Yes tune a few parameters maybe only one or two and play a great many games (not). If you and your computer are still alive you can observe the result. For it may take some time.
By the way my engine is not able to play a game on my computer within 10 microseconds.
LMR problems
Moderators: hgm, Rebel, chrisw
-
- Posts: 558
- Joined: Sat Mar 25, 2006 8:27 pm
Re: LMR problems
Certainly it can play one in 20 seconds, though? 3 games/minute, 180 games/hr, 1440 games in 8 hours overnight.Henk wrote:Yes tune a few parameters maybe only one or two and play a great many games (not). If you and your computer are still alive you can observe the result. For it may take some time.
By the way my engine is not able to play a game on my computer within 10 microseconds.
-
- Posts: 759
- Joined: Fri Jan 04, 2013 4:55 pm
- Location: Nice
Re: LMR problems
IndeedRobert Pope wrote:Certainly it can play one in 20 seconds, though? 3 games/minute, 180 games/hr, 1440 games in 8 hours overnight.Henk wrote:Yes tune a few parameters maybe only one or two and play a great many games (not). If you and your computer are still alive you can observe the result. For it may take some time.
By the way my engine is not able to play a game on my computer within 10 microseconds.
All is playing lots of games
Personnaly now I prefer 1' + 0.25'' , 10 engines , 60 games each = 600 games played in around 24 hours
Of course with the same opennings each tests
But now we are a lot hs may be...
-
- Posts: 7218
- Joined: Mon May 27, 2013 10:31 am
Re: LMR problems
- Many parameters having many values.
- Formula or source code changes.
- Mistakes.
So this means tuning may easily takes years. Nothing for me for I'm mainly interested in programming.
Actually I should not care about if chess program plays bad. Only thing that should matter is quality of the source code.
- Formula or source code changes.
- Mistakes.
So this means tuning may easily takes years. Nothing for me for I'm mainly interested in programming.
Actually I should not care about if chess program plays bad. Only thing that should matter is quality of the source code.
-
- Posts: 362
- Joined: Thu Mar 16, 2006 7:39 pm
- Location: Portugal
- Full name: Alvaro Cardoso
Re: LMR problems
I already have a good fail high on first move percentage, if the score and pv doesn't change much I can get 95%-96% on many positions, if the position is a sharp one, with a good score, it can go to 97% or 98%.Henk wrote:If you want to make your engine really slow then you should spend much CPU time on move ordering. I remember evaluating moves at MAXDEPTH on MAXDEPTH - N to get a good value for move ordering (ouch). Best value is N = MAXDEPTH and even that made it slower. So move ordering should be cheap.
Previously I got 89%-92% so now is much better, but this doesn't come cheap, I really invested time and energy on a good move ordering scheme.
I think my problems are in the evaluation function, that is the worst part of my engine, and a source of instability.
regards,
Alvaro
-
- Posts: 362
- Joined: Thu Mar 16, 2006 7:39 pm
- Location: Portugal
- Full name: Alvaro Cardoso
Re: LMR problems
Thx Alvaro, for your help.
How is your spanush checkers engine playing now?
Maybe we can play some games throught google chat.
best regards,
Alvaro
How is your spanush checkers engine playing now?
Maybe we can play some games throught google chat.
best regards,
Alvaro