Author: Andres Valverde
Date: 14:45:03 09/27/05
Go up one level in this thread
On September 27, 2005 at 00:18:15, David Mitchell wrote: >Unfortunately, I'm thick as a brick on book files, but I gather from the above >that Andres will have a different EPD line for EVERY SINGLE POSITION in his >opening book?? "Holy Crap!" I infer the advantage is it detects all transposed >positions, and the disadvantage is it makes for a HUGE file. > >Andres, are you breaking this up into multiple files? Well I try to explain my idea : 1) Read PGN collections till ply "x" (e.g: 30) 2) After every move write a text file with the position EPD string and the following move, also the result of the game is stored: rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq -;d2d4;w; 3) The file created has lot of repeated lines, so I sort it and creates a new file, saving the lines repeated "y" times (the greater "y" is the shorter the file). This new file is the book itself composed by the lines I posted. rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - pm d2d4 ce 56 (note the changes propposed by Dann. 56 is the score (in %) of the side to play) 4) The engine creates the EPD string for the current position and search the text file. It can find several lines, gather the scores, and select one of the predicted moves (randomly, the best score etc.) > >Dann (and y'all), would you comment a bit on this scheme? I'm guessing that the >text lists of the moves is the most common layout for a book file? Is this also >a popular layout for an opening book, or is it only suitable for small books, >etc.? I have tried a book with 5.000.000 lines (unique positions) and it is a matter of 0,5 secs to find the latest line in the text file. One thing I like is that the book is readable as text file and updatable manually. Thanks for your answer > >Thanks, >Dave
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.