Author: Dann Corbit
Date: 10:48:59 07/14/00
Go up one level in this thread
On July 14, 2000 at 09:24:48, Robert Hyatt wrote: [snip] >How do you do this while "minimaxing" the book scores as the book is >created? There will be "zillions" of non-CAP positions you will reach. DO >you search _each_ one to get a score to minimax back to the root? You certainly >can't minimax the book moves at game time, it would take forever... > >Cray Blitz's old book minimax code took hours to run on a Cray. The first >time I tried to minimax a book with Crafty, it took > 24 hours with no >searching, just propagating book scores back from leaf positions to the >root... The CAP project now has every board position that has ever been played at least twice in a tournament analyzed at fast time control [Not sure how many positions it is. I have only 60 million in my possession, but I think it is probably 100 million or so]. We use this data to decide future research decisions. One outcome of the whirlwind experiment is that the data could be used to fully populate any book with "really weak" CAP data. The data could subsequently be improved, however, by an iterative algorithm. The weak data could also be combined with the 6 million+ rows of strong data by using the knowledge of data depth in the calculations. [snip]
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.