Author: Andrew Williams
Date: 14:36:35 03/07/04
Go up one level in this thread
On March 07, 2004 at 17:31:26, Uri Blass wrote: >On March 07, 2004 at 17:22:24, Andrew Williams wrote: > >>On March 07, 2004 at 17:05:53, Uri Blass wrote: >> >>>On March 07, 2004 at 16:46:28, Andrew Williams wrote: >>> >>>>On March 07, 2004 at 10:24:47, Jon Dart wrote: >>>> >>>>>I ran a 40 move/40 minute match between Postmodernist 1.010a and Arasan 7.4. >>>>>Hardware was an XP2500+, 64 MB Hash for Arasan, 44MB for PM. Arasan lost this >>>>>match, +3, -10, =7. Games are available at >>>>> >>>>>http://www.arasanchess.org/pm.pgn >>>>> >>>>>I haven't analyzed the games in detail. My general impression was that PM gained >>>>>greater mobility in several games and was able to win "positionally". For >>>>>example, in game 4, PM got doubled rooks and had the better position for most of >>>>>the game, but really won in the endgame where it was able to out-manouver Arasan >>>>>and win material. >>>>> >>>>>Game 18 looks like an opening book problem (18 .. Ne8 is the standard move, but >>>>>isn't in Arasan's small book). >>>>> >>>>>--Jon >>>> >>>>Hi Jon, >>>> >>>>Thanks for this report. Speaking of book issues, I have been experimenting with >>>>an improved "preferred" book for PostModernist. I have essentially got rid of >>>>the lines which I had been adding because I think PM has done Ok with them at >>>>ICC. Instead I have been inserting "bog-standard" lines. All this was done on >>>>Peter Berger's advice. The results have been *ASTONISHING*. For example, the >>>>current match score is: >>>> >>>>improvedpreferred 108.5 - 61.5 standardpreferred >>>> >>>>The previous match score was 158 - 119 standardpreferred before I stopped it. >>>> >>>>I would NEVER have imagined a result like it, and it has been VERY easy to do >>>>this. >>>> >>>>Andrew >>> >>>I do not understand what was so easy in changing the book. >>> >>>What exactly did you change? >>My preferred book is a series of lines like this: >> >>e4!! c5 Nf3!! d6 d4!! cxd4 Nxd4!! Nf6 Nc3!! a6 Be3!! >> >>I used to have various odd lines in there - mostly as a result of looking >>through PM's performance on ICC and deciding based on that which openings it was >>"comfortable" with and which ones it wasn't well suited to. Peter Berger was >>kind enough to point out that the book was actually worse than useless. >>>What is bog-standard lines? >> >>"Bog-standard" means normal. Essentially, I've gone through a large collection >>of games (Dann's huge one, using just games where both players are > 2550 or >>something like that) and (generally) selected the moves which are most often >>played in openings. >> >> >>Andrew > >I do not see it as easy to generate bog-standard book. >I first need to write a program to read a file of pgn games and find for every >position the most often played line and I also think that the most often played >lines may be bad because it is possible that a line was played often in >1990-2000 and was refuted in 2001 but you still have a lot of games from >1990-2000 when the line was played. > I'd recommend Scid. This is what I used to help me to create my preferred book. It also gives you the "average" year in which a move has been played. I'd guess that Chessbase is better, but Scid runs on Linux. You can get Scid from here: http://scid.sourceforge.net/ >Statistics about wins may be also misleading because the refutation may be based >on a single game when before the refutation you can get many wins for the wrong >side. > True. This is why I said, "generally". In truth, I don't really have the skill to do this properly - all I am sure about is that the new version is better than the old. Andrew
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.