Author: Robert Hyatt
Date: 07:08:16 08/30/01
Go up one level in this thread
On August 30, 2001 at 01:24:25, Slater Wold wrote: >On August 30, 2001 at 00:54:03, Jeremiah Penery wrote: > >>On August 29, 2001 at 23:19:43, Slater Wold wrote: >> >>>On August 29, 2001 at 13:14:41, José Antônio Fabiano Mendes wrote: >>> >>>> Cray Blitz vs Belle [1981 North American Computer Chess Championship] >>>> The (in)famous position Dr. Hyatt referred to the other day: >>>> [D]5r1k/6p1/1n2Q2p/4p3/8/7P/PP4PK/R1B1q3 w >>>> And here Cray Blitz played 28.Qxb6 >>>> Source: http://wondersmith.com/rants/howfar.htm >>> >>>I understand that CB took only 10 seconds to play, and was limited to 1K nps. >>>However, I wanted to post this, to show only 1 thing: >>> >>>position: >>> >>> 8.01 0:00 -0.25 1.Bf4 Qxa1 2.Bxe5 Kh7 3.Qxb6 Qxa2 4.b4 Rf7 5.Qe3 Qb1 >>>(268.921) 312.3 >>> 9.01 0:02 -0.18 1.Bf4 Qxa1 2.Bxe5 Kh7 3.Qxb6 Qxa2 4.b4 Rf7 5.b5 Qd5 6.Qb8 >>>(903.942) 321.2 >>> 9.03 0:04 -0.17++ 1.Bxh6 Qxa1 2.Qg6 Rg8 3.Bf4 Rb8 4.Bxe5 Rb7 5.Qc6 Kh7 >>>6.Qxb7 Qg1+ 7.Kxg1 (1.356.976) 321.8 >>> 9.03 0:04 0.00 1.Bxh6 Qxa1 2.Qg6 gxh6 3.Qxh6+ Kg8 4.Qg6+ Kh8 5.Qh6+ Kg8 >>>(1.533.066) 325.0 >>> 10.01 0:07 0.00 1.Bxh6 Qxa1 2.Qg6 gxh6 3.Qxh6+ Kg8 4.Qg6+ Kh8 5.Qh6+ Kg8 >>>(2.517.038) 332.0 >>> 11.01 0:27 +0.01 1.Bxh6 Qxa1 2.Qe7 Rg8 3.Be3 Rc8 4.Bxb6 Qxb2 5.Bc7 Qxa2 >>>6.Bxe5 Qg8 7.Qh4+ Qh7 8.Qg4 (9.074.087) 331.3 >>> 12.01 1:00 +0.26++ 1.Bxh6 Qxa1 2.Qe7 Rg8 3.Be3 Rc8 4.Bxb6 Qxb2 5.Bc7 Qxa2 >>>6.Bxe5 Qxg2+ 7.Kxg2 Re8 8.Qxe8+ (19.746.518) 327.1 >>> 12.01 1:30 +0.68 1.Bxh6 Qxa1 2.Qe7 Rg8 3.Be3 Qb1 4.Bxb6 Qg6 5.Be3 Re8 6.Qd7 >>>Kg8 7.a4 Qf7 8.Qc6 Kh8 (29.906.848) 331.8 >>>best move: Bc1xh6 time: 3:04.325 min n/s: 321.286 CPU 98.6% nodes: 59.221.207 >>> >>>Deep Shredder here takes a little over 3 minutes, and plays Bxh6. I choose 3 >>>minutes, because it is a good time for an average tournament move. In 3 minutes >>>it hits roughly 60M nodes, at a rate of 321K nps. >>> >>>If CB had search 60M nodes, it would have taken 987 hours. However, as you can >>>see in Deep Shredder's eval, it _NEVER_ looks at Qxb6. So the knowledge is in >>>Deep Shredder. >> >>The output you pasted starts at 8 ply. Does Shredder look at Qxb6 from ply 1-7? >> I think it probably does at some point. > >I am not sure how much you pay attention to the eval of _most_ chess programs, >but most skip the first few plies. If you look at the average nps I was getting >(321,286 nps) and then look at the total nodes searched at the end of the first >shown ply (268,921) you'll see that it started on ply 8. Deep Shredder is >packed with knowledge, and is a "selective search" engine. > >Therefore, it's safe to say it "skipped" over the first 8 ply, and started to >actually "search" at ply 8. Perhaps some of the more knowledgable programmers >can give you more details on this. But just looking at the nodes, it makes >sense to me. > > >Slate Again, this is _wrong_. If DS can solve this instantly, then it can solve all the Nolot positions instantly, since _none_ of them rely on positional knowledge of any kind. This is about tactics, and at shallow searches, Qxb6 looks good. Find out how to make DS show the first 7 plies of search. In Crafty, this is done by "noise 0". I can assure you DS does _not_ start a search at depth=8. That is not how an iterated search works. It _must_ start at 1 or 2...
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.