Author: Pascal Coupet
Date: 23:36:38 11/25/99
Go up one level in this thread
On November 25, 1999 at 10:59:39, Andreas Stabel wrote: >On November 25, 1999 at 10:11:58, Ulrich Tuerke wrote: > >>I find your results very interesting. >> >>What puzzles me, is the large branching ratio from one iteration to the next one >>which you obtained. I'd thought, that you get a factor of 4 to 5, provided you >>have a resonable move ordering, whereas you have factor of ~20. Could it be that >>your move ordering is very poor ? >> >>I assume, with an improved move ordering, the effects, which you have observed, >>will also become less drastic. >> >>Nevertheless, it might be worthwhile to try your suggestion regarding en >>passant. >> >>Thanks, >>Uli > >Remember that this is NOT a chess playing program, only a pure tree scanning >one. To get a branch factor of 4 to 5 you need an evaluation and an alpha- >beta algorithm which I don't have in this system. > >Your observation that alpha-beta and a good evaluation and move ordering >will reduce the effects I have found is of course correct. But I still >think clearing the en passant target square when it is not necessary is >so simple that it may have an effect on the speed of a chess program. > It is important to set the en passant status correctly for the hashtable too. If you set it incorrectly into the hash key, you will represent the same position with 2 different keys. It is probably not very important for your hash table but you will incorrectly detect repetitions (if you use hash keys into your repetition code). It can be even worse: a pawn can be in position to take ne passant but this move may be illegal; if the king is in check or if the en passant move put the king in check. So here again, there is no en passant possibility in the position and to repeat it 2 more times will be enougth to reach a draw, according to the rules of chess. I'm not sure that all chess programs handle that correctly ;=) Pascal >Regards >Andreas Stabel
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.