Author: Guillaume MOYA
Date: 00:20:27 02/24/05
Go up one level in this thread
On February 23, 2005 at 11:26:47, Mathieu Pagé wrote: Hi guys, thanks a lot I fixed my alpha beta functions, and now, there are cut-off, there was a bug. Now I try different ways to sort the generated move, the quicksort method is rather slow, I'm going to have a try with radix sort, or just moving the 3 or 4 best moves to the beginning of the list. Thanks again for your help, Guillaume PS : I will have a look to your fen Mathieu. >On February 23, 2005 at 03:14:56, Guillaume MOYA wrote: > >>On February 22, 2005 at 13:18:17, Pallav Nawani wrote: >> >>Okays, >> >>I launched my program with the start position, to see how many nodes my program >>goes through, and I got exactly 206.603 nodes (i.e. the full tree). It's without >>moves order, so I suppose my alpha beta function is bug free. I will now look >>forward to implement a quick and efficient move order. >> >>Cheers, >>Guillaume >> >>>Have you checked the move make - unmake functions of your program to perfection? >>>If not, this may help: >>>http://homepages.caverock.net.nz/%7Epeter/perft.htm >>> >>>Best regards >>>Pallav > >Bonjour Guillaume, > >First, The perft command is supposed to be use with Minimax since not every >implementation of AlphaBeta will return the same numbers of nodes. If your >AlphaBeta return 206603 as the number of nodes visited you _DO_ have a bug, >since it mean that your AlphaBeta algorithm work as a minimax (it means youre >cutoff condition is never met). > >Secondly a "perft 4" on the starting position is far from sufficient to declare >a move generator bug free. You should use this position (from the same web page) > at a depth of at least 5: > >[D]r3k2r/p1ppqpb1/bn2pnp1/3PN3/1p2P3/2N2Q1p/PPPBBPPP/R3K2R w KQkq - - 0 1 > >Bonne chance :) > >Mathieu Pagé
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.