Author: Robert Hyatt
Date: 08:48:46 06/28/00
Go up one level in this thread
On June 28, 2000 at 05:26:35, Enrique Irazoqui wrote: >On June 27, 2000 at 21:38:36, Robert Hyatt wrote: > >>On June 27, 2000 at 19:39:42, Robin Smith wrote: >> >>>On June 27, 2000 at 17:41:33, Robert Hyatt wrote: >>> >>>>On June 27, 2000 at 14:44:10, Robin Smith wrote: >>>> >>>>>On June 27, 2000 at 08:43:49, Enrique Irazoqui wrote: >>>>> >>>>>>Thorsten posted an interesting position in >>>>>>http://www.oxford-softworks.com/cgi-bin/forums/mschess/558.html: >>>>>> >>>>>>[D]2R5/4k1pp/p3p3/4P1p1/p3N3/q1P4P/2P1P1P1/1K6 w - -; bm e4d6; >>>>>> >>>>>>Thorsten's comments: >>>>>> >>>>>>"Here how CSTal2.03 plays on my k6-3/400: >>>>>> >>>>>>1. d9 8" >>>>>>2. d12 105" >>>>>>3. d14 475" >1.17 >>>>>>4. d14 615" +4.24 e4d6 g7g6 e2e3 h7h6 e3e4 g5g4 h3g4 h6h5 g4h5 g6h5 g2g3 >>>>>> >>>>>>1.After 8 seconds for the first time the key-move , than changed into another. >>>>>> >>>>>>2.After 105 seconds it has the Nd6 again in depth 12. >>>>>> >>>>>>3.After 475 seconds it gets a fail high to >1.17. >>>>>> >>>>>>4.After 615 it knows HOW high and says +4.24 with the main-line shown above." >>>>>> >>>>>>Aside from CST 2.03, only Genius 6.5 (and older versions, I guess) finds the >>>>>>right line in less than 5 minutes and fails high. Shredder 4 and SOS pick Nd6, >>>>>>but with a wrong line and a 0.00 evaluation. Tiger, Fritz, Junior, Hiarcs, >>>>>>Nimzo, Rebel, AnMon, Crafty, CM6K, don't see that the black queen is trapped and >>>>>>white can zugzwang black. >>>>>> >>>>>>Enrique >>>>> >>>>>Very interesting position!! The pieces on both sides are in zugswang, so the >>>>>pawns decide everything. After 1.Nd6 g6 2.e3 h6 3.e4 I think 3. ... h5 instead >>>>>of 3. ... g4? is a better defensive try, but afterwards 4.g3! g4 5.h4! (5.hg? hg >>>>>=) g5 6.hg h4 7.g6 hg 8.g7 g2 9.g8=N+! and mate in two. So programs that can't >>>>>see underpromotions (Junior5) won't ever solve this one. And the underpromotion >>>>>occuring on ply 17 in a position where black threatens mate in one doesn't make >>>>>it easy for ANY program. Congratulations to CSTal! >>>>> >>>>>Robin Smith >>>> >>>> >>>>Do you really think it understood this game? Or do you think it was just >>>>"lucky"??? >>> >>>Failing high with a score > 4 pawns sure sounds like it "understood" to me. >>> >>>Robin Smith >> >> >>CST has been known to fail high to over +5, and four moves later be at -9, >>also... > >But in this case it shows the right line too, like Genius. That's no luck. > >Enrique I understand... but _here_ is the real question: In how many similar positions will it have the same eval, and it be wrong? I've seen it play brilliantly in one game, and then play like a patzer in the next four games. It is nice to find positions where a program _really_ seems to get the right idea. But then reality sets in, as you find similar (but not similar enough) positions where the program comes to the same conclusion as in the brilliant game, but it is dead wrong. This happens way too frequently with computer chess programs, unfortunately, mine included. It will play a brilliant endgame against a GM, then come back and play a completely insane endgame that no 2000 player would even consider. I don't like to brand programs as "brilliant" until they handle things most of the time, not just some of the time...
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.