 x64              : 1000 (+351,=454,-195), 57.8 %
Houdini 4 x64             : 1000 (+426,=372,-202), 61.2 %
Gull 3 x64                : 1000 (+501,=386,-113), 69.4 %

[JJJ]

But it does not only play vs K8.
Last regression test is showing 16 ELO improvement over Stockfish 6. But certainly not versus different opponent.

[APassionForCriminalJustic]

But it does not only play vs K8.
Last regression test is showing 16 ELO improvement over Stockfish 6. But certainly not versus different opponent.

The improvement is still 16 Elo (the code is simply better). It is what the framework states that matters. The 16 Elo doesn't even include all of the major SMP patches, so we know that Stockfish is much more than 16 Elo stronger on high-core count hardware versus Stockfish 6 (4 cores or greater should yield differences). Stockfish 6 is already becoming old news.

[pohl4711]

It is what the framework states that matters..

Really? I doubt that. No user plays Stockfish against Stockfish. The Framework is important for the development. But what really matters for a chessengine is, how strong it plays against opponent engines. Thats why I am doing my testwork.

Stefan

[Dr.Wael Deeb]

It is what the framework states that matters..

Really? I doubt that. No user plays Stockfish against Stockfish. The Framework is important for the development. But what really matters for a chessengine is, how strong it plays against opponent engines. Thats why I am doing my testwork.

Stefan

Totally agreed....

In my private rating list I encountered a lot of surprises when playing a big pool of opponents that are even 200 Elo below the top rated engines....
Dr.D