A mate in 3 that engines cannot solve ? WOW!

[Robert Flesher]

You must see the mate in 3 to be correct, Rh1 is the brilliant move that leads to mate in 3.

[Robert Flesher]

No they fail to see the mate is 3, they all see the a mate solution.

[Dann Corbit]

No they fail to see the mate is 3, they all see the a mate solution.

For a chess engine, this is not a wrong solution, since it has the same value as any other mate (game is won).

For a mate solver, these would be wrong solutions, because they must find the minimum mate.

[kgburcham]

Not true at all.
Every engine I tried found the mate, no problem, non of which is a brilliant move.

kgburcham

[yanquis1972]

i dont know how you managed that, as zappa & rybka display a #4 (like dann, i dont believe this to be 'wrong' in this case).

[zullil]

[d] 8/1n3Np1/1N4Q1/1bkP4/p1p2p2/P1P2R2/3P2PK/B2R4 w - - 0 0

I wonder how many GMs could find the move Rh1.

If told "Mate in 3" I bet every GM would find it.

[michiguel]

[d] 8/1n3Np1/1N4Q1/1bkP4/p1p2p2/P1P2R2/3P2PK/B2R4 w - - 0 0

Gaviota 0.68.8 finds mate in 3 in 5 plies, 0.4 seconds

Code: Select all

        16   1       0.0   +33.62  1.Nxc4 Bxc4 2.Rxf4 Bxd5 3.Qxg7
        29   1       0.0   +41.01  1.Rxf4
       386   2       0.0   +40.98  1.Rxf4 Be8
      1728   2       0.0   +41.48  1.d4+ cxd3 2.Rxf4 Be8 3.Rxd3
      2511   2       0.0   +45.64  1.Re1 Be8 2.Rxe8
      2791   3       0.0   +45.64  1.Re1 Be8 2.Rxe8
      7904   4       0.1   +48.27  1.Re1 Be8 2.Rxe8 Na5 3.Rxf4
     29699   5       0.2   +Mat_4  1.Re1 Be8 2.Nc8 Nd6 3.Qxd6+ Kb5 4.Qb6#
     87664   5       0.4   +Mat_3  1.Rh1 Be8 2.Qb1 Bxf7 3.Qb4#

Miguel

[Uri Blass]

If you give chess programs problems that they are not designed to solve them then you can expect them to fail

Chess engines also fail in solving selfmate problems.

See the following problem

[D]7k/q6P/7K/6P1/8/8/8/8 b - - 0 1

black force white to mate in 2 moves

The solution is Qf7 g6 Qf5 g7 mate but most chess engines are not going to find it.

If you claim that it is not fair because they have no way to understand the question then the same is for normal mate and chess engines do not understand that they need to find the fastest mate when you give them a position when they are winning.

It is easy to fix it by having an algorithm to find the shortest mate immediately after finding a mate but it is not going to help to increase the playing strength of the engines.

Uri

[Gian-Carlo Pascutto]

If you give chess programs problems that they are not designed to solve them then you can expect them to fail
It is easy to fix it by having an algorithm to find the shortest mate immediately after finding a mate but it is not going to help to increase the playing strength of the engines.

Surely I'm not the only one whose program gives higher scores to shorter mates. And the search finds the move with the highest evaluation, i.e. eventually the shortest mate.

If we didn't do this, we wouldn't be able to mate the opponents.

So I think this argument is wrong, i.e. chess engines are already capable of finding the shortest mates and its part of their goal.

[Dann Corbit]

If you give chess programs problems that they are not designed to solve them then you can expect them to fail
It is easy to fix it by having an algorithm to find the shortest mate immediately after finding a mate but it is not going to help to increase the playing strength of the engines.

Surely I'm not the only one whose program gives higher scores to shorter mates. And the search finds the move with the highest evaluation, i.e. eventually the shortest mate.

If we didn't do this, we wouldn't be able to mate the opponents.

So I think this argument is wrong, i.e. chess engines are already capable of finding the shortest mates and its part of their goal.

Of course it can still be filtered out by search algorithm so that the shortest mate is not found in a reasonable time.

I do wish that chess engine programmers (in general) would do the following two things in their searches:
1. If you have found a mate in N and you have exhausted ply N*2-1, then stop searching and say "I'M DONE!"

Reasoning:
If you have already found a mate in N and you have exhausted ply N*2-1, then there is literally no way to improve the score so any further searching is a total waste of CPU time. If your engine may have pruned out shallower mates, then widen the score and research at depth N*2-1 (or do whatever else is needed to make your engine exhaust this ply) but there is no need to search deeper.

2. If you found a mate in N and have NOT exhausted ply N*2-1, then stop searching only after you have exhausted ply N*2-1.

If you have found a mate in 11 and I have given you two days to search and you stop after finding a mate in 22 at 12 minutes, I find that annoying.

I have seen many engines that do not perform in the above manner, and I find it very annoying.