PDA

View Full Version : Checking solutions

Czentovic
23-06-2011, 11:32 PM
Why don't problem composers use chess analysis engines to check their solutions? For example, look at the current Endgame Study #751 chesscafe 5K1k/6pp/5p2/3PPP2/p7/1P3b2/2p4B/8 w - - 0 1
http://www.chesscafe.com/endgame/endgame.htm

Proposed Solution 1.Bf4 [1.exf6? gxf6 2.Bf4 h5] 1...g5 [1...fxe5 2.f6 g5 3.f7; 1...h5 2.d6; 1...Bxd5 2.exf6 g5 3.fxg6 hxg6 4.bxa4] 2.d6 Bc6 3.e6 gxf4 4.d7 c1Q 5.d8Q 1-0

It only takes Fritz 11 or Houdini a few seconds to discover that 1...g5 loses for Black, and that 1...Bxd5 draws for Black. The given analysis of 1...Bxd5 is simply wrong because 2...g5 loses and 2...h5 draws.

Kevin Bonham
23-06-2011, 11:54 PM
I don't think Bogoljubov would have had access to a computer to check his solutions when he composed that one in 1915!

Nonetheless it is a shame that Chesscafe offers problems without computer-checking them for cooks or at least noting when a problem might be cooked.

I've copied some lines - both those from the "solution" and after 2...h5! which is not covered - into the game viewer below. It does indeed seem to be a draw. I'm always careful not to assume it's drawn just because a computer says 0.00 because sometimes the win is hidden too far into the position for the computer to see it.

1.Bf4 [1.exf6?? gxf6 2.Bf4 h5-+] 1...Bxd5 [1...g5 2.d6 Bc6 3.e6 gxf4 4.d7 c1=Q 5.d8=Q][1...fxe5 2.f6 g5 3.f7 gxf4 4.Ke7+-][1...h5 2.d6] 2.exf6 h5! [Solution gives only 2...g5?? 3.fxg6 hxg6 4.bxa4+-] 3.fxg7+ Kh7 4.bxa4 h4 5.a5 h3 6.f6 [6.a6 h2 7.a7 h1=Q 8.a8=Q Qg2=] 6...Kg6 7.a6 c1=Q! 8.Bxc1 h2 9.a7 h1=Q 10.a8=Q Bxa8 11.g8=Q+ Kxf6 tablebase draw

Czentovic
24-06-2011, 12:01 PM
Here is another example, from thread 10395 on chesschat
8/2p5/3prpb1/4k3/4p3/4KP2/4NP2/2R5 w - - 0 1

The suggested solution is 1.f4+ Kf5 2.f5 Bxf5 but this loses for Black.

The correct solution is still quite interesting:

1. f4+ Kd5 2. f5 Ke5 3. fxe6 Be8 4. Rxc7 Kxe6 5. Kxe4 Bg6+ 6. Ke3

and Black is in serious trouble, but avoids the mate in 7 that otherwise occurs. The problem by Kakovin (1936) is from Chernev's book "Chessboard Magic" and Chernev comments that 2.f5 is "a pretty sacrifice that Black must accept". Wrong, Irving!

Kevin Bonham
24-06-2011, 12:34 PM
Here is another example, from thread 10395 on chesschat
8/2p5/3prpb1/4k3/4p3/4KP2/4NP2/2R5 w - - 0 1

The suggested solution is 1.f4+ Kf5 2.f5 Bxf5 but this loses for Black.

The correct solution is still quite interesting:

1. f4+ Kd5 2. f5 Ke5 3. fxe6 Be8 4. Rxc7 Kxe6 5. Kxe4 Bg6+ 6. Ke3

and Black is in serious trouble, but avoids the mate in 7 that otherwise occurs. The problem by Kakovin (1936) is from Chernev's book "Chessboard Magic" and Chernev comments that 2.f5 is "a pretty sacrifice that Black must accept". Wrong, Irving!

I don't agree with this one. At the end of your "solution" black is not just "in serious trouble"; rather he is completely lost, with only a bishop and two pawns against a rook, knight and pawn. After Kd4 white can force the capture of the c-pawn before much longer and once that happens you can put it in a tablebase which will tell you how many moves to mate.

The point of correct "solutions" to puzzles is not to find the line that the computer says is only +7.1 for the opponent as opposed to +7.3 or getting mated, but to find the hard-to-see key idea that means the position is a win in all lines. In saying Black "must accept" the sacrifice all Chernev is saying is that if black does not accept the sacrifice he will be in a clearly lost position, which is true. For a human it is not immediately obvious that the sacrifice allows white, three pawns down, to sacrifice his rook and knight and deliver checkmate with a pawn. For that reason the puzzle is both correct and beautiful even if a computer "solves" it in 0.000001 seconds and suggests that black stagger on pointlessly a rook down instead of getting mated in 7.

Have you told Chesscafe the Bogoljubov problem appears to be cooked? If not, I will.

Czentovic
24-06-2011, 09:13 PM
Black is lost if White is a strong player, but where there is life, there is hope. Who would win if White was an average club player, say ACF 1200, and Black was Anand? But after 2...Bxf5?? even an average club player should be able to win.

c-pawn? That was captured with 4. Rxc7 - maybe you mean d-pawn. Even that is hard to do, and still leaves 7 pieces; do you have a 7-piece tablebase?

This is not the first incorrect chesscafe puzzle I have discovered - I sent a comment as they asked but got no response last time, so I see no point in trying to tell them about this one.

Kevin Bonham
24-06-2011, 09:18 PM
Black is lost if White is a strong player, but where there is life, there is hope. Who would win if White was an average club player, say ACF 1200, and Black was Anand?

A 1200 club player not in time trouble should be able to win that against Anand without much difficulty. Actually Anand would probably have no better chance of saving it against any level of opponent than any 1800+ player with a good endgame would have. World class players do not have much experience of playing on in such positions because when those positions come up in games against each other, they resign.

c-pawn? That was captured with 4. Rxc7 - maybe you mean d-pawn. Even that is hard to do, and still leaves 7 pieces; do you have a 7-piece tablebase?

Yes I mean d-pawn; it is not that hard to win it by force. Actually with my comment on tablebases I had in mind that at the same time black would go after white's remaining pawn with his king and take it (after all he may as well) leaving KRN vs KBP which will be an easy tablebase win.

This is not the first incorrect chesscafe puzzle I have discovered - I sent a comment as they asked but got no response last time, so I see no point in trying to tell them about this one.

Any source that covers a study that is genuinely refuted should immediately publish the refutation. I'll have a go!

Czentovic
26-06-2011, 09:16 AM
I looked at Chernev's book to find other examples of problems in which the proposed "brilliant" solutions are found to be wrong using a chess engine,and chose to analyse the 4 puzzles he identifies as the most brilliant in his preface:

Seletsky #7 5q2/n2P1k2/2b5/8/8/3N4/4BK2/6Q1 w - - 0 0
Korolnikov #8 4N3/2PP2P1/6b1/2bN2R1/6pk/p7/K1pn4/8 w - - 0 0
Korolnikov #18 2B5/pR6/2pP1k2/8/6Kb/4p1P1/5p2/8 w - - 0 0
Libiurkin #117 1b6/4P3/1P2PN2/8/8/P1k5/P1p2P2/K5B1 w - - 0 0

They are all White to play and win. I will let you try to solve them, but I should warn you that the solutions given by Chernev are all wrong! #8 and #117 are clear draws, #18 ends up as KBR versus kbppp, with outcome unclear and #7 is indeed a win for White, but, by my analysis, takes at least 46 moves. I challenge you to solve #7 without using a tablebase.

Kevin Bonham
26-06-2011, 04:27 PM
They are all White to play and win. I will let you try to solve them, but I should warn you that the solutions given by Chernev are all wrong! #8 and #117 are clear draws, #18 ends up as KBR versus kbppp, with outcome unclear and #7 is indeed a win for White, but, by my analysis, takes at least 46 moves. I challenge you to solve #7 without using a tablebase.

I don't have any problem with #18 as I think that the KRB v KBPPP is not that hard to win. Not saying I can prove it is a win but I see no reason to even suspect it might not be. The white B controls the pawns and the white K and R are free to move up the board while the black B tends to become chained to defence so black is effectively playing a whole rook down or worse. The computer may take a while to cotton on but I don't see how black could save it without errors. I think any high level player would consider this won.

#7 I do have a problem with but it's a different problem. I suspect the issue you've found is that there's a sideline where white has to win KQ vs KBN. That's OK; that's a win in the vast majority of positions although the composer may not have been in a position to know it for sure at the time. Just because many of us would struggle to win that sideline in 50 moves without a tablebase doesn't invalidate the solution, given that that sideline is a clear win.

The problem I have with #7 is I'm not convinced the solution is unique. There are various lines (the simplest being the obvious one starting with 1.Ke1) that lead to KQBN vs either KQB or KQN. And while the composer might have assumed they were draws, the tablebases might eventually have something else to say about that (not sure these ones have been done yet as they are 7-pieces). So it may be there are multiple wins for white. Tablebases have refuted some famous problems in which incorrect assumptions about pure endings with one side a piece up were made.

#8 and #117 I agree look like draws (though it was a lot of work to be confident of the draw in #117) - a shame for the beautiful ideas behind them in both cases.

Czentovic
05-07-2011, 03:35 PM
The next puzzle from chesscafe is also wrong.

#753 1r6/7B/PP5p/7P/2p4n/k7/p1P5/K7 w - - 0 1

The solution given is: 1.a7 Rd8 2.Bd3! Rg8 3.Bf1! Re8 4.Be2 Rg8 5.Bg4 Re8 6.Be6 Rf8 7.Bf7 Rd8 8.Bd5 Rf8 9.a8Q+ Rxa8 10.Bxa8 Nf5 11.Be4! Nd4! 12.b7 c3 13.b8B! 1–0 [13.b8Q? Nxc2+ 14.Bxc2]

but Houdini 1.5 says that anything other than 6...Rd8 leads to a forced mate. White is supposed to reply 7. Bf7 to 6...Rf8, but 7. Bxc4! gives a quick forced mate. After 7. Bf7, Black can only avoid catastrophe by 8...Re8 (not the suggested Rf8), and 9. Bxc4 wins quickly, unlike 9. a8Q+.

There is no excuse for this kind of sloppiness when powerful chess engines like Houdini are freely available.

Kevin Bonham
05-07-2011, 05:05 PM
That one is quite odd. The idea is correct but the Chesscafe solution mucks it up by allowing that Bxc4 cook (adding only one extra move to the problem in the process). Whether that is the original published solution I don't know.

Here is what I think should be the correct solution:

1.a7 Rd8 2.Bd3 Rg8 3.Bf1 Re8 4.Be2 Rg8 5.Bg4 Re8 6.Be6 Rd8! [not 6...Rf8 of Chesscafe solution since 7.Bxc4 wins easily] 7.Bd5 Re8! [again not 7...Rf8 of Chesscafe solution for the same reason] 8.a8=Q+ Rxa8 9.Bxa8 Nf5 10.Be4 Nd4 [10...Nd6 11.b7 Nxb7 12.Bxb7 Kb4 13.Kxa2 Kc3 14.Be4 and white wins] 11.b7 c3 12.b8=B!! and wins

For a famous study by the same composer (which proved to very unluckily be unsound) see here: http://timkr.home.xs4all.nl/chess2/mitrofanov.htm

I emailed Chesscafe about the Boguljubov problem being unsound and they ignored my email as well. It's a pity; Chesscafe used to be a site that really cared about truth in chess and ten years or so ago it had a nice email bulletin board where you could write in corrections and have debates about world chess and so on. These days it seems to just go through the motions.