CEGT-40/4: actual tests

[Wolfgang]

RubiChess 2.1NN: https://cegt.forumieren.com/t1485-testi ... 2-1nn#2891

Berserk 3.3.0: https://cegt.forumieren.com/t1484-testing-berserk-3-30

Igel 3.0.0NN: https://cegt.forumieren.com/t1483-testing-igel-3-0-0nn

Stoofvlees II a18: https://cegt.forumieren.com/t1474-testi ... ees-ii-a18

Tournament with engines below 1700 CEGT-ELO: https://cegt.forumieren.com/t1480-tournament

[Wolfgang]

Fire 8.1: https://cegt.forumieren.com/t1486-testing-fire-8-1

Dragon 1.0 MCTS Linux: https://cegt.forumieren.com/t1487-komod ... rison#2895

[lkaufman]

Fire 8.1: https://cegt.forumieren.com/t1486-testing-fire-8-1

Dragon 1.0 MCTS Linux: https://cegt.forumieren.com/t1487-komod ... rison#2895

Am I correct in saying that the processor used for the Dragon test was too old for the AVX2 version? If so that would explain why the elo gain, although quite large, is still way below the elo gain on normal Dragon (on machines with AVX2).

[Wolfgang]

No.
ALL (!) tests with Dragon were played with the AVX2 compile.

We now have:

Dragon MCTS Linux: ~3462 (600 games only so far)
Dragon MCTS Windows: 3371
14.1 MCTS: 3299

So the ELO gain from Dragon MCTS Linux to (buggy) Dragon MCTS Windows is +91 and +163 to 14.1 MCTS

[Wolfgang]

DanaSah 8.8 x64: https://cegt.forumieren.com/t1488-testing-danasah-8-8nn

Pedone 3.1 x64 1CPU: https://cegt.forumieren.com/t1490-testing-pedone-3-1nn

Seer 2.0.0NN x64 1CPU: https://cegt.forumieren.com/t1491-testing-seer-2-0-0nn

TOP-Engines @ 2CPU: https://cegt.forumieren.com/t1492-testi ... -2cpu#2902

[Wolfgang]

...
Seer 2.0.0NN x64 1CPU: https://cegt.forumieren.com/t1491-testing-seer-2-0-0nn
...

Switched to v2.0.1 due to the TM bug, all games replayed
https://cegt.forumieren.com/t1491-testing-seer-2-0-1nn

[connor_mcmonigle]

...
Seer 2.0.0NN x64 1CPU: https://cegt.forumieren.com/t1491-testing-seer-2-0-0nn
...

Switched to v2.0.1 due to the TM bug, all games replayed
https://cegt.forumieren.com/t1491-testing-seer-2-0-1nn

Thanks for testing again and my apologies for creating a bugged release. I'll be curious to see how the TM fixes influence the your elo estimate. Your elo for v2.0.0 is consistent with my expectation which is surprising.

[Wolfgang]

Komodo Dragon 2.0

default: https://cegt.forumieren.com/t1496-testi ... -mcts#2909

MCTS: https://cegt.forumieren.com/t1496-testi ... -mcts#2910

[lkaufman]

Komodo Dragon 2.0

default: https://cegt.forumieren.com/t1496-testi ... -mcts#2909

MCTS: https://cegt.forumieren.com/t1496-testi ... -mcts#2910

So far +9 and +104 over Dragon 1 Windows in standard and MCTS mode; my predictions were + 10 and +100, not too bad!

[Wolfgang]

New Tournament with 4 new engines/versions:
- Stash 30.0
- Mr.Bob 1.0.0
- Clover 2.2
- Cheng 4.41
https://cegt.forumieren.com/t1502-tournament#2917