Frank Quisinsky wrote:
I like your engine too and I am testing
I can see a jumping from version 1.29 to 1.31b from around 30 ELO. Now I am testing version 1.32, thanks for it !!!
The nalimov question is for much programmers a bigger problem. Perhaps Eugene and Andrew should give the sources absolutly free for all. So far it's not OK that different programmers can use it and other programmer have to wait or can't use it.
But I don't understand why you don't use egbbs. An other way is to contact Stefan Meyer-Kahlen ... perhaps Stefan will give a permission for the shredderbases. I don't know why Stefan hold the sources for shredderbases secret ... it would be nice for others to use it, means his big effort he do for around 4 years.
Only loud and clear thinking ... (about shredderbases).
Perhaps you can explain me why you don't use egbbs. This would be interesting for me to know it.
I have several reasons not to use egbbs. The most important ones:
1) The effect of tablebases (egbb and Nalimov) on the rating of an engine with some basic endgame knowledge seems to be small (10-15 rating points). Since the time I can spend on Protector is very limited and my main goal is to increase its rating the support for any kind of tablebases is in the lower part of my TODO list.
2) To the best of my knowlegde egbbs (Scorpio) require a system-dependant DLL. Unpleasant limitation.
3) In contrast to Nalimov tablebases the path to the egbbs is a special param which the user has to configure for every engine separately. Also unpleasant.
4) In contrast to Nalimov tablebases with distance-to-mate values the results extracted from egbbs need special treatment if a winning position is indicated. Otherwise your engine will draw several won pawnless endgames by shuffling pieces around (e.g. KBBKN, KQKBB or KNNKP -- if the winning player owns a pawn the treatment is relatively easy: add a big bonus to the normal endgame eval value, pawn pushes will do the rest of the job; KQPKQ might still need some extra eval).
Actually it's a bit ridiculous if your engine requires tablebases since this means the user will have to download tons of data in order to make a 300k program fulfil its task.
Nonetheless I would love to reactivate the tablebase support in Protector because it's cool to issue these "Mate in 95" announcements
But I won't add it without the permission from Eugene Nalimov.