The problem computers can't solve

Back in mid-2001, a member of the CCC (computer chess club) proposed a remarkable puzzle that was specifically designed to be difficult to solve for computers. At that time, it would take most programs between minutes and hours to solve. It is a mate in 8. How fast can your program find the solution(s)?

Houdini calls this a mate in 10.

This puzzle is ridiculous.  Seriously.  :)

Sorry, I take that back.  Houdini called it a mate in 9.  My bad.

It took Houdini 1.5 about 4-5 minutes to solve this puzzle.  Nice puzzle!  It's just kind of unrealistic.  I don't think that could ever happen in a real game.

What's interesting is that Houdini is wrong. :P

It is most certainly a mate in 8, believe me.

Okay.

That would have been a horrible game.  If that many promotions were allowed by both sides, the game would have been horrible.  And if that were the case, checkmate would have been delivered by one side far before now.  That's just my opinion

Houdini 2.0c finds a mate in 8 in 20 seconds on my system:

4 [+66.29]  1. dxe7+ Qxe7 2. Qxg8+ Bxg8 3. Qxa8+ Ncxa8 4. Qhxe5 Qxe5 5. Qxe5  (10.99)

5 [+65.20]  1. dxe7+ Rxe7 2. Qxg8+ Bxg8 3. Qxa8+ Nbxa8 4. Bxd7 Rxc4 5. Rxc4 Rxe4 6. Qxe4  (11.01)

5 [+70.65]  1. Qxg8+ Nexg8 2. dxc7+ Qxc7 3. Ne6+ Rxe6 4. Bxc7+ Kxc7 5. Qxc5+ Kb8 6. Qxa8+ Kxa8 7. Ra2+ Kb7  (11.03)

7 [+75.42]  1. Qxg8+ Bxg8 2. Qxa8+ Nbxa8 3. Qexe5 Rxc4 4. dxe7+ Kc8 5. Bxd7+ Kb7 6. e8=Q Nc3+ 7. Rxc3 Ncxe8 8. Qxh6  (11.09)

8 [+75.01]  1. Qxg8+ Bxg8 2. Qxa8+ Nbxa8 3. Qexe5 Rxc4 4. dxe7+ Kc8 5. Bxd7+ Kb7 6. e8=Q Ncxe8 7. Bc8+ Kxc8 8. Qxc4+ Nac7 9. Qhf5+ Kb8 10. Qxh6 Nxf4 11. Qcxf4  (11.14)

8 [+80.04]  1. dxe7+ Kc8 2. Qxg8+ Bxg8 3. e8=Q+ Rxe8 4. Qexe8+ Nfxe8 5. Qxa8+ Nbxa8 6. Bxd7+ Kb7 7. Qxc5 Bf8 8. Qxf8  (11.35)

9 [M11]  1. dxe7+ Kc8 2. Qxg8+ Bxg8 3. e8=Q+ Nfxe8 4. Qca6+ Qb7 5. Qxe8+ Nxe8 6. Rxc5+ Kb8 7. Qd8+ Qxd8 8. Qexe5+ Nd6 9. Qa7+ Qxa7 10. Nc6+ Ka8 11. Qxa7#  (11.95)

10 [M10]  1. dxe7+ Kc8 2. Qxg8+ Bxg8 3. e8=Q+ Nfxe8 4. Qca6+ Qb7 5. Nfxd5 Bxd5 6. Qdxd5 Nbxd5 7. Bxd7+ Kxd7 8. Qeg4+ Re6 9. Qgxe6+ Nxe6 10. Qxe6#  (12.48)

10 [M9]  1. dxc7+ Qxc7 2. Qxa8+ Nxa8 3. Qexd5+ Rexd5 4. Qcxd5+ Rxd5 5. Qdxd5+ Qxd5 6. Qdxd5+ Nfxd5 7. Qxd5+ Ke8 8. Qxe7+ Qxe7 9. Rc8#  (12.81)

11 [M9]  1. dxc7+ Rxc7 2. Qxc7+ Qxc7 3. Qxa8+ Nxa8 4. Qexd5+ Rxd5 5. Qdxd5+ Qxd5 6. Qdxd5+ Nfxd5 7. Qxd5+ Ke8 8. Qxe7+ Qxe7 9. Rc8#  (13.95)

12 [+327.42]  1. Qxa8+ Nbxa8 2. Qgxc5 Qg1+ 3. Bxg1 Nc6 4. Qxc6 Ne3+ 5. Bxe3 Bg5 6. dxc7+ Nxc7 7. Qcxd7+ Nxd7  (17.28)

14 [M8]  1. Qxa8+ Nbxa8 2. Qgxc5 Qg1+ 3. Bxg1 Nc3+ 4. Qcxc3 Rxc5 5. dxc7+ Kxc7 6. Qhxc5+ Nc6 7. Qexc6+ Qxc6 8. Qa7#  (20.27)

We've come a long way since 2001.

My Chessmaster 9000 found 2 solutions, eventually...

TimeDepthScorePositionsMoves

41:301/7Mate084180702491.Qxa8+ Nbxa8 2.dxe7+ Kc8 3.Qgxc5

Qg1+ 4.Bxg1 Nc3+ 5.Rxc3 Rxc5 6.e8=Q+

Ncxe8 7.Qhxc5+ Kb8 8.Na6#

41:481/7Mate084209850031.dxe7+ Kc8 2.Qxa8+ Nbxa8 3.Qgxc5

Qg1+ 4.Bxg1 Nc3+ 5.Rxc3 Rxc5 6.e8=Q+

Ncxe8 7.Qhxc5+ Kb8 8.Na6#

it seems like black promoted at least 3 pawns into knights - why on earth would black have done that? 

It makes no sense to me....

It's a theoretical problem. Not a "why would black promote 3 pawns into knights" problem.

Btw, my Fritz 11 SE is still running this problem and has yet to even give an evaluation as to who stands better. It just calculates all the forcing moves. It's been doing that for maybe 5 minutes now. I'm getting a new kibitzer.

Chess Engines are definately much better than they were in 2001. It took my Stockfish only 38 seconds to find the mate in 8. Great puzzle.

I agree with DavidMertz.  There are two different mates, both in 8 moves.

[M8]  1.dxe7+ Kc8 2.Qxa8+ Nbxa8 3.Qcxc5 Qxg1+ 4.Bxg1 Nc3+ 5.Rxc3 Rxc5 6.e8=R+ Ncxe8 7.Qxc5+ Nac7 8.Qa8#  (1.57)

[M8]  1.Qxa8 Nbxa8 2.dxe7+ Rxe7 3.Qxe7+ Qxe7 4.Qcxc5 Qxg1+ 5.Bxg1 Qxe2+ 6.Qhxe2 Bf5 7.Qdxd5+ Nxd5 8.Qee7#  (1.28)

Black can't play 2.e8=Q.  If you mean what if White plays it, after 1.dxe7+, then it's not a mate in 8.  (It would be a mate in 9.)  That's why White doesn't play that move.