Author: Odd Gunnar Malin
Date: 05:45:41 11/28/03
Go up one level in this thread
On November 28, 2003 at 08:26:12, Odd Gunnar Malin wrote: >On November 28, 2003 at 07:44:36, Georg v. Zimmermann wrote: > >>Disclaimer: the following is not legally binding. It is an offer to negotiate. >>Nevertheless, it is a serious proposal. >> >>It has been repeatetly stated that a 'clone' of crafty can not be 100 or more >>elo stronger, by many people here. >> >>I do not believe so. How about a bet ? >> >>I would state that I can improve Crafty by 100 points or more within 4 month. >> > >For better comp-comp play I would have: >- Let the code be left alone or maybe removed some stuff that are unneeded for >pure play. Maybe added some commands to discard some evaluation code. >- Then made a automatic tuningtool. >- Collected position with eval/bestmove from some of the best comp-player >(Shredder/Fritz) for small ply depth (1 or 2). >- Let it run for tuning. >- Run matches against comps to tune the book, start with a pretuned book like >that Hyatt use on ICC now. > >The first three points shouldn't take more than one day each. The two last >points get the rest of the four months. > >I guess Crafty are tuned for human-comp play so the overall strength could >(probably) go down. > >Odd Gunnar Hmm... When looking at SSDF's ratinglist the difference wasn't so big that I though so 100 points was a bit too much. Maybe I would expect somewhere between 25 and 50 points from this in a comp-ratinglist. When you get near the top each rating-point gets harder to get. Odd Gunnar
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.