This is very nice. Who composed it?
The most beautiful puzzle that my engine can never solve
Sort:
May 4, 2023
0
#45
Aron_08 wrote:
Yes the solution was find but what says the engine?
My engine (Stockfish) says it's a draw.
drdos7 escribió:
Have a real one that engines can't solve:
Stockfish says that Qb8 is a mistake for black, causing a draw.
Aug 30, 2023
0
#49
nathanturk2 wrote:
drdos7 escribió:
Have a real one that engines can't solve:
Stockfish says that Qb8 is a mistake for black, causing a draw.
Stockfish doesn't understand this position, and you should not blindly think that Stockfish is correct without actually looking at the position. Black is LOSING this game no matter where he moves the Queen as both the Black King AND the Black Queen are TRAPPED in that corner of the board by the pawns and AND the White Knight on d8, and they can do nothing but move back and forth in that corner while the White King and Bishop scoop up the Kingside pawns, after that the White King comes down to assist in the capture of the c7 pawn and is able to push his c6 pawn on to victory, Black loses no matter where he moves the Queen in ALL scenarios, or no matter what Stockfish thinks.
Good study 👍
:tup
Mugo345 wrote:
ANIRUDH-KC wrote:
what if you put it to depth 1000
Impossible because no chess game can reach 1000 moves without being drawn.
A Chess puzzle does not have to have originated from a game. I have seen many extremely abstract positions as in 4 pawns on a file in the setup.
Aug 31, 2023
0
#54
Here is another study where Stockfish is absolutely helpless (along with other engines), and has no clue as to the solution:
White to play and WIN:
drdos7 wrote:
Have a real one that engines can't solve:
"Can't solve" seems to be a bit of an overstatement. In my test Stockfish finds all the correct moves for white but takes until a few moves in to realize white is winning. One can then back it up to the start and it will "remember" that it has a winning position (in its evaluation.)
Given a long enough think time, or fast enough hardware, I think it would evaluate white as winning from the start.
[edit] also props to the OP for finding a sort of click baitee title to boost engagement: "engine can never solve" 
drdos7 wrote:
Here is another study where Stockfish is absolutely helpless (along with other engines), and has no clue as to the solution:
White to play and WIN:
nice puzzle
Hi i got check mate and the computer told me it was the wrong move so idk if it is such a bad puzzle
i meant to write beautifull instead of bad sorry
After the 7th move ke3
Sep 1, 2023
0
#60
Lent_Barsen wrote:
drdos7 wrote:
Have a real one that engines can't solve:
"Can't solve" seems to be a bit of an overstatement. In my test Stockfish finds all the correct moves for white but takes until a few moves in to realize white is winning. One can then back it up to the start and it will "remember" that it has a winning position (in its evaluation.)
Given a long enough think time, or fast enough hardware, I think it would evaluate white as winning from the start.
[edit] also props to the OP for finding a sort of click baitee title to boost engagement: "engine can never solve"
Well of course if you feed Stockfish the moves it will eventually see that white is winning, but Stockfish WILL not see the second move (2.Qxc7+!) on any hardware, it DOES NOT see the Queen sac on the second move at the square c7. I have a 20 core machine and it doesn't find the second move even if I let it think in infinite mode for several hours.
Here is Stockfish 16 after 2 hours trying to find the second move using 20 threads and a 16GB hash table, and still no dice showing a draw score of 0.00 after 214 billion nodes searched:
what if you put it to depth 1000
Impossible because no chess game can reach 1000 moves without being drawn.