Author: Uri Blass
Date: 11:04:04 04/25/03
Go up one level in this thread
On April 25, 2003 at 12:46:38, Jorge Pichard wrote: >On April 25, 2003 at 12:24:29, Uri Blass wrote: > >>On April 25, 2003 at 11:53:59, Jorge Pichard wrote: >> >>>On April 25, 2003 at 11:37:10, Uri Blass wrote: >>> >>>>On April 25, 2003 at 11:08:46, Jorge Pichard wrote: >>>> >>>>>If doubling the speed increase the strength of a program by 40 points, then the >>>>>estimated strength of Fritz 5.32 should be close to 2666 on an Athlon 1.2 Ghz. >>>>>Therefore, the Opening alone can increase a program strength from 50 to 85 >>>>>points. >>>>> >>>>> >>>>> >>>>> ------ --- --- ----- --- ---- >>>>> 1 Fritz 5.0 PB30% 67MB P200 MMX 2566 25 -24 906 68% 2431 >>>>> 2 Junior 5.0 64MB P200 MMX 2556 33 -31 502 67% 2435 >>>>> 3 Nimzo 98 56MB P200 MMX 2526 25 -24 812 61% 2450 >>>>> 4 Rebel 9.0 45MB P200 MMX 2524 25 -25 802 63% 2432 >>>> >>>> >>>>I do not know how did you get the estimate that doubling the speed can >>>>increase the strength of the program only by 40 points? >>>> >>>>I think that it is more than it and the difference is bigger at the fast time >>>>control that you played your games. >>>> >>>>Uri >>> >>>Are you one of those that support that doubling the speed of a processor could >>>yield an increase of strength of more than 40 points, and think that the >>>difference should be greater when matching a computer program versus another >>>program instead of being directly proportional. Now if the match is between >>>Human versus computer than I totally agree with you, that the difference should >>>be bigger benefiting the computer, but NOT Comp VS Comp. >> >>The opposite. >>The common conjecture is that the difference in comp-comp is bigger than the >>difference in comp-human games. >> >> Finally, the SSDF >>>estimate is way overrated, since there is noway possible that Shredder 7 could >>>reach 2765 on a mere 1.2 GHz. If that is the case then Shredder 7 or 7.04 should >>>be able to beat any of the top 3 FIDE players 80% of the times on an Athlon XP >>>3000+. >> >>I do not see how do you get it. >>Your rating is based on comp-comp games so the rating against humans is >>irrelevant. >> >>My claim is that programs earn more than 40 elo in comp-comp games if I double >>their speed. > >I still don't think that programs earn more than 40 elo points in comp-comp by >doubling their speed. If that is the case Shredder 7 should reach 2765 + 40+ => >2805+ or greater by using an Athlon XP 2400+. I don't think that is the case. > >Jorge I see no reason not to believe that shredder can get more than 2805 ssdf rating on XP 2400+ Your games are also at faster time control than 120/40 and I believe that there is a diminishing return so you can expect bigger difference than the difference that the ssdf can find. At very fast time control Junior5 drew the nunn1 match against Fritz8.008 10-10 with time advantage of 3:1(3 minutes for Junior5 and 1 minute for Fritz8) By the theory of 40 elo difference for doubling the estimate for the difference between Junior5 and Fritz8 is less than 70 elo. 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.