Saturday, November 26, 2011

Chess Engines are Like Eye Glasses

Capablanca once wrote, “Chess books should be used as we use glasses: to assist the sight, although some players make use of them as though they conferred sight.”  The same thing could be said of chess engines.


The below position is from Knaak-Geller, Moscow, 1982 and was analyzed by CCGM Robin Smith in his book Modern Chess Analysis.  Smith’s book is badly outdated when it comes to engine analysis, but that doesn’t stop his “human” analysis from being instructive plus his general advice on how to properly analyze a position with an engine is still valid.  I enjoy setting up some of the positions from the book, published in 2004, and letting today’s engines give an opinion.

Smith examined the position below with Chess Tiger, Fritz, Junior, Shredder, Chessmaster, Rebel and Yace and all of them strongly favored White.  White appears to have a distinct advantage but no immediate win, but none of the programs found the correct move.  All of them preferred 22.h6 which looks quite logical and is a move many of us probably would have played.  Smith explained that after 22.h6 g4! White will have a harder time putting pressure on the black K because White’s h-Pawn actually shields Black’s K from attack.  Nimzovich’s principle that the threat is stronger than it execution applies here.  The White h-Pawn is actually restraining Black’s g-Pawn from advancing and thereby making it a permanent weakness. He ran some engine vs. engine tournaments from the position after 22.h6 g4! and Black scored decisively.
GM Mark Dvoretsky outlined the goals of Knaak’s long-term plan and pointed out that this position requires quiet maneuvering.  Dvoretsky wrote in Positional Play, “It is not immediately clear how White should best proceed…some of his pieces cannot take part in the (K-side attack.)  First he must consolidate.”
Dvoretsky then outlines the correct plan: White must regroup by placing his B on b3 from where it will also shield the b-Pawn from attack.  Then the b1R will defend the d4P.  The N will occupy the f4 square from where it also attacks d5 while also being closer to the Black K.  While this plan may not win outright, it puts a lot of pressure on Black to find the best defense.  In the game, Geller did not defend well and lost quickly.
I let Critter 1.2 64–bit, Fire 1.5 xTreme x64, Ivanhoe B47cBx64a and Houdini 1.5 examine the position for about 30 minutes just to see what they would suggest.
Critter 1.Rbc1 (0.65)
Fire: 1.Rbc1 (0.49)
Ivanhoe: 1.h6 (0.54)
Houdini: 1.Kg2 (0.51)

Note that all the evaluations are pretty close but Ivanhoe played the definitely inferior 1.h6.  Both Critter and Fire selected 1.Rbc1 then followed it up with the incorrect 3.h6.  Houdini didn’t do much better. It played a couple of pointless moves then played 3.h6.  You will also notice that after 3.h6 that all of the engines decided NOT to capture the h-Pawn except Houdini which chose the horrible defense 3…gxh6.

Another interesting thing is that in the final position Geller resigned, but according to Houdini the evaluation is only about 0.60 P’s in White’s favor and I tried several engines, but was unable to find a line that gave White even a full P advantage.  As has been said many times, when it comes to engine positional evaluations and a GM’s, always trust the GM.  I guess Geller knew he was thoroughly beaten despite what the engines say.

2 comments:

  1. It took HIARCS some time to find it.


    2r2nk1/pp2r1pn/1q2bp2/3p3P/3P4/2NB1PQ1/PP2NK2/1R5R w - -

    Engine: HIARCS 13.2 SP [Deeper] (4 MB)
    by Mark John Uniacke

    9/22 0:00 +0.25 1.h6 (112.512) 211

    9/22 0:00 +0.56 1.h6 g5 2.f4 Qc7 3.fxg5 fxg5 4.Kg2 (113.669) 211

    10/24 0:00 +0.58 1.h6 g5 2.f4 Rd8 3.Rhg1 Bf7 4.Qf3 Ree8
    5.Kg2 Qe6 (207.428) 231

    11/24 0:01 +0.62 1.h6 g5 2.f4 Rd8 3.Rhg1 Kh8 4.fxg5 Nxg5
    5.Ke1 (361.664) 232

    12/25 0:02 +0.55 1.h6 g5 2.f4 Rce8 3.Qf3 Qd6 4.Nb5 Qd7
    5.Rbg1 g4 6.Nxa7 (663.104) 241

    13/27 0:04 +0.51 1.h6 g5 2.f4 Rce8 3.Qf3 Bf7 4.Bb5 Rd8
    5.Kg2 (1.200.372) 246

    13/27 0:05 +0.52 1.f4 (1.246.755) 246

    13/28 0:05 +0.52 1.f4 f5 2.h6 Nf6 3.Qh4 N8h7 4.hxg7 Ng4+
    5.Ke1 Rxg7 (1.325.885) 245

    14/35 0:10 +0.28 1.f4 f5 2.h6 Nf6 3.Qh4 Bd7 4.Rbg1 g6
    5.Qg5 Kh7 (2.526.855) 233

    14/35 0:10 +0.29 1.h6 (2.543.238) 233

    14/35 0:15 +0.44 1.h6 g5 2.f4 Rce8 3.fxg5 fxg5 4.Kg2 Bf7 (3.363.625) 220

    14/35 0:18 +0.45 1.Qg2 (4.189.528) 227

    14/35 0:23 +0.45 1.Qg2 Qd6 (5.443.268) 234

    15/35 0:48 +0.24 1.Qg2 Ng5 2.h6 Rce8 3.Rbg1 Qxb2 4.f4 Bf5
    5.hxg7 Bxd3 6.gxf8Q+ Rxf8 (12.036.599) 246

    15/35 0:49 +0.25 1.h6 (12.078.393) 246

    15/35 0:54 +0.46 1.h6 g5 2.f4 Rce8 3.fxg5 fxg5 4.Kg1 Rf7
    5.Qg2 Rd8 6.b3 Qc7 7.Rf1 (13.394.976) 247

    16/35 1:27 +0.42 1.h6 g5 2.f4 Rce8 3.fxg5 Nxg5 4.Qf4 Nfh7
    5.b3 (21.777.348) 249

    16/35 2:15 +0.43 1.Rbc1 (34.513.401) 254

    16/37 2:41 +0.43 1.Rbc1 Rce8 2.b3 Bd7 3.f4 f5 (41.546.311) 256

    16/37 3:32 +0.44 1.Bc2 (54.910.104) 257

    16/37 4:57 +0.52 1.Bc2 Ng5 2.h6 g6 3.Bb3 Kh7 4.Rbd1 Rd8
    5.Nf4 Qc6 (76.740.499) 258

    17/37 6:09 +0.53 1.Bc2 Ng5 2.h6 g6 3.Bb3 Kh7 4.Rbd1 Rd8
    5.Nf4 Red7 6.Kg2 a6 (95.758.917) 259

    18/37 12:28 +0.60 1.Bc2 Rce8 2.Bb3 Bf7 3.Rbd1 (196.897.394) 263

    19/43 24:12 +0.60 1.Bc2 a6 2.Bb3 Rd7 3.Rbd1 Rcd8
    4.Rdg1 Ng5 5.h6 g6 6.Na4 (384.839.874) 264

    20/44 47:11 +0.65 1.Bc2 Qa5 2.Bb3 Rd8 3.Nf4 Red7
    4.Rbd1 (754.557.174) 266

    ReplyDelete
  2. Yes i am totally agreed with this article and i just want say that this article is very nice and very informative article.I will make sure to be reading your blog more. You made a good point but I can't help but wonder, what about the other side? !!!!!!THANKS!!!!!! lasik Michigan

    ReplyDelete