Author: Uri Blass
Date: 10:21:38 12/24/00
Go up one level in this thread
On December 24, 2000 at 12:56:31, Robert Hyatt wrote: >On December 24, 2000 at 02:58:19, Uri Blass wrote: > >>On December 23, 2000 at 18:41:53, Joshua Lee wrote: >> >>>Chances are against ever solving chess let alone calculating over 10^120 >> >> >>Programs do not need to calculate over 10^120 in order to be unbeatable in games >>when the opponent is not important. >> >>Uri > >I disagree. to be "unbeatable" the program will _have_ to see to the end of >the game. Anything less leaves the opportunity for an unseen loss. The 32 piece tablebases is enough to be unbeatable and you need less than 10^120 positions in this tablebases. It is also possible that calculating 80 plies forward with the right evaluation function is enough in order to be unbeatable. Calculating 1 ply is enough to be unbeatable if you have the perfect evaluation function but it is clear that it is not practical to have it. It is not clear that the case of 80 plies is the same and it is possible that 80 plies is enough to play perfect when you use Crafty's evaluation function. You do not need to know to play every position perfect in order to be unbeatable but only the positions that you can get practically in games so finding a position when 80 plies is not enough to play perfect proves nothing. 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.