Author: Enrique Irazoqui
Date: 03:47:23 03/18/98
Go up one level in this thread
On March 18, 1998 at 05:41:06, Ed Schröder wrote: >>Posted by Enrique Irazoqui on March 17, 1998 at 19:45:08: > >>Not such a sophisticated learner, by the way. A simple one, a human-like >>one. A learner that repeats the opening line when winning and avoids it >>when losing, just as we do. In the process, it takes care of double >>games. > >Here you go "a sophisticated learner"................ > >You want us to solve everything with learners???? What you call distortion produced by learners is already solved. See below. >Where is the chess engine (the main reason why people buy chess >programs) in your point of view? In 99.4 % of the games played in comp-comp form. See below. >Well, I will give you a very good and (hopefully) convincing example >that COMP-COMP learners have the potential to cheat on SSDF. > >Before doing so I like to emphasize this topic is about COMPUTER VERSUS >COMPUTER games as played on the Swedish Computer Rating List also called >the SSDF. Make it more general. It applies to all computer-computer games, played by the SSDF or by anyone else. >This topic is not about learning in normal Human-Computer >games. It is too. See the answers posted by Bruce and by Bob. >Rebel supports that and the learner software will be further improved. >This little introduction in order to avoid confusion. > >The example... > >I now know that the Swedish interrupt matches and then later restart the >match. All ok and understandable and nothing wrong about that. > >BUT.................. > >looking at my code with this "new" information I can make advantage of >that information and add special learning -ROTFL- software which will >gain at least 30-40 ELO points on SSDF. > >This is no joke. I checked the Rebel9 book learn software and I can >do it if I wish. This is crazy no? It's crazy because the 30-40 elo >improvement is counted as a gain of playing strength. That's what >the list implies or? > >And now the $64,000 question.... You said it... >Shall I? You will... >It's a cheat no? NO! >It's a cheat because it hides the REAL strength of an chess engine. You lost $ 64,000. You said it in public, we are all witnesses. Nice. :))) >The SSDF list (as I always have understood) is meant to order chess >programs in a list based on playing strength. These days my only >conclusion is that "learners" do have too much influence to succeed >that goal. False. Below. >It's my fear that competing in SSDF will end in fights between >"sophisticated learners" (as you like to name them) instead of >the original goal which is fights between two strong engines. False again. >This whole "learning" (as currently programmed and in relation to >SSDF) is a cheat in itself which I don't want to be a part of >(anymore). Look at the above example. It's crazy, no? Yes, very. Let's see. SSDF posted a total of 5,015 games played between 1991 and 1997. 122 of these games are doubles, or 2.4%. SSDF posted a total of 319 games played between latest generation programs, all with learners: Mchess 7.1, Rebel 9, Nimzo98, Fritz 5, Hiarcs 6, Shredder 2, Genius 5. Of these 319 games, 2 are double. A 0.6%. With only these same programs I played this year 415 games. 2 doubles, or 0.48% Double games are not significant anymore. Learners are working already in a way that allows you to say SSDF is measuring more than ever the value of engines. Additional info: out of those 122 double games, 60 are played by Rebel 6-7-8. Half, or as many as by all the other programs put together. It is more a specific problem of Rebel than learners. So much for the influence of learners in today's real life. In 99.4% of the games, learners did not repeat games, therefore they did not make the difference. Now, Ed, where is the cheat? More important: where are my 64,000 dollars? :))))))))) Enrique >- Ed -
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.