Stockfish 8 Suffocates! No Way To Solve This Chess Puzzle!

Stockfish 8 Suffocates! No Way To Solve This Chess Puzzle!


Hello chess lovers. Here is a very complex complex chess puzzle by Russian chess composer Novikov. This was first published in 1930 in Russian chess magazine “64”. It’s white to move and win the game. You can pause the video and try to find the solution The thing is that I have inserted this puzzle more than 20 minutes ago and still the engine is unable, Stockfish 8 is unable to find the solution. You can see that it evaluates the position as equals and goes on the wrong direction thinking that the best move is Ra7. By the way if you can’t find the solution I can give you a hint. You have to trap this future Queen so the best move in this position … just let’s have a look what engine thinks; Rd7 Rg3 Ra7. All are wrong and evaluates that those continuations are equal. But in this position the most fantastic looking move is Rh7 check Kh7 and now comes Be4 check Kg7 and now comes Kc1 Allowing black to promote to queen and Bb1 and it turnes out that black queen is trapped just no way to go and one of this pawns is being promoted to a queen Well you may ask how is white going to win after promoting to a queen but that’s very simple you can exchange queens on a2 and that will be an easy win for white. Let’s go back, what if black won’t take on h7 for example Kg8 then comes Bd5 check and now what if black will just if you take on h 7 this is the same Bd4 check and we are going on the same line. Kf8 can you find the best move for Rh1 allowing black to promote a queen again , a1Q Kc2 and it turns out that black queen is again trapped. Queen a2 and this is a simple win, you will promote your pawn to a queen white is winning it will be also very interesting to know what other puzzles do you know which can blow up chess engines? Your comments, please. Good Luck

48 thoughts on “Stockfish 8 Suffocates! No Way To Solve This Chess Puzzle!

  1. Of course, on my Laptop Stockfish 8 solved the puzzle, but it took him more than 8 minutes to find the correct starting move Rh7+. But after a few more minutes, Stockfish increases the advantage for white dramatically (over +120). Indeed, Crafty was quicker with Rh7+, but failed to analyze it to the end. I did not check it with "Stockfish Modern" and i also did not check with my Desktop PC (which is more powerfull than the Laptop). Anyway, Stockfish solved it on the Laptop.

    Maybe i will check it tomorrow on different PCs and chess engines, for today, it is too late, i have to go into bed….

  2. I'm getting better at these computer stumpers. Knowing that I could trap the Q, I found the Rh7+ move!

  3. My Houdini 5.0.1 found the solution in under 10 seconds, with around 4 million tablebase hits. I prefer Houdini for the tactics. and I challenge everyone who has houdini to load it up, since it will find it in an extremely short time.

  4. i learn alot from you i believe in time humans will develope a method for problem solving that will exceed the calculation and focus on systems of logic instead. chess is helping us reach that goal i think

  5. This is what makes chess so great.. it is taught with rules.. pieces have values.. knights on rim are dim.. bla bla.. there is only what works and what does not.. break rules and find winning moves! Kids learning to play can think for themselves.. not oh well my opponent castled.. so should I…

  6. have you tried the one Fischer could not solve? white… B on c1..Q on d1..K onE1.. B on F1. Black.. K on. E4. White mates in 3. Try it.. solve with NO computer!!!

  7. My own engine is maybe 800 Elo weaker than Stockfish 8, and it finds the right move in 12 seconds, then gets a high score in 21 seconds. This doesn't mean there is anything wrong with Stockfish: It just means that they are optimizing for playing strength, and solving problems like this one detracts from that goal.

    Stockfish probably has some pruning heuristic that makes it miss this problem. That pruning heuristic might allow the engine to search deeper in regular positions, which means that the engine is stronger as a result. As long as the exceptions are rare in real games, having the pruning in place is the correct decision.

  8. What happens if the pawn promotes to knight, escapes (with a check) and then proceeds to pick up those pawns and eventually sacrificing itself, drawing the game?

  9. Interestingly Houdini 5 finds it in less than 10 seconds but Stockfish does not. If you make the correct Rh7+ for Stockfish and restart it if finds it instantly. (i.e. it adds it memory). Seems to suggest an improvement for Stockfish. Stockfish should consider all possible 31 moves as played and go from there. My guess is that is what Houdini is doing. Thanks for posting these Suren. Peace

  10. i think it's important to note that rd1 is a vital part of the queening process at the end to protect the bishop after pushing the c pawn.

  11. Latest asmFish or Stockfish dev can easily solve this position.

    7k/6R1/8/8/2P2P2/P7/pP6/3K3B w – – 0 1

    Analysis by asmFishW_2017-05-22_bmi2:

    1.Rh7+ Kg8 2.Bd5+ Kxh7 3.Be4+ Kh6 4.Kc1 a1Q+ 5.Bb1 Kg7 6.c5 Kf7 7.f5 Ke7 8.c6 Kd6 9.f6 Ke6 10.c7 Kd7 11.f7 Kxc7 12.f8Q Kd7 13.Qc5 Kd8 14.Qd6+ Kc8 15.Qc6+ Kb8 16.Qc2 Ka7 17.Kd2 Kb7 18.Qe4+ Kc8 19.Qe8+ Kb7 20.Qe7+ Kb6 21.Qd8+ Kc5 22.b4+ Kb5 23.Qd5+ Kb6 24.Qc5+ Kb7 25.Be4+ Kb8 26.Qb6+ Kc8 27.Bf5#

    White mates: +- (#27) Depth: 65/54 00:00:14 246MN
    (, 30.05.2017)

  12. OMG! It is clearly a draw after black promotes to a knight, you just checked some random sidelines and now you think you are better than an engine elo 3000<. GOSH

  13. we humans are at an advantage over engines, as we have a mental concept of what a chess problem is. usually we will search for "unusual" moves, like major sacrifices, strange tempo moves, minor promotions etc. engines can, essentially, not see any difference between a regular position and a chess problem. would you agree? by the way, thanks for your posts, which I enjoy a lot!

  14. My Houdini 4 Pro 960 Aquarium Extreme Edition solved it completely in less than 4 minutes. I couldn't find a problem which Houdini can't solve, but I've made a draw against Houdini (Elo=3280) in spite of not being a master. Thank You for the nice Puzzle!

  15. Stockfish 9-x64 found 1.Rh7 at 11 secs at depth 46/56+ with a score of +0.07. It had White winning at depth 46/60+ at 14 secs with a score of +1.81. White had a won game at depth 46/68+ at 25 secs with a score of +5.66. At depth 48/57 at 1m 9s SF9 displayed M32 but finished in M25!

  16. This is one of the most impressive puzzles I've ever seen! It seems so simple and still it's mind blowing!

  17. What sequence of moves would permit the black pawn to occupy A2 while white pawns occupy A3 and B2? Were pieces sacrificed on B3 and A2 to the [initial position] C7 pawn?

  18. at 1:39 a funny side line is if black promotes to a knight with …a1=N , then Bc2 still traps the promoted piece 🙂

Leave a Reply

Your email address will not be published. Required fields are marked *