Author: David Rasmussen
Date: 14:10:29 12/21/00
Go up one level in this thread
On December 21, 2000 at 02:27:25, Sam Slutzky wrote: >Thanks to Brice Boissel, I was finally able to pinpoint a bug that has been in >the program for a long time. The bug wasn't as apparent in earlier versions, >but can occur in 2.04 now and then. Basically, when I detected >draw-by-repetition, I wasn't resetting the length of the PV properly. If the >program searched a move that led to a draw, the length of the PV should be the >length to that move, but instead, it was set to the length of the last move >searched which could have been several plies deeper. This meant the program >could have bad or illegal moves in the PV in these cases causing all sorts of >problems. The fixed version is on http://www.ssechess.com. Besides the bug >fix, this version is pretty much the same as 2.04. Now that the bug is fixed >though, I can implement all the good suggestions I have been receiving. :) >Also, there is a forum (although not a great one) on my web page for SSEChessII >discussions. > >Sam Thanks for posting this! I wish more programmers would post their bugfixes, for others to learn from. I'm not sure my program is affected by this though. I have pvlength[ply]=ply; at the start of search (root is ply 0), so I never have to adjust the pv as far as I can see. It is one of the things that confuse me when I look at Craftys code. There is all sorts of fiddling with the pv in the case of exact hash cutoff, draw by repetition etc. and calling SavePV. I don't understand why that approach is better than just having pvlength[ply]=ply; in the beginning of Search()
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.