Author: Michel Langeveld
Date: 09:13:50 01/21/00
Go up one level in this thread
>This isn't uncommon. the latest crafty has more endgame stuff. Which can >greatly affect the size/shape of the tree in random ways. Since the bench >command searches to a fixed depth, this can make it take a long time to >resolve some things. IE old code might lock on to one PV and stick, so that >the search goes quickly ply by ply. The new might oscillate between two or >more moves, which greatly slows the search down in that same position. > >If you look at your output, this happened at depth 11. I noticed also according to the output that the crafty oscillates between two moves. If Kf3 is the right answer, then it sounds strange to me that Crafty 17.5 finds the move almost immediatly and Crafty 17.6 takes a while to find it. I understand that you have changed the evaluations and that's why the output is different. I also understand if the evaluation is changed in such way that crafty can find different moves on different ply depths. My new questions is: Do you really think the endgame stuff in crafty 17.6 is better than Crafty 17.5? Because it's not proved by bench-pos #1. I know this position does not tell the whole story, and this position can be an exception to the rule. Thanks in advance! Michel Langeveld
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.