Author: Uri Blass
Date: 04:36:05 05/06/02
Go up one level in this thread
On May 06, 2002 at 06:00:18, Thorsten Czub wrote: >On May 06, 2002 at 01:05:23, Uri Blass wrote: > >>I think that programs like hiarcs and shredder that are called knowledge based >>programs are only 2 more programs that their main advantage relative to most of >>the amatuers is being better in tactics. >> >>Uri > >i have to disagree. > >1. shredder is not "good in tactics". tactics ís in fact one of its weaknesses. > >2. hiarcs is a knowledge program. > > >the main difference between a good search based program and a knowledge based >program is IMO not how to evaluate ONE position ACCURATE and right. > >its not a static feature ! > >Its a dynamic feature. Hiarcs failed in the dynamic test against smirin. It failed to evaluate fortress position as a draw. Humans can know that it is a draw not based on static evaluation but based on seeing that they can make no progress. They can learn from search about the evaluation. I did not test shredder or hiarcs and I prefer to test my program that is clearly weaker than the commercial programs. I believe that the main reason that my program is clearly weaker than the commercial programs is tactics and I have almost no evaluation in my program. You can see games of my program at 40/40 time control in http://home.hccnet.nl/leo.dijksman/index.html The version that is playing does not use hash tables,null move or pondering or book but inspite of all the problems it seems that it is going to promote to the 4th division(it has 19.5 from 22 games). My latest tested version is using null move pruning and seems to be significantly better than the version that is playing in the 5th division. I hope that hash tables are going to give my program another significant improvement. 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.