Is there any project coming to solve chess?

[syzygy]

Back to this thread coz the other thread (chess is a draw) becoming even worse;
NB It was already known 150 years ago that it's a draw (B/W equilibrium) by former world champ Steinitz
who als had studied two years math btw. Chesskobra mentioned some Gm's who also are mathematicians
they ofcourse all would confirm White can only win if Black makes a mistake; this is a known fact. Go ask
them ChessKobra, instead of making such silly claims (they all know, but won't like to make statements
as 'chess is a draw' because they think that's not good for the game of chess (*.

So the concept of mathematical proof still eludes you.

[Uri Blass]

It runs contradictory to your intuition of a possible narrow path. I think for the narrow path to be feasible the margins would need to be slimmer than they are.

Nothing to do with the narrow path.

The narrow winning path from the starting position "probably" does not exist. But it is not possible to prove that it does not exist (without a gigantic computational effort for which we do not have the resources). Therefore we cannot solve chess. That is the argument against "we will solve chess in a few years because SF's eval says this or that".

This has nothing to with my claim that SF is beatable from the starting position.
WIth the latter I just mean that there is at least one line (and in reality billions of lines) in which SF makes an entirely unforced blunder that moves the game from theoretically drawn to theoretically lost for SF. This line is not a "narrow line". The line has nothing to do with perfect play. It is just a line that gets SF into a drawn position where SF screws up because of the holes it has in its evaluation and search, and perhaps some unlucky hash table collisions.

I think you should appreciate our usage of "weakly solved in practice". Instead you could be facing "essentially solved" as authors of Cepheus have used in their publication in Science (they haven't really solved it but got close enough for practical purposes).

If words don't matter, then why not call checkers chess? Then chess has been weakly solved by Jonathan Schaeffer.

I guess you mean:
"It is just a line that gets SF into a lost position where SF screws up because of the holes it has in its evaluation and search, and perhaps some unlucky hash table collisions."

No, I meant what I wrote. This should not be difficult? You cannot screw up in a lost position, since you are already lost in a lost position. To lose a draw, you need to blunder in a drawn position.

Note that SF is not designed to be unbeatable and it is easy to change it by changing the code.

Duh?

1)I understand now what you meant.
Of course I understand that you need to cause stockfish to blunder in a drawn position in order to win but I thought the position you meant is the position after the blunder when I read "into" and I thought the SF blunder is before the position.

2)I mentioned the fact that SF is not designed to be unbeatable(and it is still hard to beat) to explain one of the reasons
that I am optimistic about having some unbeatable engine in the future even in case that it is possible to beat SF today at long time control.

Of course trying to have unbeatable engine can reduce the elo because one of the ideas is that the engine does not take risks and prefer to force a draw even if maybe it means missing a win against some non perfect opponent.

[towforce]

Most of you already know my views on this, but I'll quickly restate them:

Is there a project to solve chess?

No.


Is chess a draw?

Extremely likely - but not proven. For me, the strongest evidence is that for all the progress that has been made, no way has ever been found to win material from the opening position.


Is it possible to prove that chess is a draw with today's technology?

I have no proof, but I believe that it is. The key would be to stop building game trees, and instead find enough iron rules or underlying patterns. I'll give a couple of possible routes:

(1) Generate a tree of conditions to exhaustion

At the root of the tree (level 1): the conditions that (one or more of) must exist in order to checkmate (or win material, to approach from a different direction).

For each condition (level 2 of the tree): the conditions that must exist to get to a position where level 1 conditions exist

Level 3: the conditions that must exist to get to level 2

Each time a condition is generated, it will be checked against all the other conditions generated. If it has all ready been seen, this node of the tree can be terminated. Thus the tree will eventually self-terminate, and you'll have all the conditions needed for checkmate (or to win material). You can then apply these conditions to the chess starting position.


(2) The patterns of chess are complex - but most complex patterns have simple underlying patterns. This is VERY likely to be true of chess. So... find chess's underlying simple patterns, and use these to acquire a better understanding of chess than anyone has ever had before. Once these patterns have been acquired, it's likely that a program will be able to play close to perfect chess on a cheap and simple computer with almost zero thinking time.

[Uri Blass]

Most of you already know my views on this, but I'll quickly restate them:

Is there a project to solve chess?

No.


Is chess a draw?

Extremely likely - but not proven. For me, the strongest evidence is that for all the progress that has been made, no way has ever been found to win material from the opening position.


Is it possible to prove that chess is a draw with today's technology?

I have no proof, but I believe that it is. The key would be to stop building game trees, and instead find enough iron rules or underlying patterns. I'll give a couple of possible routes:

(1) Generate a tree of conditions to exhaustion

At the root of the tree (level 1): the conditions that (one or more of) must exist in order to checkmate (or win material, to approach from a different direction).

For each condition (level 2 of the tree): the conditions that must exist to get to a position where level 1 conditions exist

Level 3: the conditions that must exist to get to level 2

Each time a condition is generated, it will be checked against all the other conditions generated. If it has all ready been seen, this node of the tree can be terminated. Thus the tree will eventually self-terminate, and you'll have all the conditions needed for checkmate (or to win material). You can then apply these conditions to the chess starting position.


(2) The patterns of chess are complex - but most complex patterns have simple underlying patterns. This is VERY likely to be true of chess. So... find chess's underlying simple patterns, and use these to acquire a better understanding of chess than anyone has ever had before. Once these patterns have been acquired, it's likely that a program will be able to play close to perfect chess on a cheap and simple computer with almost zero thinking time.

You need to define what you mean to win material.

If the target of the game is to have material advantage when the game ends as soon as you get it then white wins easily.
I think I can win at least a pawn in 3 moves as white
1.e3(the threats are
2.Qf3 (a fork against b7 and f7) and you cannot protect both pawns and 2.Qh5(a fork against f7 and h7 and you cannot protect both threats.


I see no way to protect against both threats and even no way to protect against 2.Qf3.

The fact that white is losing in chess after 1.e3 e5 2.Qf3 d5 3.Qxf7+ is not relevant because in the game of winning material white already won and the game does not continue.

If the target is to have material advantage after the opponent move or to mate the opponent then I do not know if white can win the game or not.
I know no strong engine that can play with this target so I can see results of it against itself.

Edit:I think maybe you mean that the target is to win in chess or to have material advantage in the final position that is a draw by the chess rules.
I guess it is a draw also in this case.

[syzygy]

2)I mentioned the fact that SF is not designed to be unbeatable(and it is still hard to beat) to explain one of the reasons that I am optimistic about having some unbeatable engine in the future even in case that it is possible to beat SF today at long time control.

But you can't just decide to "design" an unbeatable engine.

A truly unbeatable engine is a weak solution to chess. One side in this thread argues that we lack the resources to weakly (or ultraweakly) solve chess. The other side says "but SF is a weak solution because it is unbeatable".

Since we agree that there is no reason to believe that SF is unbeatable, we agree that the argument based on SF is unconvincing.

Of course trying to have unbeatable engine can reduce the elo because one of the ideas is that the engine does not take risks and prefer to force a draw even if maybe it means missing a win against some non perfect opponent.

Actual perfect play can indeed be very boring.
But today 99.9% of engine design is still aimed at playing "as perfectly as possible". Since engines aren't perfect, that can still give exciting play.

[jefk]

Ronald de Man, aka syzygy, i've noted your Ad Hominem remark that the concept of a mathematical
proof eludes me. Maybe i should report it (*). In the meantime, apparently verbal, semantic logic
apparently eludes you.
Point is, as long as there is no clear definition of what ultraweak solving of game as chess should be,
there -at least imo- is no need for a rigorous math proof, although you (and maybe Dann Corbit)
repeatedly are claiming that. For many things in reality you don't need mathematical proofs. And chess besides
the option of representing it as a mathematical game, also exists in reality not in the least btw
because it's a finite game and that can be Proven (yes, Zermelo did that as well Lol.

My ultraweak solution, (or maybe only conjecture in your eyes), proof or not, is strengthened by the addition
with the minimax theorem by von Neumann (proven) and the results sofar from the Chinese database.
So awaiting the judge Judy in the world of Math Purists if ultraweak solving needs a Proof, I
claim that with my reasoning, stronglin resembling an ultraweak solutiong means that chess
has been 'essentially solved' (your words btw syzygy).

But because it's not a 'proof' it may never convince some the mathpurists/ number crunchers.
Some math purists even stricter tyan you might also claim you did't solve the 7 men endgame theory, you
just made some tables, who says there can't be some holes in these tables, can you Prove that ?
(see what i mean, well otherwise others will) And don't tell such a person some stuff like brute force,
*all* positions, they might not except it; they could endlessly make objections, in other words,
nitpicking, eg. Prove that your compression algorithms may not have added errors, etc. etc etc
Ad infinitum, sigh

(*) although i'm not going into details about mathematical philosophy, in my physics
education i learned enough about mathematical proofs; eg. by complete induction, (which i
btw in an original proof within mathematical digital control theoryas part my
mathematical Msc thesis in 1979 (awarded with a 9 btw in other words i'm not an idiot)).
So ofcourse i know what a proof is or should be, but i also know there often were sometimes disputes
in math, eg. between intuitists as Brouwer and purists as Hilbert. Nowadays some math purists may
continue in endless nitpicking about what a game solution should be for chess while the results
are almost as clear as for tictactoe.

[towforce]

You need to define what you mean to win material.

Sorry. What I mean is: an opening in which (probably white) can forcibly win a pawn or a piece without the opponent being able to win equivalent material back.

Obviously, winning material is only an indicator: it does not guarantee winning the game, or even drawing the game - it's possible to be ahead on material and lose the game.

However, the fact that there is no known opening that can win material, even with powerful computers having looked very deeply, tells me that the opening position is likely to be one from which the game cannot be won: if there's no way to win a pawn or a piece, then there probably isn't a way to win the game.

[Uri Blass]

2)I mentioned the fact that SF is not designed to be unbeatable(and it is still hard to beat) to explain one of the reasons that I am optimistic about having some unbeatable engine in the future even in case that it is possible to beat SF today at long time control.

But you can't just decide to "design" an unbeatable engine.

A truly unbeatable engine is a weak solution to chess. One side in this thread argues that we lack the resources to weakly (or ultraweakly) solve chess. The other side says "but SF is a weak solution because it is unbeatable".

Since we agree that there is no reason to believe that SF is unbeatable, we agree that the argument based on SF is unconvincing.

Of course trying to have unbeatable engine can reduce the elo because one of the ideas is that the engine does not take risks and prefer to force a draw even if maybe it means missing a win against some non perfect opponent.

Actual perfect play can indeed be very boring.
But today 99.9% of engine design is still aimed at playing "as perfectly as possible". Since engines aren't perfect, that can still give exciting play.

I think that even if some engine is a weak solution for chess then proving it is practically impossible today.
I do not know if stockfish at some 1,000,000,000 nodes per move is a weak solution for chess but even if it is not a weak solution for chess then I believe that we are very close to have a weak solution for chess simply because the draw margin in chess is relatively big,

[chesskobra]

I am willing to believe that jefk knows the concept of a mathematical proof. And I am not even demanding a mathematical proof (even if I insist that such a proof is not known). I would like to know precisely what empirical evidence you have that supports the conjecture that chess is a draw. I would like it to be clearly stated (e.g., you could say that X% of the positions in set S have been proven drawn, or MCTS gives ... W/D/L statistics for starting position after N simulations, or MCTS with UCT gives ... W/D/L statistics after N simulations). I would like numbers. Then we can debate about the statistical methodology, how the set S was selected, convergence properties of various algorithms, and so on. I have not much to say about vague statements. Also, I have no interest in pi and its digits (pi "remains irrational" even after million digits - that kind of stuff).

In fact regardless of this ongoing debate I wanted to ask, if anybody knows, about the W/D/L statistics from simulations with MCTS (and its variations).

[jefk]

1) ok ChessKobra,
statistics with an Mcts search with increasing times, that can be done but i would do it at first rather
superficially because i'm currelty using my comp for other things (opening
book improvement); second, as i won't do it with Lco (too weak hardware)
you would only get Shashchess output which is ('cp') score, converted to drawing chances. So I prefer to leave this initially to others,
you will see initially 55 pct for 1.d4 and 1.e4, and this will
gradually go down; depending on you Gpu.

2) in addition to earlier response to syzygy (and Towforce), talking about
logical proof (*), I do use the Reduction Ad Absurdum method which sometimes is used as (rigorous)
method of proof in math (eg. to prove pi is irrational).

And it goes as follows:
1) if there is a forced win then there must be winning strategy (Zermelo
2) but there is no winning strategy (Mcts and/or Chinese database with minima
3) So if there is no forced win for White then it's a draw

Only point 2 above is not 'rigorous' according to the math purists. However, i claim that if
there is a winning strategy, then a solid Mcts search would increase after time the indication of winning chances. In chess it doesn't.
Only thing to remain, 'prove' that the Mcts search which we use nowadays (eg in shashchess or Lco)
is solid. Well imo it's evident, while likely still some things could be improved on the Mcts
search, it's already by now is fit for purpose. Requiring to Prove it in a rigorous mathematical sense
would be imo almost just as ridiculous as proving the syzygy bases are correct.

(*) besides defining ultraweak solving in a rather 'loose' way maybe