Author: martin fierz
Date: 04:40:46 11/17/03
Go up one level in this thread
On November 17, 2003 at 04:51:55, Daniel Clausen wrote: >On November 17, 2003 at 04:20:42, martin fierz wrote: > >[snip] > >>the real question should be: why do programs like fritz play these closed >>positions worse than any 2000 player? fritz' programmers surely know about >>those weaknesses, why have they never been addressed? with a whole team of >>professionals working on it... > >Because it won't give you anything for the SSDF rating list? > >It probably takes much more work than just a few if-statements here and there, >so if you're not committed to playing against humans, it probably won't happen >that fast. it probably also takes much more than just a few if-statements here and there for fritz to be the engine it is now :-) and since frans morsch has claimed to have been optimizing against human play over the last year (lame excuse for no progress or something else?), he should definitely have addressed this issue. even if it's more than a few if then else statemtents, i refuse to believe that it's not possible! another point here is that fritz even believed it had compensation for being a pawn down, when in fact it was a pawn down AND in a hopeless position too. comps do have strange non-objective evaluation terms like "a position with 8 pawns is very bad for me (with 7 only bad)", and "if my king is attacked that is worse than if my opponents king is attacked". these asymmetric, non-objective terms come from human-comp play, where they seem to help. so it's not quite true that comps are only optimized to the SSDF list. of course these IMO ridiculous eval terms only help by masking a weakness of the comps, and once the mask drops they are just as stupid as before.... cheers martin
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.