DanaSah 9.0 and Caligula 0.9i

[pedrox]

DanaSah 9.0. This version has a new network dananet2.nnue derived from dananet1.nnue now with depth 9.

https://www.mediafire.com/file/6xs8u10d ... 0.zip/file


Caligula 0.9i. A little update.

https://www.mediafire.com/file/ghoh1nnp ... n.zip/file

[Gabor Szots]

DanaSah 9.0. This version has a new network dananet2.nnue derived from dananet1.nnue now with depth 9.

https://www.mediafire.com/file/6xs8u10d ... 0.zip/file


Caligula 0.9i. A little update.

https://www.mediafire.com/file/ghoh1nnp ... n.zip/file

Thank you Pedro.

[CMCanavessi]

DanaSah 9.0. This version has a new network dananet2.nnue derived from dananet1.nnue now with depth 9.

https://www.mediafire.com/file/6xs8u10d ... 0.zip/file


Caligula 0.9i. A little update.

https://www.mediafire.com/file/ghoh1nnp ... n.zip/file

Nice! It's already playing in my "Danger Zone" mini-tournament, live right now.

Any estimate on elo improvement compared to 8.8?

[CMCanavessi]

Is the Gaviota support working, or the engine just does not report "tbhits"?

Edit: looks like TBs are actually working (tested some positions), but no "tbhits" are being reported at all

[Rebel]

DanaSah 9.0. This version has a new network dananet2.nnue derived from dananet1.nnue now with depth 9.

Pedro, the config file states:

Code: Select all

use_nnue 	1
evalfile_path	dananet1.nnue

Should that not be dananet2 ?

Also:

Code: Select all

limitstrength 		true
elo 			1500

Should limitstrength not be false ?

[CMCanavessi]

DanaSah 9.0. This version has a new network dananet2.nnue derived from dananet1.nnue now with depth 9.

Pedro, the config file states:

Code: Select all

use_nnue 	1
evalfile_path	dananet1.nnue

Should that not be dananet2 ?

Also:

Code: Select all

limitstrength 		true
elo 			1500

Should limitstrength not be false ?

I just deleted that file and use UCI to configure everything.

[pedrox]

DanaSah 9.0. This version has a new network dananet2.nnue derived from dananet1.nnue now with depth 9.

https://www.mediafire.com/file/6xs8u10d ... 0.zip/file


Caligula 0.9i. A little update.

https://www.mediafire.com/file/ghoh1nnp ... n.zip/file

Nice! It's already playing in my "Danger Zone" mini-tournament, live right now.

Any estimate on elo improvement compared to 8.8?

[pedrox]

DanaSah 9.0. This version has a new network dananet2.nnue derived from dananet1.nnue now with depth 9.

https://www.mediafire.com/file/6xs8u10d ... 0.zip/file


Caligula 0.9i. A little update.

https://www.mediafire.com/file/ghoh1nnp ... n.zip/file

Nice! It's already playing in my "Danger Zone" mini-tournament, live right now.

Any estimate on elo improvement compared to 8.8?

Version 9 is about 60-70 points better against version 8.8. Against other engines I have not done the test.

[pedrox]

Is the Gaviota support working, or the engine just does not report "tbhits"?

Edit: looks like TBs are actually working (tested some positions), but no "tbhits" are being reported at all

DanaSah does not show information about hash tables and tablebases, I now understand that perhaps it is interesting to facilitate that it is working, perhaps I will add it in a future version.

[pedrox]

DanaSah 9.0. This version has a new network dananet2.nnue derived from dananet1.nnue now with depth 9.

Pedro, the config file states:

Code: Select all

use_nnue 	1
evalfile_path	dananet1.nnue

Should that not be dananet2 ?

Also:

Code: Select all

limitstrength 		true
elo 			1500

Should limitstrength not be false ?

I forgot to configure the "danacfg.txt" file for the xboard protocol, since I normally use the uci protocol. So effectively if the xboard protocol is used it is necessary to modify the evaluation file and put dananet2.nnue.

The executable in the zip does not have the option to limit the playing force, so it is not affected by the limitstrenght option (this is prepared for another executable that I did not include in the package, maybe a bit confusing but I tried to make it serve the same file of configuration for the 2 executables).