Author: Dann Corbit
Date: 18:20:57 01/24/06
Go up one level in this thread
On January 24, 2006 at 21:13:52, Rolf Tueschen wrote: >On January 24, 2006 at 20:51:07, Dann Corbit wrote: > >>On January 24, 2006 at 20:22:58, Rolf Tueschen wrote: >> >>>Could you also lead me to the other necessary ingrediants of a chessengine? I >>>plan to concentrate myself fully on the chess knowledge I want to implement. But >>>this in a later Beta version in Spring 2006. For now I would be satisfied with >>>an engine that already plays at 2500 Elo but still without the later fine-tuning >>>and ANTI-GM code for all. I know that I wont win Torino 2006 but then I feel >>>that Olympic spirit which says Participating is more valuable than Winning. Wont >>>you agree with me? >>> >>>For me this is all also important so that I can speak with all the collegues >>>from an equal level if I once have established my new chessengine. I dont want >>>to be a newbie anylonger. Sigh. >> >>This thing does not do bitboards at all. >>It uses 0x88 board representation. >> >>It does not contain an evaluation function. >>It does not contain a search function. >>It contains a move generator, and that is all. >> >>For the evaluation function, you can get a working one just by counting the >>wood. Then add stuff to make is smarter as you like. >> >>For the search function, look on Bruce Moreland's page about how to do pvs >>searching. It's not difficult. >> >>The basic idea of pvs searching is that you do a normal alpha-beta search of >>your predicted move. Then, you do a zero window search of all the non-pv nodes. >> That makes the search must faster if you are any good at guessing the right >>node to start with. >> >>After you do that, add in IID, which is trivial to implement and will make a >>poor searcher search much better and do better move ordering. > > >Dann, > >you sound as if you were familiar with all that. You know I planned to >concentrate myself on the more chessic content decisions. You sound as if you >were retired enough so that you could enjoy being busy here or there. How about >doing all the more, errh, routine sort of stuff for our common project? Then I >could jump onto this tuning stuff what Thorsten does in favor of Hiarcs. Man, we >two could well appear at the next WCCC! How about you? I'm game. No commitments from me. I may help you improve a chess engine if you start to write one. But I get pulled in many different directions and I never know if I will be able to assist or not. I have a few irons in the fire right now and so I can't really promise much. But if you start to write a chess engine, you can find lots of help here or in the Winboard forum. >All the above what you mentioned wasnt exactly what I would call "easy to do". >It would cost me almost a month to get into that deep enough. As I said for a >man from chess this is terrible stress! Perhaps you want a different approach. You might check SourceForge for a completed engine with a license style to your liking. Then, modify the evaluation function only. If that is your only interest you will skip a lot of steps. Be aware, though, that search is also knowledge and if the search is inferior the knowledge is inferior (as an analog, imagine an opponent who can never see more than two moves ahead is playing someone who can consistently see 7 full moves ahead -- even if he has more fundamental understanding, he will take a real beating from time to time). >How about Torino, Dann? Perhaps they have a prize for Senior Entries? Who knows? I suspect you might want to try a different track. Have you done sourceforge searches before? Have you written C or C++ code before? >Sincerely, >Rolf
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.