Are there engines to solve win material problems

Discussion of anything and everything relating to chess playing software and machines.

Moderators: Harvey Williamson, Dann Corbit, hgm

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
Uri Blass
Posts: 8828
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Are there engines to solve win material problems

Post by Uri Blass » Thu Oct 01, 2020 6:55 am

I know that engines solve mate in x plies but I do not know if engines can solve win material problem in x plies.

For winning problem there is a problem that is not in mate problems that maybe you win material but later lose it
so I think that we need definition of win material in x plies and verfication in y plies.

I will give an example



In the following position white can win a knight if you search 1 ply forward with no verfications by 4.Bxc6

If you search 1 ply forward with verfication of 3 plies white win a pawn by 4.Bxc6 dxc6 5.Nxe5
If you search 1 ply forward with verification of 7 plies then I think that white does not win material because I
think that white cannot keep the advantage that he got after 1 ply both after 3 plies,5 plies and 7 plies.

4.Bxc6 dxc6 5.Nxe5 Qd4 6.Nf3 is still enough to keep the 1 pawn advantage after both 3 plies and 5 plies but white has no way to keep 1 pawn advantage also after 7 plies.

If you search 7 plies forward without verification of more than 7 plies then again I suspect that white win a pawn because black cannot prevent a stupid sacrifice in the last ply that is refuted only if there is a verification at bigger depth.


Basically I would like an engine to get 2 numbers:
1)search depth in ply that I denote x
2)verfication depth in ply that I denote y.
The engine should find all the paths that get the biggest material gain in x plies when you also keep the advantage in x+2k plies for every case when k is positive and x+2k<=y

Note that the number can be also negative and the result can be minus 1 if the best you can do is to lose a pawn.

I think that an engine like that can be useful for generating chess problems for humans that are simply about winning material or not losing material .
It also may be interesting to see how many winning material moves based on the limited depth that computers can search are simply bad moves based on stockfish to learn about cases when positional advantage is more imprortant than material(of course with verfication that is deep enough the winning material move does not win material in this case but it is practically depth that computers cannot reach and they need to do evaluations that are not about material to get the right conclusion.

duncan
Posts: 12029
Joined: Mon Jul 07, 2008 8:50 pm

Re: Are there engines to solve win material problems

Post by duncan » Thu Oct 01, 2020 10:36 am

Uri Blass wrote:
Thu Oct 01, 2020 6:55 am
I know that engines solve mate in x plies but I do not know if engines can solve win material problem in x plies.

For winning problem there is a problem that is not in mate problems that maybe you win material but later lose it
so I think that we need definition of win material in x plies and verfication in y plies.

I will give an example



In the following position white can win a knight if you search 1 ply forward with no verfications by 4.Bxc6

If you search 1 ply forward with verfication of 3 plies white win a pawn by 4.Bxc6 dxc6 5.Nxe5
If you search 1 ply forward with verification of 7 plies then I think that white does not win material because I
think that white cannot keep the advantage that he got after 1 ply both after 3 plies,5 plies and 7 plies.

4.Bxc6 dxc6 5.Nxe5 Qd4 6.Nf3 is still enough to keep the 1 pawn advantage after both 3 plies and 5 plies but white has no way to keep 1 pawn advantage also after 7 plies.

If you search 7 plies forward without verification of more than 7 plies then again I suspect that white win a pawn because black cannot prevent a stupid sacrifice in the last ply that is refuted only if there is a verification at bigger depth.


Basically I would like an engine to get 2 numbers:
1)search depth in ply that I denote x
2)verfication depth in ply that I denote y.
The engine should find all the paths that get the biggest material gain in x plies when you also keep the advantage in x+2k plies for every case when k is positive and x+2k<=y

Note that the number can be also negative and the result can be minus 1 if the best you can do is to lose a pawn.

I think that an engine like that can be useful for generating chess problems for humans that are simply about winning material or not losing material .
It also may be interesting to see how many winning material moves based on the limited depth that computers can search are simply bad moves based on stockfish to learn about cases when positional advantage is more imprortant than material(of course with verfication that is deep enough the winning material move does not win material in this case but it is practically depth that computers cannot reach and they need to do evaluations that are not about material to get the right conclusion.

Do you know if Postmodernist and Sos do material only searches?

https://www.stmintz.com/ccc/index.php?id=404922

I was hopeful that PM would finish 35 plies over the weekend, but it didn't
happen. When I got up this morning, there had been no progress since Saturday
afternoon and I didn't want to leave my PC on all day. PM had had 43 hours since
it finished ply 34



https://www.chessprogramming.org/PostModernist

https://www.chessprogramming.org/SOS

Post Reply