Author: Uri Blass
Date: 11:56:59 10/30/00
Go up one level in this thread
On October 30, 2000 at 13:59:16, James T. Walker wrote: >On October 30, 2000 at 13:44:56, Uri Blass wrote: > >>On October 30, 2000 at 13:09:23, James T. Walker wrote: >> >>>On October 29, 2000 at 10:02:24, Uri Blass wrote: >>> >>>>On October 29, 2000 at 08:36:07, pete wrote: >>>> >>>>>On October 28, 2000 at 20:15:02, Uri Blass wrote: >>>>>>I do not believe that it is only low branching factor. >>>>>> >>>>>>Genius has other problems. >>>>>> >>>>>>1)It does less extensions than other programs and cannot extend more than 12 >>>>>>plies. >>>>>> >>>>>>It canot see lines of more than 32 plies. >>>>>> >>>>>>2)There are positions that it does not understand when other programs >>>>>>understand. >>>>>> >>>>>>Other programs worked many years about improving their evaluation when lang did >>>>>>not do it and it is natural that other programs got better evaluation function >>>>>>in part of the positions. >>>>>> >>>>>>Uri >>>>> >>>>>I tend to disagree , at least I suspect your point of view is very hard to prove >>>>>. >>>>> >>>>>It does sound very logical to assume years of work on evaluation has payed off >>>>>but it is very rare to see Genius judging positions in a ridiculous way . >>>> >>>>I expect you to see cases when Genius evaluates positions wrong if you do more >>>>games. >>>> >>>>I saw cases when Genius did not evaluate correctly king attack. >>>>It falled into a king attack against chess system tal many years ago because of >>>>wrong evaluation and I expect it also to fall into king attacks against >>>>gambittiger. >>>> >>>>It did not happen in the 3 games that you posted but it does not say that it is >>>>not going to happen. >>>> >>>>There are cases when Genius3's evaluation is superior relative to Gambit but I >>>>believe that there are also cases when it is the opposite. >>>> >>>>Uri >>> >>>Hello Uri, >>>I may be wrong but I think you cannot rely on Gambit's score for positions as >>>they are kind of artificial. Similiar in my opinion to MChess pro. Sometimes >>>the scores are elevated to force the attack when it is speculation. Only my >>>opinion though. >>>Jim >> >>You cannot trust also the scores of other programs because they are not >>speculative and sometimes can evaluate 0.00 without seeing that one side has a >>winning attack. >> >>Uri > >Hello Uri, >Well probably there is no program which gives an accurate evaluation of every >position. What I mean is that most programs give an "honest" evaluation of the >position within their own limitations. Gambit I believe is one that gives an >untrue score when trying to force an attack and therefore should not be relied >on for real evaluation to determine who is winning and by how much. Again it's >only my opinion. >Jim Hello Jim, I do not know what is "honest" evaluation of the position. I do not see something that is not honest in evaluating a positional advantage as a piece advantage. It may be right and may be wrong and not evaluating positional advantage as a piece advantage may be also right or wrong. Programs with big positional scores are wrong in different part of the cases. I did not test gambittiger but I guess that there are a lot of cases when it is right and the other programs are wrong(otherwise it could not be one of the best programs). 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.