Author: Jorge Pichard
Date: 13:31:35 02/26/04
Go up one level in this thread
On February 26, 2004 at 03:40:52, Mike S. wrote: >On February 25, 2004 at 17:52:39, Jorge Pichard wrote: > >>I honestly think that the SSDF rating list is a ridiculous comparison to human >>rating, with Shredder 7.04 UCI rated over 2800 on a mere AMD Athlon 1.2 GHz. If >>the initial purpose of the SSDF was to assign computer programs with rating >>similar to human rating, then the target is way too far off. > >Actually (AFAIK) it was never the purpose to give a rating extimation relative >to human FIDE elo, but simply to rate chess programs relative to each other. Why >do you call the list "ridiculous" based on an assumption you're not even sure >about?? > >>The SSDF either >>lower it rating by 100 points or it is time to keep the same rating as it is but >>start testing with an AMD XP 3200. > >The list has been adjusted in the past (linear), last time by -100 points. But >that has also the result that strong chess computers at the end of the list are >*underestimated* compared to human elo. > >To put it easy, if you reduce the top's ratings by level adjustments again and >again, programs and comps at the end of the list are more and more >underestimated (significantly! as many of these had better performances against >humans, i.e. at Aegon tournaments). > >(I mean the complete list; see the text version on the SSDF homepage.) > >So whats a "true" rating level? > >It seems the list is spread across a too large rating range. This is a problem >of the rating system IMO, and has been discussed in the past, IIRC with no clear >conclusion. > >OTOH, why do you doubt a 2808 rating on the top, in comparison to human GMs? > >Remeber this? > >VI MAGISTRAL DE LA REPUBLICA ARGENTINA Vicente Lopez 2003-07-24 >En CM Salud, D. F. Sarmiento 1755, Florida, Tel 4796-1900 > >No. Jugador 1 2 3 4 5 6 7 8 9 0 1 Pts S.B. Exp >01) CUBAS, Jose * 0 1 1 = 0 = 1 = 1 0 5,5 . 4.23 >02) RODRÍGUEZ, Andres 1 * = 1 0 = = 1 = 1 = 6,5 . 5.04 >03) ROSELLI, Bernardo 0 = * 0 = = 0 1 = = 0 3,5 . 4.50 >04) DE DOVITIIS, Alej 0 0 1 * 0 0 0 = 0 0 0 1,5 . 3.96 >05) PANNO, Oscar = 1 = 1 * = 0 1 = = 0 5,5 . 4.95 >06) MARQUES, Vinicius 1 = = 1 = * = = = = = 6 . 3.78 >07) VALERGA, Diego = = 1 1 1 = * 1 = = 0 6,5 . 5.22 >08) SCARELLA, Enrique 0 0 0 = 0 = 0 * 0 = 0 1,5 . 3.69 >09) SLIPAK, Sergio = = = 1 = = = 1 * 1 = 6,5 . 5.22 >10) LAFUENTE, Pablo 0 0 = 1 = = = = 0 * 0 3,5 . 4.59 >11) SHREDDER 7.0 1 = 1 1 1 = 1 1 = 1 * 8,5 . . <-- >ELO promedio: 2442,7. Categ. FIDE: VIII. Norma GM: 7 pts., MI: 5 pts. > >Shredder 7.0's elo performance was 2752, playing on an Athlon 1,6 GHz with >aggressive Setting. In the SSDF list, Shredder 7.0 is listed 2771 (+27 -25) on >1,2 GHz. In both of these tourney the players were not too familiar with Shredder, nor Tiger, but if you take the same goup of players nowadays they will reap apart Shredder 7.0 and Tiger 15 using the same CPU speed in Shredder 7 or 8 and for Tiger even by using three the the CPU speed with the same group of players and even with Tiger 15. >(IIRC, J.Sonas has explained that Shredder's opposition in that tournament was >"too weak" in terms of elos actually...) > >Or take a look at Chris Carsons Man vs. Machine perfomances list: > >http://home.interact.se/~w100107/Man%20vs%20Machine%20%20%20%2019%20Nov.htm > >In a previous Argentinia tournament, Tiger 14 on 866 MHz performed 2788. In the >SSDF list, Tiger 14 is rated 2717 on 1,2 GHz. > >Regards, >M.Scheidl
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.