The puzzle that you, Carlsen nor Stockfish can solve

stockfish 9 says white is better

I suggest you analyze these puzzles, it's challenging http://ourexcitingpuzzles.blogspot.com/

I know the thread is old, still did some analyzing with the Engines i own currently with their standard settings. Always impressed with Deep HIARCS 14 and its positional eval...may not have gotten a update in some time but still gives probably the most insight of all engines.

1.Sg4-f6+
-+ (-1.69) Depth: 25/44 00:00:19 105MN

Positive Eval first about half a minute later:

1.Sg4-f6+
+/- (1.51 ++) Depth: 25/45 00:00:43 227MN

Some more analysis later on:
1.Sg4-f6+ Kh7-g7 2.Sf6-h5+ Kg7-g6 3.Ld1-c2+ Kg6xh5 4.d7-d8D Kh5-g4 5.Lc2-d1+ Kg4xg3 6.Dd8-e8 c5-c4+ 7.Kd6-c6 Kg3-f4 8.De8-g8 Sg5-e4 9.Dg8xc4 Sa6-c5 10.Kc6-d5 Lb4-a3 11.Ld1-c2 e3-e2 12.Dc4xe2 Se4-f6+ 13.Kd5-d4 Sc5-d7 14.De2-f2+ Kf4-g4 15.Kd4xc3 La3-d6 16.Df2-f5+ Kg4-h4 17.Lc2-a4 Ld6-e5+ 18.Kc3-d2 h6-h5 19.La4xd7
+- (3.93) Tiefe: 32/59 00:14:41 5145MN

Deep Junior 13.8 Yokohama

finds it instantly but throws it also instantly away for d7-d8D.

After some more analysis it quickly corrects down the black outlook but still shows a slight advantage.

1.Sg4-f6+ Kh7-g7
+/- (1.02) Tiefe: 9 00:00:00
1.Sg4-f6+ Kh7-g7
+/= (0.58) Tiefe: 12 00:00:00
1.Sg4-f6+ Kh7-g7 2.Sf6-h5+ Kg7-g6 3.Sh5-f4+ Kg6-f5 4.Kd6-c6 Sa6-b8+ 5.Kc6-c7 Sb8xd7 6.Kc7xd7 Kf5-e5 7.Kd7-c6 Ke5-e4 8.Kc6-b5 Ke4-d4 9.Ld1-c2 h6-h5
-+ (-3.56) Tiefe: 15 00:00:01 12463kN
1.d7-d8D Sg5-f7+ 2.Kd6-d7 Sf7xd8 3.Kd7xd8 Lb4-a5+ 4.Kd8-c8 e3-e2 5.Ld1xe2 Sa6-b4 6.Sg4-e3 c5-c4 7.Le2xc4 La5-b6 8.Se3-d5 Sb4xd5 9.Lc4xd5 c3-c2
-+ (-2.32) Tiefe: 15 00:00:01 12465kN
1.Kd6-c6 Sa6-b8+ 2.Kc6-c7 Sb8xd7 3.Kc7xd7
-+ (-2.19) Tiefe: 15 00:00:01 15455kN
1.Kd6-c6 Sg5-f7 2.Sg4xe3 Sf7-d8+ 3.Kc6-b6 Sa6-b8 4.Kb6-c7 Sb8-c6 5.Ld1-a4 Lb4-a5+ 6.Kc7-d6 Sc6-d4 7.Kd6xc5 Sd4-e2 8.La4-c2+ Kh7-g8 9.g3-g4
-+ (-1.93) Tiefe: 18 00:00:02 25393kN
1.d7-d8D
-+ (-1.75) Tiefe: 18 00:00:02 27754kN
1.d7-d8D Sg5-f7+ 2.Kd6-d7 Sf7xd8 3.Kd7xd8 Lb4-a5+ 4.Kd8-c8 e3-e2 5.Ld1xe2 Sa6-b4 6.Sg4-e3 Kh7-g8 7.Le2-d1 Kg8-f7 8.Se3-c4 c3-c2 9.Ld1xc2 Sb4xc2 10.Sc4xa5
-/+ (-1.35) Tiefe: 18 00:00:03 36582kN
1.d7-d8D Sg5-f7+ 2.Kd6-d7 Sf7xd8 3.Kd7xd8 Lb4-a5+ 4.Kd8-d7 e3-e2 5.Ld1xe2 Sa6-b4 6.Sg4-e3 Kh7-g7 7.Le2-d1 Kg7-g8 8.Kd7-c8
-/+ (-1.31) Tiefe: 22 00:00:10 166MN
1.d7-d8D Sg5-f7+ 2.Kd6-d7 Sf7xd8 3.Kd7xd8 Lb4-a5+ 4.Kd8-d7 e3-e2 5.Ld1xe2 Sa6-b4 6.Sg4-e3 Kh7-g7 7.Le2-d1 La5-b6 8.Kd7-d6 Lb6-a7 9.Se3-c2 Sb4-d3 10.Ld1-e2
-/+ (-0.92) Tiefe: 30 00:10:56 9844MN

Deep Shredder 13 also finds it instantly:

1.Sg4-f6+ Kh7-g7 2.Sf6-h5+ Kg7-g6 3.Ld1-c2+ Kg6xh5 4.d7-d8D Sg5-f7+ 5.Kd6-e6 Sf7-g5+

  +/= (0.50 ++)  Tiefe: 23/41   00:00:01  16450kN

1.Sg4-f6+ Kh7-g7 2.Sf6-h5+ Kg7-g6 3.Ld1-c2+ Kg6xh5 4.d7-d8D Kh5-g4 5.Lc2-d1+ Kg4xg3 6.Dd8-e8 c5-c4+ 7.Kd6-c6 Kg3-f2 8.De8-h5 Sa6-b8+ 9.Kc6-d5 Lb4-a5 10.Kd5xc4 Kf2-g2

  +/- (1.27 ++)  Tiefe: 23/41   00:00:01  17503kN

1.Sg4-f6+ Kh7-g7 2.Sf6-h5+ Kg7-g6 3.Ld1-c2+ Kg6xh5 4.d7-d8D Kh5-g4 5.Lc2-d1+ Kg4xg3 6.Dd8-e8 c5-c4+ 7.Kd6-c6 Kg3-f2 8.De8-h5 Sa6-b8+ 9.Kc6-d5 Kf2-g2

  +- (2.25 ++)  Tiefe: 23/41   00:00:02  19076kN

1.Sg4-f6+ Kh7-g7 2.Sf6-h5+ Kg7-g6 3.Ld1-c2+ Kg6xh5 4.d7-d8D Kh5-g4 5.Lc2-d1+ Kg4xg3 6.Dd8-e8 c5-c4+ 7.Kd6-c6 Kg3-f2 8.De8-h5 Sa6-b8+ 9.Kc6-d5 Kf2-g1 10.Dh5-e2 Sb8-a6 11.De2xe3+ Kg1-h1 12.Kd5xc4 Kh1-h2

  +- (3.50 ++)  Tiefe: 23/53   00:00:02  24635kN

1.Sg4-f6+ Kh7-g7 2.Sf6-h5+ Kg7-g6 3.Ld1-c2+ Kg6xh5 4.d7-d8D Kh5-g4 5.Lc2-d1+ Kg4xg3 6.Dd8-e8 c5-c4+ 7.Kd6-c6 Kg3-f4 8.De8-g8 Sg5-e4 9.Dg8xc4 Sa6-c5 10.Kc6-d5 Lb4-a3 11.Ld1-c2 e3-e2 12.Dc4xe2 Se4-f6+ 13.Kd5-d4 Sf6-g4 14.Kd4xc3 h6-h5 15.De2-f1+ Kf4-e5 16.Df1-a1 Sc5-e4+ 17.Lc2xe4 La3-c5 18.Kc3-c4+ Ke5xe4 19.Da1-h1+ Ke4-f4 20.Kc4xc5 Kf4-g5 21.Dh1-d5+ Kg5-g6 22.Dd5-e4+ Kg6-g5 23.De4-e7+ Sg4-f6 24.Kc5-d4 Kg5-g6

  +- (6.67 ++)  Tiefe: 36/79   00:01:51  1693MN

Houdini 6 with Standard Settings didn’t find anything after 5 minutes, so i switched to Tactical Mode and Analysis Contempt. But no matter how i fiddled with it, i came up with nothing.

Komodo 12.1.1 finds it instantly but then throws it away as well for the rest of the analysis. Stockfish 10 found nothing as well.

I guess due to the aggressive pruning of modern high-end engines they seem to be less suitable for analysis. It also shows pretty clear to me that a newer engine with much higher playing ELO doesn’t necessarily contribute more to your learning, analyzing and playing development.

There is a fundamental reason why chess playing engines may fail to solve composed problems and endgames. The challenge for composers is to create positions which are only solvable by unusual, weird and paradoxical moves and plans - measured by the experience of millions of chessgame players. General purpose engines are designed to win chess games and therefore favor orthodox moves and plans in the search order. In extreme cases (the current problem is mildly extreme) the pile up of unusual choices is so massive that the engine will only succeed by a brute force tree walk. And that is still beyond the capacity of modern engines for the longer problems. See Rocky64's fascinating article on the subject https://www.chess.com/blog/Rocky64/stockfish-and-a-forced-mate-problem-it-cannot-solve

By the way, I think this even applies to Alphazero, since it learns by playing chess games. I see no way it can train itself by "inventing" interesting problems and endgames and then learn from solving them.

Amazing

I would like to find out what various engines do with the position below. Fritz 15 can find a mate in 8 in a few minutes, but doesn't seem able to find the mate in 7 at all.

White to play and mate in 7

That's the one Rocky64 treats in the other article (for several engines): https://www.chess.com/blog/Rocky64/stockfish-and-another-forced-mate-problem-it-cannot-solve

At closer inspection, I only see him running it on Popeye (and Stockfish of course).

Thanks, Arisktotle. Maybe we can get some results with other engines. I tried to turn off pruning in Fritz 15 and couldn't find a way to do it. There's probably not one.

Good luck n9531l1! I won't be of much use in this project as I have no access to any of those wonderful engines. Except the one that chess.com so kindly provides to all of us. How do I survive?

I think you survive because those wonderful engines are so little help with retro problems.

Bad joke. 

When the king gets close to the other king after the fork, the bishop would start manuvering and Eventually find a safe square to mate the king.

Fritz 15 solves #58 in 8 seconds. How long does it take Komodo to solve #49?

#58 is not the original study, #1 is and it is one move longer. Don't expect the outcome to be very different as white does not have many interesting options beyond queening.

10 Years ago, #1 was considered almost unsolvable by engines but they made great strides forward. It does not surprise me considering that the knight is really an essentially different piece from the queen giving individual significance to its variations. I suspect it is much harder to solve a challenge with several stalemate motivated bishop and rook promotions - both contained by the queen function - though I doubt such a problem exists in the orthodox domain. A friend of mine made some short direct mate problems with 2, may be 3 bishop promotions but that won't be enough to throw sand in the gears of a state-of-the-art chess engine.

I am personally proud having designed an endgame study that gets the chess.com stockfish engine sweat for a minute or so. Made me sweat for two years to create it.

Hello.. I've solved the puzzle.. with alot of difficulty.enjoy.. uhm and please get me another one to do. It was fun.. looks like a fabonacci shell movement

Have you solved #49?