Author: Scott Gasch
Date: 14:46:35 12/31/00
Go up one level in this thread
Hi, Sorry to followup my own note... but I've been experimenting with this in the past day and wanted to share some observations. First, I am trying out extended futility pruning as described by Heinz. As you might expect, when this is enabled the first move beta cutoff percentage drops anywhere from 2%..10% in my tests. This makes sense to me because I am pruning away whole branches in which a fail high is very likely. With this heuristic turned off my results are over 90% in most positions. Interestingly this feature also costs my engine about 10k nps because I do not prune checking moves but cannot tell if a move is checking until it is played. This leads to a bunch of MakeMove ... UnmakeMove (and prune) which causes the speed to suffer. It's just a question of whether the smaller tree size is worth the rate drop. Scott
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.