Author: Robert Hyatt
Date: 05:03:10 05/19/98
Go up one level in this thread
On May 18, 1998 at 22:28:22, Dave Gomboc wrote:
>On May 18, 1998 at 21:40:31, Robert Hyatt wrote:
>
>>On May 18, 1998 at 21:29:42, Dave Gomboc wrote:
>>
>>>I downloaded this yesterday, and am trying to import the PGN into CDB.
>>>I have set the compression to occur after move zero (e.g. don't look for
>>>any transpositions) in the hopes that most of the book will be imported
>>>before my (64Mb RAM) machine starts swapping like crazy. So far I am at
>>>85000 games, no swapping, but I don't know how long it will hold up.
>>>Bob, do you know how many games are in the "wall" book? If I start
>>>swapping, but I'm only 20000 games away or so from finishing, then I can
>>>just let my computer get beat up for a couple of days. But if I am
>>>200000 games away, obviously anything other than purchasing more ram
>>>would be futile!
>>
>>let me put it like this... you are about 25% done. :) IE I think
>>there
>>are 330,000 games in the mother.. :)
>
>Well, it's through 142000 games or so now, no thrashing yet. I'm just
>going to cross my fingers and hope. <grin>
>
>Of course, even if I do get them all in, not catching transpositions is
>going to be a serious bummer. Hopefully I can edit the compression
>level after the import is finished (maybe move it a few moves up at a
>time) and the software will reexamine the book that has been built
>without going into swapland. I think I'm being overly optimistic
>though, first the import has to finish. :)
>
>Bob, do you have any tool to take the learned information (presumably
>keyed by hash value and learning score) and provide a line of play from
>the start position that reaches it?
>
>
way ahead of you... look at "book.lrn"... it is already a series of
moves
that ends with a comment in { ... } with the learned scores...
Incidentally, here are a couple of games in the "wall" book that need
>some correction:
>
>Game #128155 (or so :)
>White is "Wahls, Matthias"
>Black is "Bancod, Ronald"
>Result is "1/2-1/2" in the headers but 1-0 in the game score.
>Search for the string "45. Ke2 Ra2+ 46. Ke1 Na3 47.Re5+ Kf6 48. Rf5+ Kg6
>49. Rxf4 1-0" to find it.
>
>Game #139452 (or so :)
>White is "Curdo, John"
>Black is "Kavalek, Lubomir"
>Result is "1/2-1/2" in the headers but 0-1 in the game score.
>Search for the string "16. Nxe4 c5 17. Nxf6+ Bxf6 18. Bxf6 Qxf6 19. Qh5
>g6 20. Qxc5 Qc3 21. Qd6 0-1" to find it.
>
>There was an error around game 32000, but I missed it.
>
>Note to author of CDB: Mismatched result was the cause of the 'error' I
>emailed you about. CDB detected the above errors, and also any
>additional ones that I might come across in the remainder of the book
>that might be posted in the next day or so.
>
>Dave Gomboc
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.