Author: Robert Hyatt
Date: 07:10:54 10/01/04
Go up one level in this thread
On October 01, 2004 at 05:40:19, Tony Hedlund wrote: >On October 01, 2004 at 00:43:45, Mike Byrne wrote: > >>On September 30, 2004 at 23:52:18, Robert Hyatt wrote: >> >>>On September 30, 2004 at 23:03:19, Peter Skinner wrote: >>> >>>>On September 30, 2004 at 19:59:08, Robert Hyatt wrote: >>>> >>>>>On September 30, 2004 at 18:30:40, Peter Skinner wrote: >>>>> >>>>>>On September 30, 2004 at 18:02:21, Robert Hyatt wrote: >>>>>> >>>>>>> >>>>>>>>Again Crafty lost an Nimzo-Indian with white. When will it change to 1.e4? >>>>>>> >>>>>>>It won't. >>>>>>> >>>>>>>:) >>>>>> >>>>>>Do you honestly feel Crafty plays better vs computers with 1.d4? >>>>>> >>>>>>I find in online games that Crafty does very well with 1.e4 or 1.Nf3 against >>>>>>computers. >>>>>> >>>>>>Peter >>>>> >>>>> >>>>>The problem is professional book lines in 1. e4 openings... >>>> >>>>I see your point. >>>> >>>>Would it not be better to supply an opening bookc.bin that would refute many of >>>>the lines, or at best come out of the opening even? >>>> >>>>Playing 1.d4 is simply going to get Crafty in trouble in quite a few of the >>>>games. You can only play it so many times before learning hurts you instead of >>>>helping you. >>>> >>>>Peter >>> >>> >>>The thing is, they play X games with one computer, so learning ought to fix this >>>up unless the Nimzo lines are so well plotted that everything it tries fails. >>> >>>Making a special SSDF opening book is not something I have time to do. Nor >>>anyone else I know of. IE Peter has some stuff from the WCCC, but we'd not want >>>to reveal that via SSDF testing as we will likely play in the next WCCC event >>>and will need a unknown book for it... >> >>The book for SSDF appears to be more or less a generic plain jane book based >>on top GM games. My own testing with new clean books like this -- it will take >>at lest about 20 -25 games to really get some "good" learning action. So it >>may take a little bit of time. > >The lrn-files is still empty. Is this ok? Will it take more games for Crafty to >start learning? Or is the learning stored somewhere else? This is bad news. The file "book.lrn" is an ASCII file that contains all the "learned" stuff that has been added to book.bin. If book.lrn is empty, learning is turned off. Can you email me a log.nnn file from a complete game? I can look and find out what is up. IE learn=0 will turn it off. Or being unable to write to book.bin will also disable learning... The log file will point the problem out... > >>I like the fact that Tony is rotating through >>multiple opponents as opposed to playing 20 against Junior than 20 versus >>Ruffian etc. > >I'm not playing multiple opponents. It will be 40 games against Ruffian 1.0.1 >and 40 against Deep Junior 8. Hopefully other testers also play with Crafty. > >Tony > >>IMO, there is no doubt that good hand tuned "secret" book for small swiss 14 >>round tournaments is vital to surivive. They all appear to be doing it. >> >>One area that has seen a noticable chnage over the last 10 years is "opening >>book preparation" There is so much more information available to everyone and >>there are good tools to use the data -- all one has to do is to committ the time >>and the energy to do it. The top professional programs are clearly doing >>that. Why not , they are the professionals and a great result in one of these >>reknowned toutnaments will sensales their way. So what if the tournamnet is >>only 14 games - it will be the headlines to the winner that generates sales. >>The book usually plays a pivotal role between the top finishers - since the >>programs are so close to each other anyway.
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.