Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Challenge position for engines! (Lots of luck)

Author: Uri Blass

Date: 07:29:47 03/05/02

Go up one level in this thread


On March 05, 2002 at 08:12:17, Albert Silver wrote:

>On March 05, 2002 at 06:36:58, Ed Schröder wrote:
>
>>On March 05, 2002 at 05:15:38, Uri Blass wrote:
>>
>>>Yace also likes f5 and has Rh3 in the main line so it seems that finding a
>>>better move than Rh3 is not easy.
>>>
>>>The question is if Rh3 is winning.
>>
>>I don't know, a further investigation is needed IMO. Rebel gives a +2.30 score
>>for Rh3 after 15 plies but probably most of that is coming from eval.
>>
>>00:00:01.5	1,34	7	416116	Rh3 Kf7 exf5 Bxb3 Qxg6+ Ke7 Rh7+ Rf7
>>00:00:02.5	1,34	8	948561	Rh3 Kf7 exf5 Bxb3 Qxg6+ Ke7 Rh7+ Rf7
>>00:00:04.9	1,34	9	2379413	Rh3 Kf7 exf5 Bxb3 Qxg6+ Ke7 Rh7+ Rf7
>>00:00:08.7	2,22	10	5564762	Rh3 Kf7 exf5 Bxb3 Qxg6+ Ke7 Rh7+ Rf7
>>00:00:16.6	2,10	11	12107255	Rh3 Kf7 exf5 Bxb3 Qxg6+ Ke7 Qg5+ Rf6
>>00:00:42.6	1,78	12	31920754	Rh3 Kf7 Qh7+ Kf6 Qh4+ Kg7 Bxe6 Qxe6
>>00:02:02.0	1,62	13	95158047	Rh3 Kf7 Qh7+ Kf6 Qh4+ Kf7 exf5 Bxb3
>>00:06:49.0	2,30	14	322198950	Rh3 Kf7 Qh7+ Kf6 Qh4+ Kg7 Bxe6
>>00:14:38.6	2,30	15	688286326	Rh3 Kf7 Qh7+ Kf6 Qh4+ Kg7 Qh7+ Kf6
>>
>>
>>
>>>If Rh3 is not winning then it means that Rebel found the right move for the
>>>wrong reasons
>>
>>That's a popular statement in this forum, I once had similar thoughts but
>>changed my mind. It does not matter if Nxg6! is played by eval or search, the
>>only thing that matters is playing the good move.
>
>Yes and no. Suppose it finds Nxg6 but yields a draw score with a perpetual check
>in its main line. Nxg6 is a winning move, so if Rebel were to play it thinking
>it drew, how do you know it won't content itself with the draw? Would you
>consider it found Nxg6, the _winning_ move, just because it played it? That for
>me is the crucial question. I often see programs find a very delicate sacrifice
>but with a draw score, and even by feeding them the responses, they do not win
>the game. So what then?
>
>Uri's comment here is relevant also because it isn't shown why Rebel immediately
>sacs the bishop back with Bc1. No doubt this is the best reply for Black, but
>did Rebel do this because it saw a genuine follow-up for White if f5 was played,
>or did it see a ghost?
>
>As to the comment on Rh3, it wasn't so much a complaint (I saw the smiley BTW)
>as comment on my challenge: find the strongest continuation. In this case, let's
>broaden this to find the best continuation after 25.Nxg6 hxg6 26.Qh6 Be6 27.Rd3
>f5.
>
>The strongest continuation BTW isn't 28.Rh3, it is 28.Rg3! Rf6 29.Rh3 Kf7
>30.Qh7+ Kf8 31.exf5!

Yace can learn +1.20 for white from this line
(I was too lazy to teach it more)

I added 31...Bxb3 32.axb3 Ne7 33.Qh8+ Ng8 34.b4 or 32...Rxf5 33.b4

+1.20 is still not enough to convince yace to play Bc1 because the score after
Bc1 is 1.41 for white.

New position
[D]4qrk1/8/pBn1b1pQ/1p2pp2/4P3/bBPR4/P4PPP/6K1 w - f6 0 1

Analysis by Yace 0.99.56:

4.exf5 Bxb3 5.axb3 Rxf5
  -+  (-1.69)   Depth: 1   00:00:00
4.Rg3 Rf6 5.Rh3 Kf7 6.Qh7+ Kf8 7.exf5 Bxb3 8.axb3 gxf5 9.Rg3 Qe6 10.b4 Ke8
11.Rg8+ Rf8 12.Qh5+ Qf7 13.Rxf8+ Kxf8 14.Qh6+ Ke8 15.Qxc6+ Qd7 16.Qg6+ Qf7
17.Qd6 Qe7 18.Qb8+ Kf7
  ±  (1.20)   Depth: 1   00:00:00
4.Rg3 Rf6 5.Rh3 Kf7 6.Qh7+ Kf8 7.exf5 Bxb3 8.axb3 gxf5 9.Rg3 Qe6 10.b4 Ke8
11.Rg8+ Rf8 12.Qh5+ Qf7 13.Rxf8+ Kxf8 14.Qh6+ Ke8 15.Qxc6+ Qd7 16.Qg6+ Qf7
17.Qd6 Qe7 18.Qb8+ Kf7
  ±  (1.20)   Depth: 2   00:00:00
4.Rg3 Rf6 5.Rh3 Kf7 6.Qh7+ Kf8 7.exf5 Bxb3 8.axb3 gxf5 9.Rg3 Qe6 10.b4 Ke8
11.Rg8+ Rf8 12.Qh5+ Qf7 13.Rxf8+ Kxf8 14.Qh6+ Ke8 15.Qxc6+ Qd7 16.Qg6+ Qf7
17.Qd6 Qe7 18.Qb8+ Kf7
  ±  (1.20)   Depth: 3   00:00:00
4.Rg3 Rf6 5.Rh3 Kf7 6.Qh7+ Kf8 7.exf5 Bxb3 8.axb3 gxf5 9.Rg3 Qe6 10.b4 Ke8
11.Rg8+ Rf8 12.Qh5+ Qf7 13.Rxf8+ Kxf8 14.Qh6+ Ke8 15.Qxc6+ Qd7 16.Qg6+ Qf7
17.Qd6 Qe7 18.Qb8+ Kf7
  ±  (1.20)   Depth: 4   00:00:00
4.Rg3 Rf6 5.Rh3 Kf7 6.Qh7+ Kf8 7.exf5 Bxb3 8.axb3 gxf5 9.Rg3 Qe6 10.b4 Ke8
11.Rg8+ Rf8 12.Qh5+ Qf7 13.Rxf8+ Kxf8 14.Qh6+ Ke8 15.Qxc6+ Qd7 16.Qg6+ Qf7
17.Qd6 Qe7 18.Qb8+ Kf7
  ±  (1.20)   Depth: 5   00:00:00  16kN
4.Rg3 Rf6 5.Rh3 Kf7 6.Qh7+ Kf8 7.exf5 Bxb3 8.axb3 gxf5 9.Rg3 Qe6 10.b4 Ke8
11.Rg8+ Rf8 12.Qh5+ Qf7 13.Rxf8+ Kxf8 14.Qh6+ Ke8 15.Qxc6+ Qd7 16.Qg6+ Qf7
17.Qd6 Qe7 18.Qb8+ Kf7
  ±  (1.20)   Depth: 6   00:00:00  76kN
4.Rg3 Rf6 5.Rh3 Kf7 6.Qh7+ Kf8 7.exf5 Bxb3 8.axb3 gxf5 9.Rg3 Qe6 10.b4 Ke8
11.Rg8+ Rf8 12.Qh5+ Qf7 13.Rxf8+ Kxf8 14.Qh6+ Ke8 15.Qxc6+ Qd7 16.Qg6+ Qf7
17.Qd6 Qe7 18.Qb8+ Kf7
  ±  (1.20)   Depth: 7   00:00:00  232kN
4.Rg3 Rf6 5.Rh3 Kf7 6.Qh7+ Kf8 7.exf5 Bxb3 8.axb3 gxf5 9.Rg3 Qe6 10.b4 Ke8
11.Rg8+ Rf8 12.Qh5+ Qf7 13.Rxf8+ Kxf8 14.Qh6+ Ke8 15.Qxc6+ Qd7 16.Qg6+ Qf7
17.Qd6 Qe7 18.Qb8+ Kf7
  ±  (1.20)   Depth: 8   00:00:01  796kN
4.Rg3 Rf6 5.Rh3 Kf7 6.Qh7+ Kf8 7.exf5 Bxb3 8.axb3 gxf5 9.Rg3 Qe6 10.b4 Ke8
11.Rg8+ Rf8 12.Qh5+ Qf7 13.Rxf8+ Kxf8 14.Qh6+ Ke8 15.Qxc6+ Qd7 16.Qg6+ Qf7
17.Qd6 Qe7 18.Qb8+ Kf7
  ±  (1.20)   Depth: 9   00:00:05  2723kN
4.Rg3 Rf6 5.Rh3 Kf7 6.Qh7+ Kf8 7.exf5 Bxb3 8.axb3 gxf5 9.Rg3 Qe6 10.b4 Ke8
11.Rg8+ Rf8 12.Qh5+ Qf7 13.Rxf8+ Kxf8 14.Qh6+ Ke8 15.Qxc6+ Qd7 16.Qg6+ Qf7
17.Qd6 Qe7 18.Qb8+ Kf7
  ±  (1.20)   Depth: 10   00:00:18  8819kN

(Blass, Tel-aviv 05.03.2002)

>
>I fed this to Fritz 7.0.0.6 and let it run all night on it. It found Nxg6 after
>17 plies, but changed it's mind again, and on ply 19 it preferred Nf3 again.
>Could you let Rebel run for an hour on the original to see the lines it
>produces?
>
>                                          Albert
>
>
>>>and the question in this case is if Rebel can practically find the
>>>winning move in the following position(I did not investigate to find the right
>>>move):
>>>
>>>[D]4qrk1/8/pBn1b1pQ/1p2pp2/4P3/bBPR4/P4PPP/6K1 w - f6 0 4
>>>
>>>
>>>Here is yace main line before f7-f5
>>>
>>>Analysis by Yace 0.99.56:
>>>
>>>
>>>3...f5 4.Rh3 Kf7 5.exf5 Bxb3 6.fxg6+ Ke7 7.axb3 Rf6 8.Qh7+ Ke6 9.g7 Ne7 10.Rd3
>>>Qg6 11.Qh3+ Nf5 12.Bd8
>>>  ²  (0.26)   Depth: 12   00:10:16  199936kN
>>
>>
>>[d]4qrk1/8/pBn1b1pQ/1p2pp2/4P3/bBP4R/P4PPP/6K1 b - - 0 1
>>
>>Following Rebel's analysis
>>
>>1. Rh3 Kf7 2. Qh7+ Kf6 3. Qh4+ Kg7 4. Bxe6 Qxe6 5. exf5 Qxf5 6. Qh6+ Kf7 7. Rf3
>>Qxf3 8. gxf3 1-0
>>
>>or...
>>
>>1. Rh3 Kf7 2. Qh7+ Kf6 3. Qh4+ Kg7 4. Bxe6 Qxe6 5. exf5 Qxa2 6. Qh7+ Kf6 7.
>>Qxg6+ Ke7 8. Rh7+ Rf7 9. Rxf7+ Qxf7 10. Qxc6 Qxf5
>>
>>White is a healthy pawn up while the black king is draughty, Rebel's score still
>>over 2 pawns, +2.39 to be exactly.
>>
>>Ed
>>
>>
>>
>>
>>
>>>(Blass, Tel-aviv 05.03.2002)
>>>
>>>
>>>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.