Author: Uri Blass
Date: 04:11:22 11/05/04
Go up one level in this thread
On November 04, 2004 at 21:59:06, Mike Byrne wrote: >I have commenced the effort to put forth my interpretation of the best of Crafty >for the year 2004. > >This modified Crafty will contain my interpretation of the best of Crafty 19.xx >for the year 2004. It will have all the latest modified SE PB code, 4 new GM >personalities ( 28 GM personalities in total) with a personalized book for each >GM created that will play the openings that GM plays. So far, I have decided >to add Kramnik and Leko -- I'm undecided about the other 2 GM personalities. > >Please let me know which other GM personalities you would like to see. They can >be past or present, it helps if they were/are near the top. I would prefer not >to add GMs under the age of 25 - Peter Leko just made it for my age requirement. > >I am also considering Vishy Anand but feel free to submitt anyone you would >like to see. > >The code will not be from any one Crafty. I had narrowed the base down to >between 19.12 and 19.8 -- and 19.8 won a 200 game tournament 61-40-99 (55%). I >believe these earlier versions are stronger than the current Crafty 19.17 (just >my opinion ). As a sanity check , I am now running this modified Crafty 19.8 >(named Crafty Classic 2004 SE PB) against Crafty 19.17 . It is not only >modified with the SE/PB code - I have also taken snippets of code from 19.12 and >19.17 that I believe "help" Crafty. So it is in essence a modified Crafty with >SE/PB code but in standard personality it will be unlike any Crafty that exist >today. Hence, the name "Crafty Classic 2004 SE PB" Whether it is better or not >than Crafty 19.17 , I will let you decide. Being better than Crafty19.17 is an easy job and it seems clear that almost every version of Crafty is better than 19.17. There is also some evidence that 19.15 is better than 19.17 and 19.13 is better than 19.17 http://www.husvankempen.de/nunn/ratingaegt2310.txt http://f27.parsimony.net/forum67828/messages/205.htm http://f27.parsimony.net/forum67828/messages/187.htm You can add the fact that the ssdf also found that 19.17 is probably weaker that previous version that they tesed. You can also add frank quinsky results when Crafty19.17 came last place with 11/34 when Crafty19.15 had 15.5/34 against similiar opponents in round 3 and Crafty19.13 had 15/34 in round 2 and round 1. The opponents were changed but there was no big difference in the strength of the opponents. http://www.playwitharena.com/ (see ATL league) I remember more tests that I was too lazy to search links for them. One of these tests may not be significant but when I combine all the data that different people found that 19.17 is weaker then it seems to be very significant. I remember no single test when 19.17 performed better than 19.15 or older version so there is probably some big bug in 19.17(in case that a version is only slightly weaker I can expect to see at least one tournament when it performs better out of many tournaments). Uri
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.