Author: Johan de Koning
Date: 15:16:46 04/04/04
Go up one level in this thread
On April 03, 2004 at 11:23:25, Marc Bourzutschky wrote: >On April 02, 2004 at 20:17:44, Johan de Koning wrote: > >>On April 02, 2004 at 04:58:25, Sune Fischer wrote: >> >>>On April 02, 2004 at 02:13:53, Johan de Koning wrote: >>> >>>>But as Theron pointed out some years ago, one should avoid *any* >>>>probe inside a search. >>> >>>Do you recall the argument? >> >>Blindly probing after captures results in lots of redundant probes. >>Most positions are way off balance (outside [alpha,beta]) and are >>interesting only if a 1 or 3 ply tactic exists. >> >>About 2 years ago Chritophe posted he was working/planning on a >>set of rules to decide for each material config whether to probe >>or not (depending on local depth I guess). Then he went on doing >>Palm stuff and other, more important, improvements. >> >>The idea is sound I think, because 1 probe that misses the EGDB >>cache is already awfully expensive. But on the other hand, with >>6 men, building a set of rules (more imprtantly exceptions) will >>be quite a daunting task. >> >>... Johan > >A set of general rules will be difficult, but it is a fact that when analyzing >endgame studies programs that probe tablebases beyond the root perform >significantly better, because almost by definition the probed positions will >often be non-trivial. That's an interesting observation. Yet another possible paramater to tune EGDb probing, I guess (fear). >But perhaps those of us who use comps for such analysis >are in the minority compared to people who run engine-engine "tournaments" or >hardware contests on the servers... That would be a safe bet. Particilarly since it is virtually impossible to disprove. :-) ... Johan
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.