Author: blass uri
Date: 03:50:37 10/10/98
I think that you can compress distance to conversion to distance in number of mistakes The idea is that often the computer can find the right move by evaluation function and if it is the case then you do not need the tablebases. You have to save for a position only the maximal number of times that you will need the tablebases in the future If you can decrease this number by 2 plies search then it is a proof that the evaluation function cannot help and you use the tables. If you cannot decrease this number by 2 plies search then you can use the evaluation function to play the right move. The problem with this idea is that I do not see how to compress it more by division by 2 . It is a good idea only if the static evaluation function(or the evaluation after a very smalll search) of the program does not do a lot of mistakes. It is better to use the evaluation after search of a second per position but you need a lot of time to generate hash tables in this way and you have not many years to generate the hash tables. 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.