Author: Mike Byrne
Date: 04:50:47 03/06/03
Go up one level in this thread
On March 06, 2003 at 07:36:33, Uri Blass wrote: >On March 06, 2003 at 07:22:50, Mike Byrne wrote: > >>On March 05, 2003 at 10:12:39, Uri Blass wrote: >> >>>On March 05, 2003 at 10:04:14, Djordje Vidanovic wrote: >>> >>>>Thank you for the information regarding the new Crafty version. However, in my >>>>own testing (more than one thousand games at quick controls), Ruffian 1.X gets >>>>about 70-75% against Crafty 19.X. The older Crafties perform a little better, >>>>especially the 17.13 version. >>>> >>>>The ELO difference ranges between 80-120 pts in favour of Ruffian. >>>> >>>>-- Djordje >>> >>>I guess that the problem is to choose the right X and the right modifications. >>> >>>Modified 19.x should not prerform worse than 17.13 because people can delete all >>>the counter productive changes from 17.13 to 19.03 >>> >>>Uri >> >>That's an over simplification -- one (applicable to most people) cannot simply >>take 19.xx code and roll it back to 17.xx code -- there has been way too many >>changes from 17.xx to 19.xx. If you had compared the code, you would have never >>said that. > >I did not compare the code but the point is that a later version should not >perform worse if it is modified correctly. Should not -- but that's a big if --- it's simple to modifify --- it is much tougher to modify it and have it play better -- most modifications make it play _worse_ -- _some_ might make it play better. The code that is released is well tested before release > > > IMO, 19.xx is better than 17.xx - no question. I would agree that >>the issue of 18.xx is better than 17.xx is up for debate. > >The question is still open > >http://www.digichess.gr/infiniteloop/ratings/IL3b_rate.txt > >Crafty17.9 is leading >Crafty19.1 has not a good result. > >http://www.digichess.gr/infiniteloop/ratings/rapid_rating_il2r.txt > >Crafty17.9 is leading >19.1 is not there but I do not expect 19.1 to be first place after >reading the following link when Crafty19.1 got only 2.5 out of 9. I would prefer to see them use 19.3 but it's not my call. 19.1 was a major overhaul esp. with extensions which seemed to be an area that needed workover 18.xx and it was the first release. 19.3 is more fine tuned with the new code. > >http://f11.parsimony.net/forum16635/messages/44045.htm > >Number of games is not enough to be sure and it is only fast time >control so I do not claim that 17.xx is better than 19.xx but only >that we have not enough evidence to say that 19.xx is better. Agree, more games are needed. I will start a new LTC (40moves/2 hours) 48 round double RR tourney soon - I will include Chessbase native v17.9 and then we will have better sense. > >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.