Author: Uri Blass
Date: 15:35:07 03/20/02
Go up one level in this thread
On March 20, 2002 at 18:14:02, Sune Fischer wrote: >On March 20, 2002 at 17:29:08, Uri Blass wrote: > >>On March 20, 2002 at 16:57:54, Sune Fischer wrote: >> >>>On March 20, 2002 at 16:21:53, Uri Blass wrote: >>> >>>>On March 20, 2002 at 16:07:11, Sune Fischer wrote: >>>> >>> >>>>>> >>>>>>>Even three or four days still will not play to perfection. >>>>>>>Just look at the numbers 40 to the 30 power will still not give perfection and >>>>>>>might take years to complete one move. >>>>>>>But we are working on it. >>>>>>>Bill >>>>>> >>>>>>It is not a proof that perfection is impossible. >>>>>> >>>>>>better searching rules(better pruning and extensions) together with better >>>>>>evaluation may convince programs to find always the best move. >>>>>> >>>>>>You do not need to search everything to the end of the game in order to do it >>>>>>and the question how many plies you need to search is dependent on the >>>>>>evaluation and on the extensions and the pruning rules. >>>>>> >>>>>>Uri >>>>> >>>>>Hmm, I think you need to search pretty deeply to find which is best of 1.d4 and >>>>>1.e4 :) >>>>>And you prove nothing with pruning rules and nullmove, only way to be _really >>>>>sure_ is to do a fullwidth search :( >>>>> >>>>>-S. >>>> >>>>The point is that you do not need to prove that you play perfect game >>>>in order to play perfect game and I guess that both 1.e4 and 1.d4 >>>>lead to draw. >>> >>>Sure all you need to play perfect is luck, but how will you _know_ that you play >>>the perfect game, you can't even tell which is better of e4 and d4? :) >>> >>>-S. >> >>I cannot know but if I see that I never lose games I am going to start to >>suspect that I play perfect. >> >>It is not a proof but if I lose games when I do not always win with one of the >>colors then it is a proof that I do not play perfect. >> >>Uri > >Well unless you're playing perfect opponent, then you might win (almost) every >game without perfect play. > >I think this is the problem we face today with Kasparov, how close is he to >perfect play. We can't really say since there is nothing above him to make a >reference to. >Suppose the engine you build is 3500, it will win almost all games with just a >seldom draw now and then, but if perfect play is 4000 there is still a long way >to go, you'll just never realize it. >Anything that is 500 elo stronger than the opponents will appear almost perfect >IMO. > >-S. If I see that the results of it against itslef is always the same and more time does not help then I am going to suspect that it plays perfect. Today things like this do not happen. 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.