Aegon 1996 Report by Thorsten Czub
Bruce Moreland
In article <31A067...@tott.kih.no>, ma...@tott.kih.no says...
>
>We proofed this at the championship 1995 in Paderborn, where Chess System
>Tal made big Tal-games, beat Genius with a nice sacrifice...
>
>[Event "13th World Microcomputer Chess Championships"]
>[Site "Paderborn (Germany)"]
>[Date "Oct 11 1995"]
>[Round "4"]
>[White "Chess Genius"]
>[Black "Chess System Tal"]
>[Result "0-1"]
>
>1. c4 e5 2. Nc3 Nf6 3. Nf3 Nc6 4. g3 d5 5. cd5 Nd5
>6. Bg2 Nb6 7. 0-0 Be7 8. a3 0-0 9. b4 Re8 10. Rb1 Bf8
>11. d3 a5 12. b5 Nd4 13. Nd2 Rb8 14. e3 Ne6 15. Nf3 Ng5
>16. Ng5 Qg5 17. Ne4 Qd8 18. Qc2 Bg4 19. f4 ef4 20. Rf4 Be6
>21. Rf2 f5 22. Nc3 Bf7 23. Ne2 Qd7 24. Qc3 Nd5 25. Bd5? Qd5
>26. Nd4 Re5 27. Qc7 Ra8 28. Qc3? Rae8 29. Nf3? Re3!!
a b c d e f g h
+---+---+---+---+---+---+---+---+
8 | |///| |///| r |/b/| k |///| 8
+---+---+---+---+---+---+---+---+
7 |///| p |///| |///| b |/p/| p | 7
+---+---+---+---+---+---+---+---+
6 | |///| |///| |///| |///| 6
+---+---+---+---+---+---+---+---+
5 |/p/| P |///| q |/r/| p |///| | 5
+---+---+---+---+---+---+---+---+
4 | |///| |///| |///| |///| 4
+---+---+---+---+---+---+---+---+
3 |/P/| |/Q/| P |/P/| N |/P/| | 3
+---+---+---+---+---+---+---+---+
2 | |///| |///| |/R/| |/P/| 2
+---+---+---+---+---+---+---+---+
1 |///| R |/B/| |///| |/K/| | 1
+---+---+---+---+---+---+---+---+
a b c d e f g h
4rbk1/1p3bpp/8/pP1qrp2/8/P1QPPNP1/5R1P/1RB3K1 b - - 4 29
The key is Rxe3 and it does appear to win. I am intrigued by this
sacrifice because it came against a very strong opponent, namely Genius,
and was made by a program whose design includes speculative chess.
I wanted to see if my program could find Rxe3 and it can, sort of. On the
P6, a much faster machine than CST was running in Paderborn, Ferret finds
Rxe3 in 20 seconds (9 plies), but discards it in favor of R5e7 at 1:05 (10
plies), then finds it again and sticks with it at 3:36 (11 plies). The PV
at 15 plies (about 40 minutes) still showed it at -0.57, so it's not like
it could see this out to mate.
00:40:26.359 15 -57 Rxe3 Bxe3 Rxe3 Nd4 g6 Re2 Rxe2 Nxe2 Qa2 Qe1 Qxa3 Qd2
Bb4 Qe3 Bc5 Nd4
I have heard that both CST and Genius had Genius ahead after Rxe3.
This is an interesting tactical problem if anyone else would like to try
to solve it.
bruce
Chris Whittington
Latest version of CS Tal (which is probably over-speculative right now)
0 secs ...Rxe3 +0.28 initial speculative stuff
2 +0.28
9 -0.25 gets worried there's defensive resources
20 -0.25
71 > +0.11 but now it looks ok
91 +0.86
231 +0.50
446 +0.58
Chess System Tal sees:
Bishops raking the g1-a7, h1-a8 diagonal with white Kg1
Possible discovered checks (Bc5, Re3)
Nasty pin on Rf2
Mate threats g2,h1
pawn hanging a3
pawn break on f5, with more stuff against the white king
all this and black king tucked away nicely.
Its not just one of these things but the combination of all of
them which makes white's position untenable. Too many threats.
CST decides to have a go and sac the rook, but, to be fair,
sometimes in positions like these there's a deep defence.
Opponent looks like he's in a real mess, but everything holds
somehow, but CST has sac-ed - and so then loses. 50-50 ?
Unfortunately, all the time required to calculate this stuff
in the evalauation function doesn't come cheap. 1800 n/s for a
P90 in this position.
Best regards,
Chris Whittington
Al Cargill
In article <31A067...@tott.kih.no> ma...@tott.kih.no "Mads Brevik" writes:
> .., and showed
> different bugs
>
> [Event "13th World Microcomputer Chess Championships"]
> [Site "Paderborn (Germany)"]
> [Date "Oct 15 1995"]
> [Round "11"]
> [White "Chess System Tal"]
> [Black "XXXX"]
> [Result "0-1"]
>
> 1. Nf3 d6 2. d4 g6 3. e4 Bg7 4. Bc4 e6 5. Nc3 Nc6
cut
> 71. c6 Kb6 72. Rb8 Ka7 73. c7 Bb4 74. Kc6 g2 75. c8=Q???? 0-1
>
Sorry but why are there these question marks and the result 0-1 as
this forces mate on move 76 for white when I played it through on CB?
Al
Bruce Moreland
In article <832793...@alcarg.demon.co.uk>, A...@alcarg.demon.co.uk
says...
The move actually played was c8=N. I haven't gone over the game here, but
I saw it live. CST had a mate in 2 or so, but instead decided to get a
knight instead of a queen, drop a rook, and allow XXXX to queen, all on
this one move.
I heard that the problem was one of communication between the engine and
the user-interface, something ended up being zero that shouldn't have
been, and zero was a knight.
It happens. If you hook your program up to ICC or FICS it will happen
less often. It's better to lose 30 rating points on the net than it is to
drop a game in this kind of event.
bruce
--
The opinions expressed in this message are my own personal views
and do not reflect the official views of Microsoft Corporation.
Chris Whittington
Aaaaarrrrrggggghhhhhh.
I've gone paranoic :)
This is of course a conspiracy to keep on and on and on about
c8=N :)
Well, it proves CST has an underpromotion algorithm, I suppose.
Chris Whittington
Al Cargill
In article <83285645...@cpsoft.demon.co.uk>
chr...@cpsoft.demon.co.uk "Chris Whittington" writes:
> Aaaaarrrrrggggghhhhhh.
>
> I've gone paranoic :)
>
> This is of course a conspiracy to keep on and on and on about
> c8=N :)
>
> Well, it proves CST has an underpromotion algorithm, I suppose.
>
> Chris Whittington
>
>
Well it just proves that CST has a more human playing style than the
other computers!! That really must have been an "I just don't believe
it" moment.
Al