Author: Uri Blass
Date: 11:29:13 11/11/00
Go up one level in this thread
On November 11, 2000 at 11:53:16, Hermano Ecuadoriano wrote: >On November 11, 2000 at 09:35:06, Uri Blass wrote: > >>On November 11, 2000 at 08:24:24, Thorsten Czub wrote: >><snipped> >>>the hiarcs of today is the result. >>>a knowledge based program, the version 7.32 gets normally outsearchered by the >>>latest fritz and junior programs. but hiarcs is still >>>very good. >> >>I disagree here. >> >>I do not think that Junior outsearches Hiarcs at tournament time control. >>I believe that one of the main problems of hiarcs is the hash tables bug. > >I didn't know about a bug. >Would you please describe it? The bug is the fact that hiarcs cannot see things in games that it has no problem to see when it analyze a position with no previous game. I found cases when hiarcs did tactical mistakes that cannot be reproduced without a game when it does not learn from previous search. I found cases when Hiarcs failed low only at big depth because of this bug and had not enough time to solve the fail low and find a better move when it could fail low at low depth or even avoid considering the bad move as best without learning from previous search. In most of the cases it has no problem to solve the fail low but even in these cases it is losing time on the clock. It happens mainly at long time control. >Do you know if the author is fixing it? No 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.