Author: Sune Fischer
Date: 08:56:44 10/04/02
Go up one level in this thread
On October 04, 2002 at 11:42:09, Omid David wrote: >On October 04, 2002 at 11:38:01, Vincent Diepeveen wrote: > >>On October 04, 2002 at 11:36:16, Omid David wrote: >> >>>On October 04, 2002 at 11:33:00, Vincent Diepeveen wrote: >>> >>>>On October 04, 2002 at 11:25:12, Daniel Clausen wrote: >>>> >>>>>On October 04, 2002 at 11:22:05, Vincent Diepeveen wrote: >>>>> >>>>>>On October 04, 2002 at 10:34:43, Omid David wrote: >>>>>> >>>>>>had fritz implemented such a scheme it would play 300 points weaker >>>>>>anyway. >>>>> >>>>>"Obviously" it's "impossible" to implement such a scheme, because you've tried >>>>>that already for 5 minutes and it didn't work? SCNR :) >>>>> >>>>>Sargon >>>> >>>>I played with it with DIEP a few years ago and it performed horrible. >>> >>>If you apply such method to each and every node in the search tree, it's natural >>>that the whole search is slowed down in a horrible manner. >> >>no you really are an idiotic programmer if you slow down that >>much by just a few heuristics. > >"just a few heuristic" isn't the case here. It's a complicated (pretty >time-consuming) algorithm, and this, to make sure its results are 100% correct. > >>I already do a lot of scans on the >>board of diep. a full eval is like 100k clocks. so a scan over >>the board is not so expensive. i already scan the pawn structure >>extensively in fact. it's just a few functions i need to turn on >>to find otu things. >> >>>> >>>>when you have code that 'overrules' other code then you take away hundreds >>>>of patterns which normally compensate knowledge. it means that if that >>>>single pattern is incorrect, that you lose directly a game. >>> >>>Of course. And that's why your heuristic should be 100% correct and reliable. >> >>the problem of the whole chess game is that you have never 100% correct >>and reliable evaluation except if your opponent is mated. > >Wrong. I've conducted hundreds of tests, and in no single case has my heuristic >returned an inaccurate result. There exist blockades that it fails to detect, >but it never declares a false draw. Perhaps this is too restrictive too. You could have some confidence variable, if you draw detector cannot say with 100% certainty that it's draw, then interpolate the evaluated score with the confidence level: score=eval()*(100-confidence)/100. It might help the engine to steer clear of those drawish positions? -S. >Omid. > >> >>>Omid.
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.