Arasan 24.2

[jdart]

FYI if you are experiencing issues, adding "-t" to the Arasan command line will produce additional debug info, which can be helpful in diagnosis.

> in Cutechess it doesn't work
what's the issue? When I run in UCI mode under cutechess-cli, I see learning activity in the logs, such as:
395957 <Arasan-24.2.1(1): info string learning position, fen = r4rk1/pp1nqppp/2pp4/4p3/1PP1P3/P1NPPQ1P/6P1/R4RK1 w - - score = +0.32 depth = 24

FYI I have done a test match on Linux, 24.2.1 vs. 24.1, 1:0+0.8 time control. Results:
LLR=3.08 [-2.94, 2.94] 3252 games, score: 51.25%
result=H1
ELO: 8.66 +-5.1 (95%)
LOS: 100.0%

[peter]

> in Cutechess it doesn't work
what's the issue? When I run in UCI mode under cutechess-cli, I see learning activity in the logs, such as:

Tried only Cutechess- GUI, no .lrn- file is created by 24.2.1.

[Modern Times]

This may not be a complete fix: Graham was saying he still had issues. But I'm releasing it because it does fix a real problem.

Well, ChessGUI is the GUI where there is the greatest potential for time losses, as it is I think Graham has to set a huge overstep margin to mitigate it. But many engines do run just fine there.

[peter]

This may not be a complete fix: Graham was saying he still had issues. But I'm releasing it because it does fix a real problem.

Well, ChessGUI is the GUI where there is the greatest potential for time losses, as it is I think Graham has to set a huge overstep margin to mitigate it. But many engines do run just fine there.

In 350 game run by Cutechess GUI no time loss, yet 1 after 10 games in Shredder (13), regards

[jdart]

I have made another patch release (24.2.2) to fix UCI position learning (was working with ponder on, but not off). Any other fixes will have to wait until the next major release.

Note that under UCI, the learning info is reloaded from disk after receiving "ucinewgame." I am not sure all interfaces send this. cutechess-cli does, as does Arena.