Author: Dieter Buerssner
Date: 06:08:55 03/20/05
Go up one level in this thread
On March 20, 2005 at 03:33:05, Fabien Letouzey wrote: Hi Fabien, >Did you store separate drafts for the two bounds? Yes. I had one draft/score for lower-bound and exact, and one draft/score pair for upper bound. Only one move, for the lower-bound. Hmmm, perhaps this was bad. Normally (with the hash entry with only one score) my engine also stores a move for upper bound (the best move of a fail-soft search, which surprisingly often is really a good/the best move, and in tests done long ago, this was clearly better than not storing a move for fail low). As I said, for some reason in general depth preferred seems more critical, than always replace. I thought, the problem could be, that with depth preferred you have deep entries, that you keep. But the situation changes, and those entries have the wrong bound and are useless for cutoffs. So I hoped, with 2 bounds, I could tune the HT a bit more on the direction of depth preferred (and it even seemed to work a bit in some test positions), but in practice it seemed almost 100 Elo worse. It was a bit frustrating, because I could not find anything. Perhaps I gave up too early. Cheers, Dieter
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.