Author: Roman Hartmann
Date: 04:55:53 02/20/06
Go up one level in this thread
On February 20, 2006 at 07:50:52, Tord Romstad wrote:
>On February 19, 2006 at 22:11:11, Sean Mintz wrote:
>
>>As I try to figure this out, perhaps you have an idea as to what it might be:
>>
>>[D]8/2p5/3p4/KP5r/1R3p1k/8/4P1P1/8 w - -
>>
>>depth correct what I get
>>
>>1 14 14
>>2 191 191
>>3 2,812 2,812
>>4 43,238 43,238
>>5 674,624 674,624
>>6 11,030,083 11,030,078
>>
>>So, at a depth of 6 ply, I come up 5 moves short.
>
>I don't have any good advice about this particular position, but this is
>how I solve this sort of problems in general:
>
>First, you need an open source program which you know give correct perft
>scores (i.e. Crafty). Modify this program to print perft numbers not only
>print the perft number for the position for the board, but also each
>subtree below each move at ply 1. For the position above, you should get
>output similar to the following:
Or just use Sharper or Roce which both have this feature ('divide #') included
already. Very usefull to track bugs like this down.
Roman
>Subtree below move e2e3: 745505
>Subtree below move g2g3: 271220
>Subtree below move e2e4: 597519
>Subtree below move g2g4: 892781
>Subtree below move b4b1: 1160678
>Subtree below move b4b2: 818501
>Subtree below move b4b3: 941129
>Subtree below move b4a4: 745667
>Subtree below move b4c4: 1027199
>Subtree below move b4d4: 957108
>Subtree below move b4e4: 860971
>Subtree below move b4f4: 174919
>Subtree below move a5a4: 868162
>Subtree below move a5a6: 968724
>perft(6) = 11030083
>
>Make your own program print similar statistics for each move. For at least
>one of the moves, the numbers will differ between the two programs. Execute
>this move in both programs, and do a perft(5) from the resulting position. By
>using this procedure recursively, you will quickly be able to zoom in on the
>exact position where your program generates too few (or too many) legal moves.
>
>Good luck!
>
>Tord
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.