Author: Uri Blass
Date: 22:09:56 10/01/01
Go up one level in this thread
On October 01, 2001 at 18:58:30, Slater Wold wrote:
>On October 01, 2001 at 18:54:41, Uri Blass wrote:
>
>>On October 01, 2001 at 18:39:42, Slater Wold wrote:
>>
>>>On October 01, 2001 at 13:36:16, Robert Hyatt wrote:
>>>
>>>>On September 30, 2001 at 17:36:18, Slater Wold wrote:
>>>>
>>>>>When Junior 7 was released a few months back, most people who used it said that
>>>>>it was too aggressive, and sometimes would give material for almost no reason.
>>>>>
>>>>>Those claims were hushed shortly after it won the WMCCC.
>>>>>
>>>>>This game was a game against Chess Tiger 14.0, played on playchess.com. It
>>>>>shows a great example of Deep Junior giving away a piece, for little if any
>>>>>compensation.
>>>>>
>>>>>I understand that in essence, it is trading a piece for 3 pawns. But any human
>>>>>can look at the board and see, it's a trade that won't pay off in the end.
>>>>>
>>>>>
>>>>>Chess Tiger 14.0 - Deep Junior 7 [D53]
>>>>>
>>>>> 1.d4 Nf6 2.c4 e6 3.Nf3 d5 4.Nc3 Be7 5.Bg5 h6 6.Bxf6 Bxf6 7.Qd2 dxc4 {Last book
>>>>>move for Deep Junior 7} 8.e4 c5 9.d5 exd5 10.e5 Nd7 11.exf6 Nxf6 12.Qe3+ Qe7
>>>>>13.Qxe7+ Kxe7 14.0-0-0 Be6 15.Re1 Rhc8 16.g3 d4 17.Nb5 c3 18.bxc3 a6 19.Na3 Kf8
>>>>>20.Bc4 Bh3 21.Ne5 Rc7 22.cxd4 cxd4 23.Kd2 Rac8 24.f3 Bf5 25.Rc1 Re8 26.Rhe1 Bh7
>>>>>27.Bb3 Rxc1 28.Kxc1 b5 29.Nc2 Rc8 30.Kd2 Bxc2 31.Bxc2 Rc3 32.f4 a5 33.Bb3 g5
>>>>>34.fxg5 hxg5 35.Bxf7 b4 36.Be6 a4 37.Ng6+ Ke8 38.Bb3+ Re3 39.Bxa4+ Kf7 40.Bc2
>>>>>Rxe1 41.Kxe1 Ng4 42.h4 Kf6 43.h5 Ne3 44.Be4 Kg7 45.Ne5 g4 46.Nc6 Kh6 47.Nxb4 Nc4
>>>>>48.Bg6 Kg5 49.Nc6 d3 50.Bxd3 Nb6 51.Bg6 Nd5 52.Ne5 Nb6 53.Kf2 Na4 54.Nxg4 Nc5
>>>>>55.Kf3 Ne6 56.a4 Nd4+ 57.Ke4 Ne2 58.Be8 Nxg3+ 59.Kd5 1-0
>>>>>
>>>>>
>>>>>The key position is this:
>>>>>
>>>>>[D]rnbqk2r/pp3pp1/5b1p/2ppP3/2p5/2N2N2/PP1Q1PPP/R3KB1R b KQkq - 0 10
>>>>>
>>>>>Here Deep Junior 7 plays 10...Nd7?? with this eval:
>>>>>
>>>>>W,S - Deep Junior 7
>>>>>rnbqk2r/pp3pp1/5b1p/2ppP3/2p5/2N2N2/PP1Q1PPP/R3KB1R b KQkq - 0 1
>>>>>
>>>>>Analysis by Deep Junior 7:
>>>>>
>>>>>10...Bh4 11.Nxh4 Qxh4 12.Qxd5
>>>>> ³ (-0.32) Depth: 3 00:00:00
>>>>>10...Bg5 11.Nxg5 Qxg5 12.Qxd5
>>>>> ³ (-0.33) Depth: 3 00:00:00
>>>>>10...Bg5 11.Nxg5 Qxg5 12.Qxd5
>>>>> ³ (-0.33) Depth: 3 00:00:00
>>>>>10...Bg5 11.Qxd5 Qxd5 12.Nxd5 c3 13.Nxc3
>>>>> = (-0.10) Depth: 6 00:00:00 3kN
>>>>>10...Bg5 11.Nxg5
>>>>> = (0.20) Depth: 9 00:00:00 94kN
>>>>>10...Bg5 11.Qxd5 Nc6 12.Qxc4 Nd4 13.Rd1 Nxf3+ 14.gxf3 Qc7 15.f4 Bh4 16.Qa4+ Bd7
>>>>>17.Bb5 Bxb5
>>>>> = (0.23) Depth: 12 00:00:01 1645kN
>>>>>10...Be7 11.Qxd5 Nc6 12.Bxc4 Qxd5 13.Nxd5 Bd8 14.0-0-0
>>>>> = (0.21) Depth: 12 00:00:02 2218kN
>>>>>10...Be7 11.Nxd5 b5 12.b3 Bg4 13.bxc4 Bxf3 14.gxf3 Nd7 15.Nxe7 Qxe7
>>>>> = (0.18) Depth: 15 00:00:07 12048kN
>>>>>10...Nd7 11.exf6 Nxf6 12.Qe3+ Qe7 13.Nb5 Be6 14.Qe5 Ng4 15.Qxg7 0-0-0 16.h3 Rhg8
>>>>>17.Nxa7+ Kb8 18.Qc3 Nxf2 19.Kxf2 Kxa7
>>>>> = (-0.01) Depth: 15 00:00:37 23534kN
>>>>>10...Nd7 11.exf6 Nxf6 12.Qe3+ Qe7 13.Qxe7+ Kxe7 14.Ne5 Be6 15.0-0-0 d4 16.Na4
>>>>>Ne4 17.Bxc4 Bxc4 18.Nxc4 Nxf2
>>>>> = (-0.12) Depth: 18 00:03:58 448237kN
>>>>>
>>>>>(W, 30.09.2001)
>>>>>
>>>>>
>>>>>
>>>>>Slate
>>>>
>>>>
>>>>That seems wrong, of course. I get a score of near zero for Be7. Bg5
>>>>is the suggested book move. After Nd7 I get a score of roughly +1 for
>>>>white. A piece for three pawns almost always loses unless it ends up in
>>>>a king and pawn ending quickly and the pawns are mobile.
>>>>
>>>>This certainly isn't that case..
>>>
>>>Well, that was kinda my *sentiment*. Show a 1500 player this, and they'll laugh
>>>at the idea. Because the idea is 1 knight for 3 pawns. That's it. Not a
>>>knight and 3 pawns for mobility, or an easily won endgame, or anything
>>>reasonable.
>>>
>>>Then tell that 1500 that Deep Junior 7 getting 1.88M nps makes this move, and
>>>they'll laugh. (This is from an actual test of my own.)
>>>
>>>Perhaps this is a case of taking the good with the bad...........
>>>
>>>
>>>
>>>Slate
>>
>>They may laugh but I am practically sure that 1500 players are going to lose
>>with the piece against the 3 pawns of Deep Junior.
>>
>>I believe that Junior can see some positional bonus in this specific case
>>because it is not the program that is usually happy to sacrifice a piece for 3
>>pawns.
>>
>>I remember that there was a game when shredder lost for sacrificing a piece for
>>3 pawns when Junior7 understood that the human has the advantage after the
>>trade.
>>
>>Uri
>
>True enough.
>
>But it's like Petr said after defeating Deep Shredder on my dual Xeon's:
>
>"The problem with chess computers is they play like a 1500 player, and 2800
>player, all in the same game."
>
>
>
>Slate
I looked in the relevant game and I see other problems
[D]r1r5/1p2kpp1/p3bn1p/2p5/3p4/N1P2NP1/P4P1P/2K1RB1R b - - 0 19
Junior could do better by playing 19...dxc3 but it prefered to play 19...Kf8 and
20...Bh3
Junior for some reason likes it's bishop at h3.
Maybe it helps it for some king attacks but it is not good at this case
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.