Author: Andreas Herrmann
Date: 07:04:50 07/17/03
Go up one level in this thread
On July 17, 2003 at 09:55:43, Uri Blass wrote: >On July 17, 2003 at 09:51:46, Andreas Herrmann wrote: > >>On July 17, 2003 at 02:26:19, Albert Bertilsson wrote: >> >>>A great way to find buggs in move generation is this: >>>1. Have a command that instead of calculating perft calculates the perft values >>>for all the children, and prints the list like this: >>>b1c3 9755 >>>b1a3 8885 >>>g1h3 8881 >>>g1f3 9748 >>>a2a3 8457 >>>... >>>2. Have a correct version of your program or another program that can do this >>>too. >>>3. When there is a difference at any perft value you see exactly which move that >>>leads you closer to the bug, do the move and calculate the perft values with >>>depth - 1, very soon you'll be at level 1 and when you compare the list of >>>children you'll see which move is missing, or illegal. >>> >>>If you need an engine to compare with you can use Sharper, I call the command >>>"divide". >>> >>>http://www.albert.nu/programs/sharper >>> >>>/Regards Albert >> >>Hi Albert, >> >>you can get the same results, if you do one possible move from starting position >>and then start perft n-1. Just do this for every legal move from the starting >>pos. and you have the same results. So you can compare it with each program, >>which has perft implemented. >> >>Andreas > >Yes but doing this is a work and it may be better to have the list of moves and >number in 2 windows so you can compare fatser to see the mistake in case that >the numbers are different. > >doing and undoing all the legal moves and calculating perft may take >significant time and I say it from experience because I did it when movei had >problems with perft. > >Uri >Uri off course. I just tryd to explain, how to compare the results with other engines, that havn't implemneted it in that way. Andreas
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.