Author: Tony Werten
Date: 01:06:50 04/12/02
Go up one level in this thread
On April 12, 2002 at 04:00:51, Artem Pyatakov wrote: >Hi everyone! > >After a long break (because of other work), I am back to working on my chess >program!! > >I am currently picking up where I left off - opening book. So, I've got several >questions for the experts (and simple good listeners) out there: > >1) I've got my program basically working with the MEDIUM opening book from Dr. >Hyatt's site (although move selection based on the stored book info still needs >to be improved). So question: which PGN files should I use to make my final >large opening book? There are so many to pick from: large, enormous, wall, >rock... Which seem to work better? Which seems to work better for the weaker >engines (around 2350 on ICC)? > >2) Has anyone experimented with using opening book stuff anywhere besides at the >root of the search? My implementation allows to check if the position is in the >book very quickly, so may be I could use this to guide the search (like by >setting higher priority for some moves) at plies 1-3 or so? Is this a bad idea? It helps your engine to get back in the book. > >3) I know it's a big problem to find an opening book which does not leave the >engine in positions which it does not understand... Has anyone tried to prune >the opening book at construction time by actually evaluating the book positions? Yes >I know this is computationally intensive, but I've got acccess to a cluster of >computers, so I should be able to do this. A cluster helps. I had my book build over a network, with max 6 machines working on it for a couple of 1000 hours of cpu time. >Or may be I should evaluate the book >move briefly during game play before sticking to it? Then it's too late cause you're already in the variation. Tony > >Thanks in advance. > >Artem
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.