Author: Uri Blass
Date: 01:25:18 03/24/04
Go up one level in this thread
On March 24, 2004 at 04:16:42, Uri Blass wrote: >I think that some test to test the quality of free source code evaluation may be >interesting. > >The test should be done in the following way: > >1)Modify arasan's evaluation to be identical to phalanx's evaluation >Modify it also to be identical to fruit's evaluation and the same for the other >free programs. > >Do the same for other free chess programs so you have many versions for every >free program with source code. > >2)Do most of the program slower by adding loops that do nothing. >For example if you needed to do arasan 10 times slower in nodes per second to do >it's evaluation identical to resp and it was the worst case then take care that >all the program are 10 times slower in nodes per second by adding loops that do >nothing. > >3)Now start to compare between programs and find rating for every program in >blitz 1+1 time control by hundred of games(if you started with 10 free programs >then you have 100 programs) > >Note that it may be possible to have problems in part of the cases if free >source code chess programs use parts of the evaluation for pruning decision and >not only the standard chess techniques but I think that the main problem is to >modify the free chess programs with no bugs. > >How much time do you think that a good programmer need to do the last task? > >Uri I forgot to say that the test that I suggest also compare between quality of chess programs search but I am more interested in the importance of evaluation and using one program can lead to misleading results because it is possible that evaluation of program A is better only for the rule search of A and not for the rules of search of B. 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.