Is there any project coming to solve chess?

[Uri]

Sometimes I worry about the future of our species.

If chess is too difficult for humans to solve then how are humans going to solve more difficult problems like how to get to Mars or perhaps even to another solar system.

Sometimes I ask why couldn't real life be more like an X-Men or a Star-Trek movie where everything is advanced and everyone is happy?

You know this really makes me wonder if technological utopianism is achievable at all in our reality or is just a fantasy that is never going to be attained. Maybe in an X-Men kind of world a technological utopia is attainable but just not in ours.

https://en.wikipedia.org/wiki/Technological_utopianism

[towforce]

Sometimes I worry about the future of our species.

If chess is too difficult for humans to solve then how are humans going to solve more difficult problems like how to get to Mars or perhaps even to another solar system.

Sometimes I ask why couldn't real life be more like an X-Men or a Star-Trek movie where everything is advanced and everyone is happy?

You know this really makes me wonder if technological utopianism is achievable at all in our reality or is just a fantasy that is never going to be attained. Maybe in an X-Men kind of world a technological utopia is attainable but just not in ours.

https://en.wikipedia.org/wiki/Technological_utopianism

Firstly, as I've said before, I think chess can be solved (proven to be a draw) with today's technology. The problem with this forum is that most of us are "game tree people" at heart, and the simple fact of exponential growth makes it almost impossible to prove the outcome of chess with a game tree.

Secondly, today's technology can make you happy - you just have to choose to use it that way. Think of the things that make men happy:

* doing things which are fun and challenging

* having a sense of purpose/mission

* having a sense of making progress

* having equal of higher status to the other men in your environment

* getting a girlfriend you like

Used well, technology can help with all of those things.

[paulcolley]

(1) who indeed mentioned a certain Paul Colley, only relevant Colley I could find so far only has a
Bachelors degree (and now is head of an information dpt in Berkeley)

I am the Paul Colley cited in Victor Allis' thesis, I can be reached at paulcolleyster@gmail.com. I did some work on games during the 1990's but nothing relevant since. For what it's worth, Victor Allis' thesis slightly confused the credit — he misattributed which two of the three terms I suggested.

My discussion with Victor Allis was about what it means to "solve" a game. If you talk to a random person and claim you have "solved" a game, most players would think you mean that you can play it perfectly. That they could, for example, show you a position from the middle of a game and you would know what the perfect move was, and who would win or draw. Perhaps they could just dump chess pieces on a board, and if you claim you've "solved" chess, they'd expect you to be able to play from that random position.

Programmers would generally mean, "my program achieves the best possible result when it plays the entire game, in reasonable time" In particular, most programmers would not want to be handed positions which can not turn up in a real game — Lots of chess programs would crash if asked to start from a position with multiple unpromoted pawns on the final row, or three kings on the board. Many "solved by program" games require opening books to reach game positions they can play perfectly (checkers was one such, when it was first solved.)

(Note that it's easy to write a chess program that provably plays perfect chess, and the hardware requirements are trivial — less hardware than you'd find in a modern digital watch — so "reasonable time" is a very important part of that statement)

Mathematicians generally mean, "I have proven something relevant," the better of them, "I have proven something interesting to other mathematicians!" The best, "I have proven something publishable!" I am a mathematician; the fact that hex has a trivial proof showing it's a first-person win even though we can't play it well on non-trivial board sizes is fascinating. I do not have any non-mathematician friends who agree with me.

To put it another way, we know that the opening position for hex (no pieces on the board) is a first-player win; once you start putting even one piece on the board we generally don't know who would win. The team that "solved" nine-men's morris announced it on Usenet claiming it was solved because they knew they could play perfectly against a perfect opponent; it was not clear from their announcement whether they could play perfectly against a fallible opponent (I assume they either could, or rectified the problem, since it would not have required much more computation).

I won a computer tournament for a very obscure game (Lewis Carrol's numerical croquet) with a program that did not play perfectly against opponents who had perfect programs. It turns out that "wrong" moves can be harder to respond to than right ones; and my program provoked errors from more of the imperfect entrants than the perfect programs did. I don't recommend this approach; I was just lucky that my program didn't make any mistakes against the perfect entrants.

The team that solved checkers noted the same problem: When their program reached positions that it knew were tied, it would often make moves that were very easy to respond to, and hand the human opponents an easy draw. It's a difficult problem to understand, among the set of drawing checker moves, which of them are hard for a fallible human to respond to and which are easy. You want to play the hard ones.

All this is behind the desire for a terminology which would succinctly label, "I can play this game from the starting position and achieve the game-theoretic value" vs "I can play this game starting from any reachable position and achieve the game-theoretic value" vs "I can tell you the game-theoretic value of the starting position, and many other interesting positions, but I can't actually play the game."

This terminology seems off-topic for discussing rules of correspondence chess.

(2) whether that's an ultraweak solution in the opinion of prof vd Herik i really don't care, i'm not interested in
nittygritty, math purists nomenclature, arbitrary definitions nor nitpicking ad in finitum. Like i already stated my
main motivation was in pointing out the need for a rule modification in ICCF correspondence chess, not in a
endless discussion with math purists about definitions. The thesis of Allis btw is pleasant reading, you
might be able to learn from that, as you -contrary to Victor Allis- don't have a Phd (whereby as an IP 'lawyer' i would suppose you
would be able to understand what the definition of a definition is. Anyway, i have 'determined' the definition
of an ultraweak solution in game theory is not well established and although it might be subject for further
research depending on further outcomes as subsequently possibly accepted by some (preferably more) academic
professors in game theory, I decide not to use this definition anymore (because it's almost meaningless),
and while vd Herik now is almost with pension, i expect this lousy definition will not survive in game theory anyway. Amen

Please avoid ad-hominem attacks.

[towforce]

Let's say there are two levels of solving a turn-based game:

Level 1: You prove what the result of perfect play would be from the starting position. Checkers has been solved to level 1.

Level 2: You have a "god algorithm" that can correctly tell you the outcome with perfect play from any given position without having to build a game tree (so it can be done quickly).

Regarding the problem of perfect play often making it too easy for the opponent to get a draw in a drawn position: here's what you should do:

1. In a winning position, choose the first move on the shortest path to the win.

2. In a drawing of losing position, choose the move that maximises the complexity of the position.

Point (2) above works for both humans and engines: humans get confused when there's too much to think about, engines lose search depth when there are lots of pieces attacking (or potentially attacking) each other.

[paulcolley]

In a winning position, choose the first move on the shortest path to the win.

Yes, that is reasonable. (I'd prefer the term "quickest" over "shortest" because sometimes we have to balance computation time vs number of moves, and "quickest" nicely encompasses both issues).

2. Choose the move that maximises the complexity of the position

This is a concise statement of the problem ... but how do we measure the complexity of a position? What is hard for a program is not the same as what is hard for a human (and somehow it seems against the spirit of chess for a program to play differently against a human than against a computer!)

The obvious metrics that are already computed by typical chess programs are not accurate predictors of complexity. E.g., Longest series of moves, or highest branching factor, are not very predictive of how hard a position is to play.

Endgame databases just make these issues more obvious.