Kayra TR vs Stockfish:

[mehmet123]

Kayra TR vs Stockfish:

Program Elo + - Games Score Av.Op. Draws

1 Kayra TR 0.90 : 3002 3 2 1000 50.2 % 3000 98.6 %
2 Stockfish 281026 : 3000 2 3 1000 49.8 % 3002 98.6 %


Individual statistics:

1 Kayra TR 0.90 : 3002 1000 (+ 9,=986,- 5), 50.2 %

Stockfish 281026 : 1000 (+ 9,=986,- 5), 50.2 %

2 Stockfish 281026 : 3000 1000 (+ 5,=986,- 9), 49.8 %

Kayra TR 0.90 : 1000 (+ 5,=986,- 9), 49.8 %


Game Conditions: Cute Chess Gui, Core-i7 12700h, 1 Cores, 1 min + 1 sec TC, Balsa Opening Book (5 moves), 128 Mb Hash, Ponder Off
https://www.mediafire.com/file/30yk5be6 ... 3.pgn/file


Kayra TR is a chess engine based on Triumviratus. There haven't been many changes to the original Triumviratus engine, as I haven't made any modifications to the source code. I hope to do that in future versions of Kayra TR.
However, I have modified most of Triumviratus's search parameters. I have made some adjustments to more than 800 parameters.

[mehmet123]

Kayra TR 0.90 Engine:
https://www.mediafire.com/file/ncmn0du0 ... e.zip/file

Kayra TR 0.90 Source:
https://www.mediafire.com/file/r48oqr7z ... e.zip/file

[NathanDrake]

Kayra TR vs Stockfish:

Program Elo + - Games Score Av.Op. Draws

1 Kayra TR 0.90 : 3002 3 2 1000 50.2 % 3000 98.6 %
2 Stockfish 281026 : 3000 2 3 1000 49.8 % 3002 98.6 %


Individual statistics:

1 Kayra TR 0.90 : 3002 1000 (+ 9,=986,- 5), 50.2 %

Stockfish 281026 : 1000 (+ 9,=986,- 5), 50.2 %

2 Stockfish 281026 : 3000 1000 (+ 5,=986,- 9), 49.8 %

Kayra TR 0.90 : 1000 (+ 5,=986,- 9), 49.8 %


Game Conditions: Cute Chess Gui, Core-i7 12700h, 1 Cores, 1 min + 1 sec TC, Balsa Opening Book (5 moves), 128 Mb Hash, Ponder Off
https://www.mediafire.com/file/30yk5be6 ... 3.pgn/file


Kayra TR is a chess engine based on Triumviratus. There haven't been many changes to the original Triumviratus engine, as I haven't made any modifications to the source code. I hope to do that in future versions of Kayra TR.
However, I have modified most of Triumviratus's search parameters. I have made some adjustments to more than 800 parameters.

Hi, did you actually perform any tuning of the parameters or have you just multiplied them randomly?

686 of the 1119 parameters differ, but 401 of them only in their min/max bounds, which do not change play. 285 default values changed, and 253 of those are our value times 0.97 or 1.03, about half up and half down. Did you actually tune them (SPSA or similar, at which time control), or were they scaled by hand?

In our test, same code and net, your values against ours: -4.4 +- 3.5 Elo over 9,414 games at 10+0.1.

If you want to experiment with the values, I suggest doing some actual tuning, so that it can be helpful for Triumviratus. And avoid changing them just to clone the engine.

[j.t.]

lol