problems that stump computers
Tom
I'm researching an article on computers and chess, and am interested in
finding a few problems that a master could solve, but that baffle even the
best chess engines. The more patently obvious to the master's discerning
eye -- ie the more they rely on "instinct" (however one is to define that!)
rather than calculation -- the better they'll suit my purposes.
Any suggestions?
Best regards,
Tom Mueller
Clifford Stern
Since no one has yet responded, here is an elegant example from the
problem world that I originally posted on October 5, 1999:
8/p1p1p3/2p3p1/6Pb/p3P1k1/P1p1PNnr/2P1PKRp/7B
Mate in 100
Frederik W. Nanning
Tijdschrift vd. KNSB
January 1935
A perceptive player would see that White needs to repeatedly lose a
tempo in order to force Black to exhaust all of his pawn moves so he
would be finally forced to move his rook with the White king at f2,
permitting RxN mate. The tempo maneuver consists of White moving his
king to the a1 corner where he triangulates, then returns to f2. This is
an 11-move sequence, which is repeated eight times to force Black to use
up all of his pawn moves. The tempo sequence is repeated one last time
to arrive at the mating position at move 100. A human solves this
because he is familiar with the winning concept, but a machine, which
doesn't operate on a conceptual level, is helpless to make progress in
the position.
Source: 96 Citaat-Problemen - edited by Dr. Meindert Niemeijer
Wassenaar, 1960
This is a booklet of problems accompanied by quotations. Here is the
one given for this problem:
Je n'ai fait celle-ci plus longue que parceque je n'ai pas eu le loisir
de la faire plus courte
Pascal
(Lettres provinciales, 16, 14 dec. 1656)
According to AltaVista's Babel Fish Translation, this means, "I did this
one longer only because I did not have the leisure to make it shorter."
Clifford Stern
ax...@lafn.org
NoKetch
I have this position in my database as a mate in 34. Here is the proposed
solution:
[Event "Mate in 34"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "1-0"]
[SetUp "1"]
[FEN "8/p1p1p3/2p3p1/6Pb/p3P1k1/P1p1PNnr/2P1PKRp/7B w - - 0 1"]
1. Ke1 c5 2. Kd1 e6 3. Kc1 c4 4. Kb1 a6 5. Ka1 e5 6. Ka2 c6 7. Kb1 c5 8. Kc1
Rh4 9. Kd1 Rh3 10. Ke1 Rh4 11. Kf2 Rh3 12. Ke1 Rh4 13. Kd1 Rh3 14. Kc1 Rh4 15.
Kb1 Rh3 16. Ka1 Rh4 17. Ka2 a5 18. Kb1 Rh3 19. Kc1 Rh4 20. Kd1 Rh3 21. Ke1 Rh4
22. Kf2 Rh3 23. Ke1 Rh4 24. Kd1 Rh3 25. Kc1 Rh4 26. Kb1 Rh3 27. Ka1 Rh4 28. Ka2
Rh3 29. Kb1 Rh4 30. Kc1 Rh3 31. Kd1 Rh4 32. Ke1 Rh3 33. Kf2 Rh4 34. Rxg3# 1-0
Dave
Luigi Caselli
news:20031210162511...@mb-m18.aol.com...
> >Since no one has yet responded, here is an elegant example from the
> >problem world that I originally posted on October 5, 1999:
> >
> >8/p1p1p3/2p3p1/6Pb/p3P1k1/P1p1PNnr/2P1PKRp/7B
> >
> >Mate in 100
> >
> >Frederik W. Nanning
> >Tijdschrift vd. KNSB
> >January 1935
>
> solution:
>
> [Event "Mate in 34"]
> [Site "?"]
> [Date "????.??.??"]
> [Round "?"]
> [White "?"]
> [Black "?"]
> [Result "1-0"]
> [SetUp "1"]
> [FEN "8/p1p1p3/2p3p1/6Pb/p3P1k1/P1p1PNnr/2P1PKRp/7B w - - 0 1"]
>
> 1. Ke1 c5 2. Kd1 e6 3. Kc1 c4 4. Kb1 a6 5. Ka1 e5 6. Ka2 c6 7. Kb1 c5 8.
Kc1
> Rh4 9. Kd1 Rh3 10. Ke1 Rh4 11. Kf2 Rh3 12. Ke1 Rh4 13. Kd1 Rh3 14. Kc1 Rh4
15.
> Kb1 Rh3 16. Ka1 Rh4 17. Ka2 a5 18. Kb1 Rh3 19. Kc1 Rh4 20. Kd1 Rh3 21. Ke1
Rh4
> 22. Kf2 Rh3 23. Ke1 Rh4 24. Kd1 Rh3 25. Kc1 Rh4 26. Kb1 Rh3 27. Ka1 Rh4
28. Ka2
> Rh3 29. Kb1 Rh4 30. Kc1 Rh3 31. Kd1 Rh4 32. Ke1 Rh3 33. Kf2 Rh4 34. Rxg3#
1-0
In your solution Black doesn't play his best moves.
He has to always move the Rook from h3 to h4 and from h4 to h3 and move the
pawns only when the white King is back to square f2.
In this case White needs 100 moves to checkmate.
Luigi Caselli
Mike Leahy
I pasted this into Bookup 2000 Professional with Ruffian and it started
giving White the advantage after abou 45 seconds. Definitely didn't see
the mate though.
Mike Leahy
"The Database Man!"
www.bookup.com
"Luigi Caselli" <luigic...@iol.it> wrote in message
news:VqMBb.181612$e6.67...@twister2.libero.it...
Mike Ogush
wrote:
This is a good example of how computers can make inferior moves
because thay cannot plan. For example Fritz needs to go to 18 ply to
realize that 1.Ke1 is better than 1.e5 and even here it thinks that
1.K1 is = (0.00) evaluation.
However, I think this example is somewhat flawed: After 1.Ke1 Rh4
checkmate in 99 moves is guaranteed after 1...Rh4 by playing 2.Kd1,
but although the variations are more complex White can actually mate
much sooner by playing 2.Rxh4. It turns out White can win the
exchange and the pawn at h2 and then either capture black's weak pawns
to create a passed or sacrifice on h5 ro create a passde g-pawn.
Some sample variations:
2.Nxh4 Kxh4 3.Rxh2+ Kxg5 4.Bf3 e5 5.Rh3 Bxf3 6.exf3 Nh5 7.Ke2 and
either White is going to be able to eat the queenside pawns with his
king or rook (if Black unblocks the h-file)
2.Nxh4 Kh3 3.Nf3 Bxf3 (3...Nxh1 4.Rxh2+ Kg3 5.Rxh5! Nf2 6.Rh6 and
after 7.Rxg6 Black will have to give up his knight to stop the g-pawn
from queening.) 4.Rxg3+ Kxg3 5.exf3 e5 6.f4 Kg4 (6...exf4 7.e5! and
white queens shortly.) 7.fxe5 Kxg5 8.Kf2 Kh6 9.Kg3 Kg7 10.Kxh2 Kf7
11.Kg3 Ke6 12.Kf4 and Black will be forced to give way. White will
queen an e-pawn and mate by around move 30.
When I look at Fritz's analysis after 1...Rh4 part of the reason the
engine gets so confused is after analyzing for about a minute and
getting to 17 ply it thinks that White has a strong advantage +1.25
pawns after 2.Nxh4 after two minutes it reaches 18 ply and gives an
evaluation of +1.50 a winning advantage to White.
Mike Ogush
Dana Turnmire
Try this position. Whoever moves first wins. The main line goes.
[FEN "4k3/5ppp/8/8/8/8/PPP5/3K4 w - -"]
1. Ke2 Kd7 2. Kf3 Kc6 3. a4 h5 4. c4 f5 5. Kg3 Kb6 6. b4 g5 (6.. g6 7. a5+ Ka6
8. c5 Kb5 9. Kg2 g5 (9.. h4 10. Kh2 f4 11. Kg1 g5 12. Kg2 g4 13. Kg1 {etc.} )
10. Kg3 g4 11. Kf2 f4 12. Kg2 h4 13. Kg1 f3 14. Kf2 h3 15. Kg3 {and now the
Black king must move, allowing a Qhite pawn to queen.} ) 7. a5+ Ka6 (7.. Kb7
8. c5 ) 8. c5 h4+ 9. Kh3 f4 (9.. Kb5 10. Kh2 {as in the variation above} ) 10.
c6 f3 11. b5+ Ka7 12. c7 g4+ (12.. Kb7 13. b6 g4+ 14. Kh2 g3+ 15. Kg1 h3 16.
a6+ Kc8 17. a7 ) 13. Kxg4 f2 14. c8=Q f1=Q 15. b6+ 1-0
-- -- -- -- BK -- -- --
-- -- -- -- -- BP BP BP
-- -- -- -- -- -- -- --
-- -- -- -- -- -- -- --
-- -- -- -- -- -- -- --
-- -- -- -- -- -- -- --
WP WP WP -- -- -- -- --
-- -- -- WK -- -- -- --
From "Basic Chess Endings" revised by GM Benko page 61 No.132
Tom Mueller
here?
Kerry Liles
"Tom Mueller" <tbmu...@libero.it> wrote in message
news:9jlDb.191159$e6.72...@twister2.libero.it...
Mike Ogush
Tom,
You may want to check out a book published in the last few months:
"How to Use Computers to Improve Your Chess" by Christian Kongsted.
The first half or so of the book deals with limitations of computers
as chess players. It has a number of positoins that computers
misevaluate, although the positions may not be "problems" (e.g. White
to move and win).
Mike Ogush
Tom Mueller
Tom
Damir Ulovec
Artificial Intelligence to play better? Or they still didn't "surmount"
babyish diseases...?
"Mike Ogush" <mike_...@hp.com> wrote in message
news:3fde1a1a....@news.corp.hp.com...
> ...
> It has a number of positoins that computers
> misevaluate, although the positions may not be "problems" (e.g. White
> to move and win).
> ...
Damir Ulovec
that is on advantage, and irony is that he was evaluate 0.00 (or near it)
untill I played 99. Kf2.
1. = (0.00): 99.Kf2
2. = (0.00): 99.Kd1
3. -+ (-3.72): 99.Rxh2 Rxh2 100.Nxh2+ Kh3 101.Nf3 Nxh1 102.Nxe5 Ng3 103.Nxc4
Nxe4 104.Nxa5 Nxg5 105.Nc4 Ne4 106.Nb6
4. -+ (-5.28): 99.Rxg3+ Rxg3 100.Nxh2+ Kh3 101.Nf3 Bxf3 102.Bxf3 Rxg5 103.Kf2
Kh2 104.Kf1 Rg1+ 105.Kf2 Rc1 106.Bg4
5. -+ (-5.47): 99.Nxh2+ Kh4 100.Kd1 Nxh1 101.Nf1 Ng3 102.Nxg3 Rxg3 103.Rxg3
Kxg3 104.Ke1 Kh4 105.Kf2 Kxg5 106.Kg3
==========
After played 99.Kf2...
1. +- (#1): 99...Rh4 100.Rxg3#
"Clifford Stern" <ax...@lafn.org> wrote in message
news:3fd75c18....@news.lafn.org...
HD
> Yes, really, what went wrong with computer chess programs? Is they really need
> Artificial Intelligence to play better? Or they still didn't "surmount"
> babyish diseases...?
>
Well, likes humans, programs "think" different about a lot of positions,
and the programmers haven't yet implemented a genuine human intuition
into the beasts. You'll have to wait a while before the computers become
humans, sorry. But ofcouse, chessprograms and computers are alike: They
are all buggy!
HD
darrz
wrote:
This is FEN (Forsythe Edwards Notation) for a chess position. Start at
the top side of the board.
Numbers represent the number of blank squares, always counting from
left to right, so "8" represents a blank row on the topmost row of the
board.
The "/" represents an end of the row marker. Lower case letters like
"p" represent the black pieces (usual suspects, p=pawn, n=knight,
r=rook, k=king, etc.), while upper case letters denote white's pieces.
"8/p1p1p3/2p3p1/6Pb/p3P1k1/P1p1PNnr/2P1PKRp/7B", alone, would not be a
full FEN position, however.
after the above notation, there would also be an indicator of castling
rights, denoted by KQkq if both sides can castle to either king of
queen's sides of the board, any en-passant captures, which side to has
the move, and move counter.
It may look like greek at first, but it actually works well with a
little experience with it.
Darrz