The old Nemeth's trojan is still working.

Hello,

http://pagesperso-orange.fr/lefouduroi/eddy1.htm#int

While reviewing some stuff about Nemeth's anti-computer play on this page (sorry it is in french, there is also a german version since Nemeth is, i don't know if it has ever been translated into english ?, but you can download the PGN games from http://pagesperso-orange.fr/lefouduroi/config/en.zip ) :

Since this article was written in early 2000's i thought all this was no more up to date now, while chess engines have made tremendous progress.

FEN : r2q1rk1/p2b1pp1/2nb4/1pp1pPPQ/3pP3/1P4NN/P1P3P1/R3KB1R b KQ - 1 16


But let's take this position achieved in game 2, and most of engines i have tested just do not see the threat, not only a threat but a HUGE and crushing win for white (because of the open h file, with Q+R on it, and the poor white king has nowhere to go).

Thus many of the engines in 2009 still see 16... Qa5+ as their best move, and just cry all their tears after 17. Kd1!
I mean, they do evaluate the position as +1.0 and after the king move as -4.0 not to say even worst, this is a big big gap.

Nevertheless 2 engines are seing it correctly, Spike1.2 (-12.0) and Hermann_24 (-8.0) and in a small Rybka 2.3.2a (-2.1), but new Rybka 3.1 does not (-0.30) and do not foresee Kd1!, it sees Ke2 or Kf2 instead.

And even though, if the engine correctly evaluate the threat at ply 15., we just have to go back 1 or 2 ply earlier and it is blind again...

Here is the PGN :

Code:


[Event "Blitz30"]
[Site "?"]
[Date "2001.??.??"]
[Round "?"]
[White "Nemeth"]
[Black "Shredder 5"]
[ECO "B20"]
[Result "1-0"]

1. e4 c5 2. d3 d5 3. Nc3 d4 4. Nce2 Nc6 5. b3 Nf6 6. Ng3 e6
7. Bg5 h6 8. Bd2 b6 9. h4 Bd6 10. f4 e5 11. f5 O-O 12. Bg5
Bd7 13. Nh3 hxg5 14. hxg5 Nxe4 15. dxe4 b5 16. Qh5 Qa5+
17. Kd1 g6 18. Qh6 Qc3 19. Nf2 Qxa1+ 20. Ke2 d3+ 21. Kd2
Qc3+ 22. Kxc3 b4+ 23. Kb2 Bxf5 24. Qh8# 1-0


As explained in Nemeth's article, it is 13.Nh3 that comes to hide the mate threat to the computer, by deepening is tree analysis, and reject the mate threat beyond the evaluation horizon.

Well, this surprise me even more since very strong chess engines have became universally used to analyse postmortems. Seing that the old Nemeth's gambit is still puzzling computers when any human can see instantly that this h file is a crap and will soon or later cause terrible damage on kingside !

Another reason to learn the general principles teached in ICS course, and not put all your confidence in chess engines post-analysis.

I hope Bilbo is reading