Author: Uri Blass
Date: 06:29:06 11/30/02
Go up one level in this thread
On November 30, 2002 at 08:58:25, Mike S. wrote: >On November 29, 2002 at 20:21:12, John Merlino wrote: > >>(...) >>>>[D]8/2q1p1k1/3pP1p1/2pPNpNp/2P2B2/1p1B2K1/1P1R4/r7 w - - 0 42 >>>>CM9000, on a humble P3-733, finds 42.Nxg6! in 21 seconds. How do other engines >>>>fare? >>> >>>Analysis by Nimzo 8 (P3/700, 64 MB hash): >>> >>>1.Nd7 Qa7 2.Rf2 Qa2 3.Kh2 >>> ± (0.87) Depth: 1/11 00:00:00 58kN >>> ± (0.76) Depth: 5/17 00:00:00 58kN >>>1.Nc6 Ra8 2.Nf7 Re8 3.Rf2 Kh7 4.Bg5 h4+ >>> ± (0.77) Depth: 5/17 00:00:00 58kN >>> +- (1.62) Depth: 10/25 00:00:07 2694kN >>>1.Ngf7 Kh7 2.Nxg6 Rg1+ 3.Kh2 Rxg6 4.Bxf5 Kg7 5.Re2 Rg4 6.Bxg4 >>> +- (1.63) Depth: 10/25 00:00:13 5667kN >>> +- (2.34) Depth: 11/26 00:00:43 19670kN >>>1.Nxg6 Rg1+ 2.Kh2 Rg4 3.Rf2 Kxg6 4.Bxf5+ Kg7 5.Bxg4 hxg4 6.Rg2 >>> +- (2.35) Depth: 11/26 00:00:46 21624kN >>> +- (3.21) Depth: 13/29 00:01:23 38980kN >>> >>>(...) >>>After 1.Nxg6 Kxg6? 2.Kh2 (as you mentioned), Black has no defense against the >>>threat 3.Rg2 and 4.Nf7+ Kf6 5.Bg5+ Kg7 6.Bxe7+ Kh7 7.Bxf5# > >>Thanks for that analysis. I wondered if Nxg6 could be played later or if it was >>necessary immediately. But, according to your output above, doesn't Nimzo decide >>on Nxg6 after 46 seconds? I would be curious to see what its output is after a >>few more minutes, since you appeared to stop the analysis after less than a >>minute and a half. Does Nimzo eventually switch to Ngf7 like Yace does? > >Yesterday I did wait a bit longer, but Nimzo didn't come up with a new >mainvariant after ~3 minutes. To repeat the test, I removed the learn file and >let it run for 15 minutes (while surfing the web): > >Analysis by Nimzo 8: > >1.Nd7 Qa7 2.Rf2 Qa2 3.Kh2 > ± (0.87) Depth: 1/11 00:00:00 > ± (0.76) Depth: 5/17 00:00:00 >1.Nc6 Ra8 2.Nf7 Re8 3.Rf2 Qb7 4.Bh6+ Kg8 5.Bg5 Kh7 6.Kh2 > ± (0.77) Depth: 5/17 00:00:00 > +- (1.63) Depth: 10/25 00:00:08 2363kN >1.Ngf7 Kh7 2.Nxg6 Rg1+ 3.Kh2 Rxg6 4.Bxf5 Kg7 5.Rd3 Qb6 6.Bxg6 > +- (1.64) Depth: 10/25 00:00:20 7487kN > +- (2.50) Depth: 11/26 00:01:04 27245kN >1.Nxg6 Rg1+ 2.Kh2 Rg4 3.Rf2 Kxg6 4.Bxf5+ Kg7 5.Bxg4 hxg4 6.Rg2 > +- (2.51) Depth: 11/26 00:01:06 28244kN > +- (3.37) Depth: 15/32 00:09:22 275483kN > >I have now idea why some evaluations today are different from yesterday's, but >anyway, Nimzo likes 1.Nxg6 more. Now, I switched to multi-pv mode during the >analysis above, and let it calculate for some more minutes: > >Analysis by Nimzo 8 (during depth 13): > >1. +- (3.21): 1.Nxg6 Rg1+ 2.Kh2 Rg4 3.Rf2 Kxg6 4.Bxf5+ Kg7 5.Bxg4 hxg4 6.Rg2 >2. +- (1.94): 1.Ngf7 Qa5 2.Nc6 Rg1+ 3.Kf3 Qa1 4.Nxe7 Rg4 5.Rh2 Rxf4+ 6.Kxf4 > >It seems that 1...Qa5 is more annoying for White after 1.Ngf7 here (than after >1.Nxg6 Kxg6 2.Kh2 as you mentioned it), with counter-threats for Black as his >queen can enter White's position to support the rook. > >So it seems it's not just a simple transposition (as I thought yesterday), and >these two first moves are not exchangeable... but probably both winning. Yace's move was Ngf3 and not Ngf7 I believe that white has a lot of good moves(for example Nc6 is suggested by movei with verifecation search +2.25/12 before failing high on Nxg6(+2.26 after 27315139 nodes slightly better than Nimzo8 that needs 28244 knodes) I prefer to use verification search for long time control. Movei fails high again on Nxg6(29146651 nodes) with score of 2.55 but solving the fail high problem seem to take a big number of nodes(111961781 to get 2.74 pawns that is also the final score for depth 12). The score is going up to +2.89 at depth 13(129583889 nodes) and Movei finishes iteration 13 after 133863837 nodes. The score continues to go up at depth 14 to +2.99(171208764 nodes) 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.