Don't blindly believe the computer!

Aug 25, 2011, 6:24 AM |

The danger of relying on computer evaluations!


This endgame study was shown by James Plaskett to a group of strong Grandmasters at the 1987 Brussels super-GM tournament and was later published in the Telegraph in an article by David Norwood so acknowledgements and thanks also to Andy Foulds from my Daventry chess club for showing this position and article to me.

Myself and a chess pal managed to solve the position but only with prompting... i.e. “that’s the right move” and then being able to move the pieces as the position unfolded. Of course, I also knew that it was “white to win”. I managed, by logical reasoning, to realise what needed to be done but getting there was another matter entirely. So am I better at problem solving than super-GM’s? No, I think not! However, three of us spent a very enjoyable hour (possibly more!) looking at the problem over a pint. Andy, thanks for torturing us with this beautiful puzzle!

The position is “white to play and win” and this had flummoxed all of the players at the tournament except the late, great, Mikhail Tal who apparently looked at the position for a few minutes, went for a walk in the park before returning with the solution.

Remember, this was 1987 so it wasn’t as though Tal could refer to Sciddy on his Smart Phone or relay the position to someone with Fritz or Deep Rybka to obtain the answer.

In the time it has taken to type this, I have had the position analysing on my own computer using Fritz 11. Okay, it’s a laptop but it’s still got a pretty decent processor. There are, by my calculations, 19 legal moves for white to play some of which include promoting the pawn to other than a Queen on the first move.

So far, Fritz has come up with:

New game - Fritz 11, Blitz:0' Northampton

8/3P3k/n2K3p/2p3n1/1b4N1/2p1p1P1/8/3B4 w - - 0 1

Analysis by Fritz 11:

1. -+  (-2.16): 1.d8Q Nf7+ 2.Kd7 Nxd8 3.Kxd8 Ba5+ 4.Kd7 Bc7 5.Nxe3 Bxg3 6.Kc6 Kg6 7.Kb5 Nc7+ 8.Kc4 h5 9.Kxc3 Bf4 10.Kd3 h4 11.Bg4 Be5 12.Ke4

2. -+  (-2.45): 1.d8R e2 2.Bc2+ Kg7 3.Re8 c4+ 4.Kc6 Nf3 5.Rxe2 Nd4+ 6.Kb6 Nxe2 7.Kxa6 Nd4 8.Ne3 Kf7 9.Bb1 Be7 10.Nc2 Nf5 11.Kb5 Nxg3 12.Nd4 h5

3. -+  (-2.55): 1.Kc6 Nf7 2.Nxe3 Nd8+ 3.Kb6 Nb8 4.Kc7 Nbc6 5.Ba4 Ba5+ 6.Kd6 Nd4 7.Kxc5 N8e6+ 8.Kd5 Kg7 9.Ke4 Ne2 10.Kf3 N6d4+ 11.Ke4 Kf7 12.Nc4 Bc7

4. -+  (-2.72): 1.Nxe3 Ba5 2.Be2 Nb8 3.Ke7 Nc6+ 4.Kd6 Nd4 5.Bd1 Kg7 6.Kxc5 Nge6+ 7.Kc4 Kf7 8.Ba4 h5 9.Kd3 Nf3 10.Ke4 Nfg5+ 11.Kd5 Ke7 12.Nc4

5. -+  (-3.19): 1.Ke7 e2 2.Bxe2 c4+ 3.Kf6 Ba5 4.Bd1 Ne4+ 5.Kf7 Nb4 6.Ne3 Nxg3 7.Nxc4 Bc7 8.Ne3 h5 9.Nc2 Nc6 10.Ne3 h4 11.Nd5 Be5 12.Ke6

6. -+  (-3.44): 1.Kd5 Nc7+ 2.Kd6 Nce6 3.Nxe3 Kg6 4.Bc2+ Kf7 5.Nf5 h5 6.g4 hxg4 7.Nh6+ Kf8 8.Nxg4 Ba5 9.Ne3 Bc7+ 10.Kc6 Ke7 11.Nd5+ Kd8 12.Nxc7 Nxc7 13.Bd3 Nce6

7. -+  (-3.46): 1.Bb3 Ne4+ 2.Kd5 Ba5 3.Nxe3 Nxg3 4.Kc4 Nb8 5.Ba4 Kg7 6.Kb5 Bd8 7.Kxc5 Kf7 8.Kd6 Ne4+ 9.Kd5 Nf6+ 10.Kd6 Be7+ 11.Kc7 Nbxd7 12.Bxd7 Nxd7 13.Kxd7 h5

8. -+  (-3.49): 1.Bc2+ Kg7 2.Nxe3 Ba5 3.Bb1 Bd8 4.Kd5 Nf7 5.Kc4 h5 6.Nd5 Ne5+ 7.Kxc3 Ba5+ 8.Kb3 Nxd7 9.Nf4 Nf6 10.Bd3 Nc7 11.Kc4 Bb4 12.Bf5

9. -+  (-4.01): 1.Ke5 Ba5 2.Nxe3 Nb4 3.g4 Kg6 4.Kd6 Ne4+ 5.Ke6 Bc7 6.Bc2 Nxc2 7.Nxc2 Nf6 8.Ne3 Kg5 9.Ke7 c4 10.d8Q Bxd8+ 11.Kxd8 Kf4 12.Nxc4

10. -+  (-4.09): 1.Nf6+ Kg7 2.Nh5+ Kg6 3.Nf4+ Kf5 4.Nd5 Ne4+ 5.Kc6 Nb8+ 6.Kc7 Nxd7 7.Nxe3+ Kg5 8.Kxd7 Nxg3 9.Ba4 Kf4 10.Nd5+ Ke5 11.Ne7 h5 12.Kc6 Ke6 13.Ng6 c4 14.Bc2

11. -+  (-4.31): 1.Be2 Nf7+ 2.Ke7 Kg7 3.Nxe3 c4+ 4.Ke6 Nc5+ 5.Kd5 Nxd7 6.Kxc4 Ba5 7.Bd1 Nd6+ 8.Kd5 Ne8 9.Bc2 h5 10.Nc4 Nef6+ 11.Kd4 Bb4 12.Bf5 Nc5

12. -+  (-4.66): 1.Ne5 Ba5 2.Nc4 Bc7+ 3.Ke7 e2 4.Bxe2 Nb4 5.Ne3 Kg7 6.Bd1 Nc6+ 7.Ke8 Nf7 8.Nf5+ Kf6 9.Ne3 Nd6+ 10.Kf8 Nb4 11.Ba4 Nb7 12.g4 Bf4 13.Nf5

13. -+  (-6.62): 1.Ba4 Nf7+ 2.Ke7 Kg7 3.d8Q c4+ 4.Ke8 Nxd8 5.Nxe3 Nf7 6.Nd5 Ba5 7.Ne3 Bc7 8.Ke7 Bd6+ 9.Ke8 Be5 10.g4 Nd6+ 11.Ke7 Nc5 12.Bc2 Nc8+

14. -+  (-8.43): 1.Bf3 Ba5 2.Nxe3 Nxf3 3.Kd5 Kg6 4.Kc4 Nb8 5.d8Q Bxd8 6.Kxc3 Bc7 7.Kc4 Nd7 8.g4 Kg5 9.Kd5 Bf4 10.Ke4 Bxe3

15. -+  (-9.77): 1.Nh2 Ba5 2.Ng4 e2 3.Bxe2 Ne4+ 4.Ke7 c2 5.Nf6+ Kh8 6.Nxe4 Nb8 7.d8Q+ Bxd8+ 8.Kxd8 c1Q 9.Nd6 Qe3 10.Bc4 Qxg3 11.Kc7 Qe5 12.Kxb8 Qxd6+ 13.Kb7 Kg7 14.Kc8

16. -+  (-13.16): 1.d8B e2 2.Bxe2 c4+ 3.Kc6 c2 4.Nf6+ Kh8 5.Nd5 c1Q 6.Bb6 Ne4 7.Bf3 Qc2 8.Bd4+ Kg8 9.Nxb4 Nxb4+ 10.Kb5 Qd3 11.Bxe4 Qxd4 12.Kxb4 Qxe4 13.Kc3 Kf7

17. -+  (-14.61): 1.d8N e2 2.Bxe2 c4+ 3.Kc6 c2 4.Nf6+ Kg6 5.Nd5 c1Q 6.Nb7 Ne6 7.Kb6 Nac7 8.Nxc7 Qe3+ 9.Kc6 Nxc7 10.Bxc4 Na8 11.Kb5 Qb6+ 12.Ka4 Qxb7 13.Bd3+ Kg7 14.Bf5

18. -+  (-16.56): 1.Nxh6 Kxh6 2.Kc6 Nf7 3.Kb7 c2 4.Bxc2 e2 5.Bd3 e1Q 6.Bxa6 Qe7 7.Bb5 Nd6+ 8.Kb6 Qd8+ 9.Kc6 Nxb5 10.Kxb5 Qxd7+ 11.Kc4 Kg6 12.g4 Kg5 13.Kb3 Qd3+

19. -+  (-28.06): 1.Nf2 exf2 2.Bc2+ Kg7 3.Kc6 Nb8+ 4.Kb7 Nxd7 5.Bd3 Ne5 6.Bf1 c2 7.Kb6 c1Q 8.Bg2 Kh7 9.g4 Nxg4 10.Kb5 Ne3 11.Bf1 Qxf1+ 12.Kb6 Kg6

That’s analysis for all 19 legal moves..... and Fritz had been thinking for around 9 minutes but the evaluation for the best move has not shifted by more than a few hundredths of a pawn. Normally, Fritz is brilliant at seeing forced checkmates. I’ll give you a clue... It’s mate in 14 – Fritz doesn’t see it until mate in 10 and only then when the moves are played through but to be fair to the silicon beast, it does see a  mate in 9 in very short order!

Therefore I conclude that it is not always wise to believe the computer’s evaluation, even on some modern machines. I daresay a faster computer using Deep Rybka may well see things more deeply but it’s still food for thought (perhaps whilst taking a nice walk in the park?).