Author: Peter Stein
Date: 23:20:38 11/21/00
Go up one level in this thread
On November 22, 2000 at 01:00:22, David Dahlem wrote: > >On November 22, 2000 at 00:49:31, David Dahlem wrote: > >> >>> >>>Ok, I think I've identified the "weights" which seem to be numbers in the range >>>of -125 to 125. What values should these weights be set to so that moves are >>>chosen based on a user specified percentage? For instance, for a set of 4 >>>possible moves I want 40, 30, 20, and 10 percent respectively as the frequencies >>>at which these moves are chosen. I've fiddled around with the weights and the >>>"book options", but can't for the life of me figure out the relationship between >>>these weights and selection frequencies (it's not any intuitive 1:1 or 1:n). >> >>I used a calculator dividing the number of games into the number of moves to get >>a percentage. >> >>Dave > >Bot I was trying to get equal frequencies so that method would't work for you. >Playing with the up/down arrows changes the numbers in the percent column. Figured out one of the problems. The "influence of learn value" slider was in the wrong position. Once this was corrected I was able to assign frequencies of my choice although with difficulty. However, I discovered another gotcha. If the tree was generated by importing from a database (rather than entering moves manually) then certain moves will be assigned a frequency. I discovered that only these moves can have their frequencies altered, i.e. moves which were not assigned a frequency by the import could not be given a frequency. I'd like to ask the authors why this has to be so difficult. Especially when older playing software with book editors let the user set the frequencies. What would be so difficult about letting the user directly specify the frequencies? Thanks for your help. This discussion has been much more helpful than the "T notes".
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.