Author: Uri Blass
Date: 21:17:31 10/11/04
Go up one level in this thread
On October 11, 2004 at 20:05:04, Robert Hyatt wrote: >On October 11, 2004 at 15:17:20, Uri Blass wrote: > >>On October 11, 2004 at 14:55:14, Pallav Nawani wrote: >> >>> >>>I was testing new version of my engine on WAC and I got a surprise: >>>Natwarlal scores 263/300 in 1 second searches and 253/300 in 10 second searches! >>>I am using 1.4Ghz athlon. >>> >>>I have read on ICC people talking about WAC busts. Are there any wrong solutions >>>in the WAC test suite? Should I switch to WACnew (assuming that it is a >>>corrected version of WAC)? >>> >>>Best regards, >>>Pallav >> >>There are cases when there is more than one solution and this is the reason for >>wacnew but if your engine scored less in WAC in 10 second search it probably has >>a bug and you should look at the positions it solved in 1 second and did not >>solve in 10 seconds in order to detect the bug. >> >>Uri > > >Might not be a bug either. Might be just a case of right move, wrong reason, >and at 10 secs/move, wrong move for right reason. IE a deeper search might >produce a different move where a shallow search produced the right move for the >wrong reason. It is possible but the more common case is the case when the program did not see the right move and changed it's right for the right reason. It is logical to have slightly less solutions in 13 seconds per position relative to 12 seconds because of that reason and the reason of the few mistakes in WAC but if you have significantly less solutions in 10 seconds relative to 1 second then the only logical explanation is a bug. Uri
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.