New engine releases & news H2 2022

[RubiChess]

Engine needs a grace margin of 1s to not lose on time??
Come on, you can do better.

[Gabor Szots]

Captain 3.0

https://github.com/n4h/Captain-chess

[MOBMAT]

I have tried Rebel 16.1 under Arena on a Windows 10 machine.
I loaded some FENs and tried to search and Rebel starts but then locks up.
I have Arena set to use Rebel as an UCI engine.
Has anyone else had any issues?

V

[Rebel]

I have tried Rebel 16.1 under Arena on a Windows 10 machine.
I loaded some FENs and tried to search and Rebel starts but then locks up.
I have Arena set to use Rebel as an UCI engine.
Has anyone else had any issues?

V

Please post some of these FEN's.

[Guenther]

Engine needs a grace margin of 1s to not lose on time??
Come on, you can do better.

fully agreed

[Gabor Szots]

Captain 3.0

https://github.com/n4h/Captain-chess

Captain 3.1 is out, same as 3.0 with version names correctly displayed.

[Modern Times]

Engine needs a grace margin of 1s to not lose on time??
Come on, you can do better.

Agreed. In Cutechess I have a 500ms margin set and that is as far as I'm prepared to go. I very rarely have an engine lose on time.

[chrisw]

Engine needs a grace margin of 1s to not lose on time??
Come on, you can do better.

Agreed. In Cutechess I have a 500ms margin set and that is as far as I'm prepared to go. I very rarely have an engine lose on time.

Hahaha!! The exponential effect of Chinese Whispers.

Rebel does NOT need a timemargin of 1s to not lose on time.
Rebel will be fine with a timemargin of 500ms, or even 250ms

Testing at 20+0.05 with timemargin set to 1000 resulted in ZERO time forfeits in 24000 games.

Testing at 20+0.05 with timemargin=0 results in time losses, unsurprisingly.
Testing at 20+0.05 with timemargin=100 results in 1 time losses per 100 games.
Testing at 20+0.05 with timemargin=250 results in 0 time losses per 100 games.
Testing at 20+0.05 with timemargin=500 results in 0 time losses per 100 games.

Set timemargin to what you like, 250 or 500 should be okay, 1000 even better, although high concurrency settings and/or slow old PC's might need more. Basically it is not really an engine programmer's problem to second guess the communication overhead between engine and GUI, nor the speed of the PC used, nor the concurrency setting.

We can make a worst case estimation for those people who test with timemargin=0, and produce a time-crippled executable, but that doesn't seem too sensible.

[Modern Times]

Excellent, that looks all good

[chrisw]

Engine needs a grace margin of 1s to not lose on time??
Come on, you can do better.

Indeed, we already do.
Funnily enough, I thought I'ld put your RubiChess through the same testing procedure as Rebel and Chess System Tal.
At timemargin=0, lots of forfeits, which is to be expected.
At timemargin=100, 2 forfeits per 100 games, twice the rate of Rebel which forfeits at 1 per 100 games.
At timemargin=250, all seems okay.

Happy New Year to you!