Author: Dann Corbit
Date: 12:44:18 02/01/00
Go up one level in this thread
On February 01, 2000 at 01:57:38, David Blackman wrote: >On January 31, 2000 at 16:18:35, Dann Corbit wrote: >>On January 31, 2000 at 09:44:09, Harald Faber wrote: >>>BTW has anyone done the conversion from the book.bin and books.bin into a >>>Fritz-tree? I guess not, I don't know of any converter that does this job, and >>>even creating the tree from the pgn-file does not take the priorities... >> >>I don't think it is invertible. The hash is stored in the book, not the >>position. You would have to write out some kind of key-file at creation time. > >Start at the starting position. Generate all legal moves. Try each legal move >and then generate the hash, see if it matches any in the book. For the ones that >do, generate all legal moves, and so on recursively. You can re-create all the >book positions in time roughly proportional to book size times number of legal >moves in a position times whatever it costs to access a single position in a >crafty book given the hash. > >The only bit that looks at all hard is generating Fritz-tree format at the end. >Is that format documented? > >I'm not interested in doing this personally since i don't own Fritz. But i did >tree taversals for various reasons with my own hashed book format, until i >decided the book was more trouble than it was worth. I sometimes process from 70 to 100 million *distinct* board positions when I create an opening book. The method you post would be "tedious" and since we already have the positions themselves (we generated the book from PGN or EPD or *something*, didn't we?) then it seems a lot more sensible to simply write the positions that we use out as we go. If we wanted to do that sort of thing, that is. I would like to see an interface to a program like Chess Assistant. It could tell us all sorts of statistics about the current position and subsequent positions. It is also aware of C.A.P. data, won/loss/draw statistics and the ELO of players in its database who made the choices etc. It also has temporal data, which could show which moves are currently thought to be best. It does little good if Morphy [strong but relatively 'ancient' in modern chess history] made the move, but it has been repudiated by later analysis. If Kasparov and Leko are choosing it, it seems like a very good choice. Seems like it would be an unbeatable combination for a chess opening book.
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.