Making A Plan

[towforce]

I recently suggested using a directed graph of states to make a plan (link). It may be there are too many states in chess for this to be practical. Maybe a program could generate a graph of states which are similar to the current state, and use this to build a plan.

But then, what's the point of a plan anyway? Most chess programs are designed to select the least-bad move, and this seems to be the best approach to playing chess. I remember reading the importance of having a plan in an old chess book that probably predated computer chess: plans seem to be useful as a goal-oriented framework for human thought.

Two questions:

* is there any value in trying to get a chess program to make a plan? Maybe a more human style of play? Maybe a teaching aid?

* would the use of shortest path algorithms in a directed graph of states be a good way of making a plan?

[glav]

Interesting project. Today program play is often beyond comprehension of the vast majority of human players. Therefore, having a program able to explain the reasons behind his reasoning would be a very welcome addiction for humans, even if this tool should be hundreds of points weaker than the the strongest ones.

[smatovic]

Yes, reasoning would be a benefit over current kind of engines.

--
Srdja

[Harald]

The idea of computing plans in chess is very old. For nostalgic reasons you can look at the book

Chess Skill in Man and Machine, Edited by Peter W. Frey, 1977, 1983

Example chapters:
6. Plans goals and search strategies for the selection of a move in chess,
Russell M. Church and Kenneth W. Church
10. Using chess knowledge to reduce search,
David Wilkins, SRI International

You can also find pfd versions of that book in the internet. But may be that is illegal.

[smatovic]

Another one:

At the Advances in Computer Chess 4 conference 1984, Kai von Luck and his fellow Bernd Owsnicki discussed a method of plan construction in computer chess with results from its application in a program called N.N

https://www.chessprogramming.org/N.N.
https://www.chessprogramming.org/Kai_von_Luck

--
Srdja

[towforce]

The idea of computing plans in chess is very old. For nostalgic reasons you can look at the book

Chess Skill in Man and Machine, Edited by Peter W. Frey, 1977, 1983

Example chapters:
6. Plans goals and search strategies for the selection of a move in chess,
Russell M. Church and Kenneth W. Church
10. Using chess knowledge to reduce search,
David Wilkins, SRI International

You can also find pfd versions of that book in the internet. But may be that is illegal.

I actually read that book in a library in about 1983/84. I still remember some of it! My favourite part was the analysis of human play: apparently, to become a GM, you need to be able to recognise the magic number of 50,000 different patterns that occur in chess games.

The reality is, though, that the top chess computers rarely did much that humans would describe as "planning".

[towforce]

Another one:

At the Advances in Computer Chess 4 conference 1984, Kai von Luck and his fellow Bernd Owsnicki discussed a method of plan construction in computer chess with results from its application in a program called N.N

https://www.chessprogramming.org/N.N.
https://www.chessprogramming.org/Kai_von_Luck

--
Srdja

Can you give us an outline of Bernd's plan to make plans?

[smatovic]

Nope, I did not read the papers, yet.

--
Srdja