Here's 2 I posted in another thread on this topic:
This is just a basic endgame study and stockfish has trouble solving it.
And a simple mate in 67
Mates that are difficult for engines
Sort:
Apr 23, 2023
0
#23
EndgameStudier wrote:
Here's 2 I posted in another thread on this topic:
This is just a basic endgame study and stockfish has trouble solving it.
And a simple mate in 67
Actually the latest development version of Stockfish solves the mate in 67 now, even faster than Crystal does:
I'll try your other one out here and see if it can solve that WITHOUT tablebases because the Tablebases would solve it instantly.
Ok, even without table bases the latest development version of Stockfish solves the mate in 18 instantly, but says it is a mate in 17 instead of 18.
I love this thread so much. Amazing stuff
drdos7 wrote:
Here is a puzzle known as "Plaskett's Puzzle" that is very well known for computer chess, it is actually a chess endgame study created by the Dutch endgame composer Gijs van Breukelen around 1970, it was origanally supposed to be a "white to play and win study", but with the advent of computer engines I found out that there is actually a forced mate in this position.
I was one of the first to see this creation in the nineteenseventies not knowing it would be famous one day. Gijs and I were members of the same chess club in Utrecht. He recently passed away.
Btw, all win studies are forced mates but generally people don't care to find the checkmates. They only become special when the fastest win also turns out to be a sound directmate like in Bondarenko's well known #21 with 2Q vs 6R+P. Some studies are converted to directmates to avoid duals with longer forced checkmates.
Apr 24, 2023
0
#26
Arisktotle wrote:
drdos7 wrote:
Here is a puzzle known as "Plaskett's Puzzle" that is very well known for computer chess, it is actually a chess endgame study created by the Dutch endgame composer Gijs van Breukelen around 1970, it was origanally supposed to be a "white to play and win study", but with the advent of computer engines I found out that there is actually a forced mate in this position.
I was one of the first to see this creation in the nineteenseventies not knowing it would be famous one day. Gijs and I were members of the same chess club in Utrecht. He recently passed away.
Btw, all win studies are forced mates but generally people don't care to find the checkmates. They only become special when the fastest win also turns out to be a sound directmate like in Bondarenko's well known #21 with 2Q vs 6R+P. Some studies are converted to directmates to avoid duals with longer forced checkmates.
I kind of figured there might be a "Dutch connection" there 
, also of course I already knew that all win studies are directmates, but this one just happened to be within reach of the computers horizon within the last 3-4 years, and finally I knew of Gijs's recent passing from a chessbase article, or it may have been from one of the forums I visit.
I believe you are talking about Bondarenko's study here:
Which you commented on. 
Btw, I have no particular preference for either endgame studies or directmates. If a composition fits both categories it is up to author to decide where to place it. And still anyone else can make a different choice.
Apr 24, 2023
0
#28
Arisktotle wrote:
Btw, I have no particular preference for either endgame studies or directmates. If a composition fits both categories it is up to author to decide where to place it. And still anyone else can make a different choice.
Btw, here is Rybka 2.3.2a (an engine from 2007) finding the win in Gijs's study even faster than Threadripper, only taking 32 seconds. Ironically I think this engine won't run on a Threadripper cpu...it might be too old, of course I could be wrong but I know that some older engines won't run a Threadripper.
Interesting. I always wondered whether or not "finding" a checkmate during analysis would guarantee that further searches could not "undo" that result. I recall having seen an engine reporting "+M16", then retract it later and replace it with something less definite (don't recall exactly what). Such an observation would place all mate-evaluations into a speculative category where we don't want it!
Apr 24, 2023
0
#30
Arisktotle wrote:
Interesting. I always wondered whether or not "finding" a checkmate during analysis would guarantee that further searches could not "undo" that result. I recall having seen an engine reporting "+M16", then retract it later and replace it with something less definite (don't recall exactly what). Such an observation would place all mate-evaluations into a speculative category where we don't want it!
That happens sometimes, but what I do is let it show the mate three times and that way I know it's for sure. The only problem I run into is that the engines don't always find the shortest mate.
drdos7 wrote:
The only problem I run into is that the engines don't always find the shortest mate.
It's their human touch. They hate it when we view them as perfection nerds (The world according to Sheldon Cooper).
Apr 24, 2023
0
#33
Good luck engines, white to move and mate in 26:
If you have a mate finder engine it might find this one, but I doubt it 
.
.
Apr 24, 2023
0
#34
Come On Engines (and Threadripper), I'm Trolling You! white to move and Mate in 12:
Make me look like the fool I really am!
P.S. Crystal solves this one Btw in just under 2 minutes on my comp.
Here is my mate finder "ChestUCI 5.2" finding the solution in a fraction of a second:
Apr 25, 2023
0
#35
It's Blatantly obvious what must be done to deliver the mate in this one, the white King CANNOT move from where he is at or he'll free up Black's pieces, so the White Queen must deliver the checkmate herself by forcing the black King to the only square (a3) that she can mate him at from the a5 square. The question is can engines figure out what humans can do?
White to move and mate in 23:
Good luck engines.You can solve this one.
Apr 25, 2023
0
#36
EndgameStudier wrote:
The black king could have made a few moves toward the center instead toward the edge, this doesn't look as forced as the other ones. Would be nice to see some of the variations with this one.
Ok, I appended the diagram and added a variation with black playing 1...Kc6 which is an even quicker mate. If black varies from the moves it results in a quicker mate.
There were multiple points at which the king could have made other moves. It may still be forced, I didn't analyze it completely.
Apr 25, 2023
0
#38
EndgameStudier wrote:
There were multiple points at which the king could have made other moves. It may still be forced, I didn't analyze it completely.
If you want to just post where you would want to change the move and I'll refute each one, but it will have to wait until I get home from work.
Btw, were you Able to get the Stockfish development version?
Apr 25, 2023
0
#39
Ok, I'm going to show some mercy on the engines here, with a mate in 10 position from an actual game, here it is:
White to move and mate in 10, remember THIS IS A MATE IN 10 from an obviously winning position, so you don't get ANY credit unless you can demonstrate the mate in 10:
If your computer is fast enough, Stockfish should solve this in a somewhat reasonable amount of time like less than 10 minutes.
Stockfish see a mate! And finds the solution.
Ok, post the mate when you get it, ESPECIALLY if it is less than 32 moves!
It is not going to show the exact mate. As it is seeing the mate in TB. And the DTM will prevent that number. Showing as a mate in x.
I bet you never thought to see if that one was a forced mate, I accidentally discovered that it was after letting my comp think about it over night with Tablebases disabled.