chess engine not solving a position
Sort:
CM
thechesscorner64
Jan 24, 2018
0
#1
I’d like to know if there are positions where stockfish 8 or leading chess engines do not find the best move. I am not talking about wrong evaluation like in other posts. I am talking about the engine best move suggestion after 15 minutes being inferior to another move. If you will , the engine would eventually agree with the move you would force over its first recommendation. I’d like to see suggestions from real games or studies. It is getting harder to find positions where top chess engines err in their move recommendations
Jan 24, 2018
0
#2
Here's a couple from a user Suren on YouTube posted. https://youtu.be/tnHKUtUXaCM
Jan 24, 2018
0
#3
Here's a couple from a user Suren on YouTube posted. https://youtu.be/tnHKUtUXaCM
Jan 24, 2018
0
#8
sorry for so many posts with so many links but if you check his channel out it's got good stuff
Jan 24, 2018
0
#9
sorry for so many posts with so many links but if you check his channel out it's got good stuff
Jan 25, 2018
0
#10
Some interesting examples given above from YouTube.
You'd think that an engine like Stockfish would be best at finding a forced mate, but here's a Mate-in-9 problem from this thread that Stockfish is unable to solve. After a few minutes, the engine found a M13 with 1.Rxh7, and after a couple of hours, it found the right key 1.Kb1! but got the continuation wrong, 1...Nd5 2.exd5? which let the BK escape to f5, and this mates only on move 12. I'll post the 9-move solution later, in case anyone wants to try "beating" Stockfish by solving this.
Jan 25, 2018
0
#11
I believe that certain engines solve this one and others don't. Not sure if programmers are back-testing their engines using studies rather than GMs games
Jan 25, 2018
0
#12
does your chess program solve this one ? if so how long does it take ?
Jan 25, 2018
0
#13
White to play. Takes several minutes before switching to correct move
Jan 25, 2018
0
#14
Go to 'The Secret of Chess Thread' and you'll learn all about how stupid Stockfish 8 is.
Jan 25, 2018
0
#15
Lommer Study. Fritz/Stockfish cannot solve it but Houdini/Deep Shredder can. solution attached
Give one side a big material advantage and need to break through a pawn wall via piece sacrifice to go get the weaker king. Stockfish wont do the sacrifice and will loose via 50 move. Stockfish does not want to sacrifice material if it can't win it back soon.
Find a position where White can checkmate Black in 5-6 moves, so a computer will see it but a human might not. Change it by making it Blacks turn. Place Black's queen in danger of immediate capture. Give Stockfish the black pieces. I bet Stockfish uses the extra tempo to prevent mate in 6, letting it's queen be captured. It should save the queen first, since the probability of being mated when a queen down is higher than the probability of your human opponent seeing the mate in 6.
Give stock fish a trapped king who needs two moves to make breathing room before a rook does a back rank mate the poor king only has one move though. There are 5 pieces that can be placed in the rook's path one at a time to delay checkmate, almost luring it to the mating rank. The correct move is to quietly move stuff to make breathing room for the king, yet every program out there except Jester will sack it's entire army to postpone a mate an amateur may not have noticed. This is just mates. What about this concept applied in other aspects of the game? Our goal is to win more games, not just drag out the ones we know we will lose by another 3 moved.
Jan 26, 2018
0
#19
Here's the solution to the Mate-in-9 from the post above. White's plan is to mate on e7 with the B (while avoiding stalemate from capturing the BN). Black is almost in zugzwang - most N moves allow the B mate and ...Nd5/Nc6 is answered by the waiting move Bc5!, after which the N has to move and unguard the mating square. But the WK must move to deal with ...Nc2+. 1.Kb2? fails to 1...Nxd3+! 2.Kb1 Nc5 3.Bb4 Nxe4, giving the BK an escape square on f5. So 1.Kb1! is the key, but after 1...Nd5, 2.Bc5? is premature because of 2...Nc3+!, again winning the vital e4-P. White needs to move the K again, and can make no progress if it stays on the Q-side, since the BN can continue to switch between b4 and d5 and to threaten check on c3 or c2. If 2.Kc2?, then 2...Ne3+! followed by 3...Nf5 would free the BK. So 2.Kc1! forcing 2...Nb4, then 3.Kd1! Nd5.
By this point, if you have some experience in solving composed problems, you could guess the rest of the solution by using the "Wouldn't it be nice?" principle. Since composed problems are supposed to show an interesting idea, you may guess that the nice idea here is that WK marches all the way to h1 from a1 without leaving the first rank. So not 4.Bc5? or 4.Ke2? because of the fork 4...Nc3+ again, but 4.Ke1! Nb4 5.Kf1! Nd5 6.Kg1! Nb4 7.Kh1! Nd5, and finally the WK is safe from checks, 8.Bc5! N-any 9.Bxe7 mate.