Author: Dann Corbit
Date: 23:54:29 07/18/03
Go up one level in this thread
On July 19, 2003 at 01:49:58, Kurt Utzinger wrote: >On July 18, 2003 at 21:37:41, Dann Corbit wrote: > >>On July 18, 2003 at 13:19:47, Kurt Utzinger wrote: >> >>>English >>>If we compare the following, nobody would believe that the same engines had >>>played the same tournament with only a small difference: the books >>> >>>Français >>>Si on compare les résultats suivants, on ne peut guère croire que les memes >>>modules ont joués le meme tournoi avec une petite différence: les livrets >>>d'ouvertures: >>> >>>Deutsch >>>Betrachtet man den nachstehenden Vergleich, ist kaum zu glauben, dass dieselben >>>Programme dasselbe Turnier gespielt haben mit einem kleinen Unterschied: >>>verschiedene Bücher: >>> >>>RK 2003 tournament 120'/60+60'/30+30' >>>Results with draw book remis.ctg for all engines >>>after 84 games [each engine 21 games] >>> >>> Program Score % Av.Op. Elo + - >>> >>> 1 TK 3.23 SKR : 15.5/ 21 73.8 2476 2656 129 141 >> >>2656 - 141 = 2515 >> >>> 2 Fritz 8 : 11.5/ 21 54.8 2495 2528 158 118 >>> 3 Shredder 7.04 : 11.0/ 21 52.4 2497 2513 162 109 >>> 4 Hiarcs 8 : 11.0/ 21 52.4 2497 2513 162 96 >>> 5 Chess Tiger 15 : 10.5/ 21 50.0 2499 2499 122 122 >>> 6 Ruffian 1.0.5 : 9.0/ 21 42.9 2505 2455 97 153 >>> 7 Deep Sjeng 1.5 : 8.0/ 21 38.1 2509 2425 100 145 >>> 8 Junior 8 : 7.5/ 21 35.7 2512 2410 148 142 >> >>2410 + 148 = 2558 >> >>IOW, you cannot discern between the top and the bottom here. >> >>> >>>RK 2003 tournament 120'/60+60'/30+30' >>>Results with own books for all engines >>>after 84 games [each engine 21 games] >>> >>> >>> Program Score % Av.Op. Elo + - >>>Draws >>> >>> 1 Shredder 7.04 : 13.0/ 21 61.9 2490 2574 145 100 >> >>2574 - 100 = 2474 >> >>> 2 Junior 8 : 13.0/ 21 61.9 2490 2574 145 117 >>> 3 Fritz 8 : 11.0/ 21 52.4 2498 2515 162 109 >>> 4 Ruffian 1.0.5 : 11.0/ 21 52.4 2498 2515 162 96 >>> 5 Deep Sjeng 1.5 : 10.5/ 21 50.0 2500 2500 133 133 >>> 6 Hiarcs 8 : 9.5/ 21 45.2 2504 2471 89 158 >>> 7 The King 3.23 skr : 8.5/ 21 40.5 2508 2441 107 149 >>> 8 Chess Tiger 15.0 : 7.5/ 21 35.7 2513 2411 91 142 >> >>2411 + 142 = 2553 >> >>IOW, you cannot discern between the top and the bottom here. >> >>>Further details at: >>>Détails additionnels chez: >>>Weitere Details unter: >>> >>>http://www.utzingerkurt.com (section RK 2003) >> >>If you look at the error bars, I will think that you cannot draw any >>conclusions. >> >>It will take 210 games (not 21) to find out which setup really is stonger, and >>even at that you will not be able to discern between the closest entries. >> >>IMO-YMMV. >> >>"If ten men can build a house in one day, then 864,000 can build it in one >>second." > > Hi Dan > I fully agree with you. With my example I only wanted to have some > discussion and to demonstrate that such things can happen. I do not > think that the different books are responsible. If we look at the > above we just see the result of two "tournaments" and not more. People > often forget this and try to draw conclusions that are not at all > justified. It is (of course) possible that the conclusions people jump to are correct. It is just that in this case, it is not fully justified by the evidence.
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.