Chess will never be solved, here's why

Calls to a non-existent algorithm are just like believing in the paranormal.  There's no basis for it.  There's only one method currently underway to solve chess, and that is building tablebases backwards from all possible mates.  Significant enough progress on that will not happen in our lifetimes, or even within any possible technological future that we can make concrete plans toward.  This thread, like all the other threads on this topic, was over before it began.

As I have pointed out in the past, trying to build a bridge to the steel structure that tablebases represent using the mud and reeds and Tygxc proposes is not a worthwhile effort.  But at least it's an attempt of some sort, deluded as it may be, which is more than can be said for Optimissed's talking in circles, which is pure trolling and pseudo-intellectual twaddle.  

"there's chess positions with exact solutions. Either e4, or d4, or c4, etc. "

This sentence perfectly shows the Descartian small-mindedness of the OP.

I only  have this message for you:

To understand chess and why it can be solved, you have to completely let go of all you have learned in your current education system, i.e. Descartian logic, and step to chaos theory i.e. probabilistic logic as the fundament of your thinking. The brain and everything in the universe is probabilistic, even a computer chip - it's only an illusion that it's deterministic because technocrats like to think away organic substrate.

Chess cannot and will never be 'solved' because engines keep getting stronger. Meaning concretely, if you think there is a fixed set of best moves this isn't true because a stronger engine might find a long line based on a different prior move that turns out stronger, hence changing 'theory'.

This search for ultimate knowledge is a mental remnant of a Babylonic attempt to control the world, now simulated on the microworld of the chess board. Even if we think Western civilization is modern, this mindset is actually centuries old and seen through.

It renders this topic obsolete. Chess is a wicked problem, it cannot be solved. As in here: https://en.wikipedia.org/wiki/Wicked_problem

We have to understand this.
The Descartian age is long gone, guys and gals. Forget about traditional beta subjects. Logic, and in the end, AI, is something alive just like all people and other creatures.

It is the size of the game that puts it beyond reach. Other than that chess is similar to checkers which has been solved (in the sense of the peer-reviewed literature - exhibiting complete, game theoretic optimal strategies for both sides).

Thankyou but that is completely beyond me. I did maths up to the end of first year degree level mechanical engineering and dropped out of university first time round, eventually getting a degree in philosophy. My wife, who has an MSc in psychology, had to study maths to be able to do the fairly mathematically simple research element. It amazed both of us that we produced a mathematician between us! However, now he's a mechanical engineer!

I can remember Edmund counting the stars when he was about two. His method was this: "one, one, one, one, one, one, one, one ....

See, these are the kinds of posts that make you seem redeemable...hard to reconcile with the imbe-/idio-/incomp- words, though.  

and...

Optimissed wrote: 
I can remember Edmund counting the stars when he was about two. His method was this: "one, one, one, one, one, one, one, one ....

...a valid counting method, given that each star is unique at a certain level of detail. 

I was going to add a bit more, @Elroch. I would think that the Riemann hypothesis, which I don't understand, sets conditions for ensuing equations. The significant theorem must tie in with the conditions where the Riemann is true and also where it's false but the two sets of equations are exclusive to one-another. Riemann being true and false would be hypothetical in both cases and thus it wouldn't matter if it were true or false provided the methodology worked.

In Edmund's thesis, he had to depict magnetism in terms of fermionic spin. He realised that, similarly, it splits into two parts. It took him six months just to write down the equation and four years to solve it. The first six months was taken up with learning physics to degree level, from scratch, since he'd never studied it. He told me that solving the equation hadn't been a proof and it was necessary to hold all his variables as constants, his constants as variables and do it again. Only this time, he thought, the mathematical methodology didn't exist and might take as much as 20 years to be developed (he thought). If both methods upheld the idea, it was then necessary to use the equations to predict which metallic alloys could be used to empirically test the hypothesis. Success there would mean it had become theory. I asked him if he would get any credit and he thought not. The credit would go to those who managed to solve the harder part and then demonstrate it empirically.

Apparently there are several theorems in number theory where it is convenient to prove them in two halves as I mentioned (and perhaps loosely analogous to your suggestion). It is rather fascinating, as you wouldn't instinctively expect it to be possible for both a proposition and its negation to be useful in proving a different result.

https://en.wikipedia.org/wiki/Riemann_hypothesis#Excluded_middle

Tablebase is 140 terabytes. We need Chess 2.0 to come out soon.

(Chess is solved with 7 or less pieces on the board)

But that is actually used routinely, you don't need any esoteric examples to demonstrate it.

Not quite.

It's not solved for the majority of positions with castling rights.

For the remainder it depends on what you mean by chess.

FIDE chess is insoluble because no ordered yield is defined for the players in the case of various events occurring simultaneously e.g. simultaneous resignations. FIDE chess is not a zero sum game.

It is a simple matter to fix the FIDE laws to make it a zero sum game, but there are different possible ways. 

If, for example, the resignation and agreed draw rules were excised then the Syzygy tablebases would provide a strong solution of the remainder for both players in FIDE's basic rules game since 2017 (when the 50 move and triple repetition rules were dropped).

Nothing would currently provide any sort of solution of all the remainder for either player in FIDE's previous basic rules (when the 50 move and triple repetition rules were still in force). Those rules (with some further fixes) have mostly been regarded as chess since Ruy Lopez.

The same would be true of FIDE's current competition rules (though those would obviously need more comprehensive amendment to describe a soluble game).

For example, Black can draw under FIDE competition rules in the final position shown here by 26...Ka1, but all current tablebases, including Syzygy, give only 26...Kc1 which loses. 

What the Syzygy tablebases do is provide a weak solution of all 7 man positions that neither repeat previous positions considered the same under FIDE art. 9.2.2 nor are preceded by such repeats with the same material. That is sufficient for perfect play by either player in a game, because the first such position to occur cannot include such repeats.

Also current tablebases except Syzygy (e.g. Nalimov and Lomonosov before it was sabotaged) fail to solve many ply count 0 positions under FIDE pre 2017 basic rules or current FIDE competition rules.

(What's Chess 2.0?)

@8661

"It is the size of the game that puts it beyond reach"
++ Chess is 1000 times larger to weakly solve than Checkers.
Checkers was weakly solved using 10^14 relevant positions of the 10^20 legal.
Chess needs 10^17 realevant positions of the 10^44 legal.
Schaeffer used 50 desktops to weakly solve Checkers.
He used 200 desktops to generate his 10-men table base.
For Chess the 7-men table base is already there.
The 50 desktops anno 2007 correspond to 3000 desktops now or 3 cloud engines.

@8659

"There's only one method currently underway to solve chess,
and that is building tablebases backwards from all possible mates."
++ But that would lead to a 32 men table base strongly solving Chess for all 10^44 legal positions, taking too much time and storage.
Such a strong solution would contain all weak solutions.
There is only one viable method: weakly solve Chess as Schaeffer did for Checkers and as Sveshnikov proposed: calculating from the opening to the 7-men endgame table base.
That takes 5 years to exhaust all 10^17 relevant positions and costs $ 3,000,000.

@8660

"if you think there is a fixed set of best moves this isn't true because a stronger engine might find
a long line based on a different prior move that turns out stronger"
++ A stronger engine does not play better moves, it plays less bad moves in the same time.
You can weakly solve chess with any engine, however weak or strong.
It only needs to calculate all the way to the 7-men endgame table base.
A weak engine will take longer to do that than a strong engine, as it will run astray more often.

I didn't say it could be completed, I said it was the only valid method currently underway.  Your method will not work, and certainly not in the timeframe you claim, as has been pointed out to you countless times now by numerous posters.

@8674

"I said it was the only valid method currently underway."
++ Only an 8-men endgame table base is underway.
It is no valid method as it takes up too much time: billions of years and storage: 10^44 bit.
That is not how Schaeffer weakly solved Checkers: he only generated a 10-men endgame table base and then calculated towards it from the initial position.
That is just what Sveshnikov proposed to do in 5 years for Chess.

"Your method will not work" ++ It works, it worked for Checkers.

"not the the timeframe you claim" ++ It does. Three 10^9 nodes/s cloud engines piloted by 3 grandmasters can in 5 years exhaust all 10^17 positions relevant to weakly solving Chess.

"pointed out to you countless times now by numerous posters"
++ By ignorant posters, like yourself: you erroneously take the prohibitive time to strongly solve Chess to a 32-men table base as the same time to weakly solve Chess.

tgyxc, your proposed method is flawed. There's no doubt.

Congratulations!

You managed to forget the explicit example I posted in #8670, which no tablebase constructed using current methods solves ultra-weakly, weakly or strongly in chess with 50 move and 3-fold repetition rules (like the version you claim you will solve) - in the space of one post.

No wonder you can't remember all the other occasions I've pointed this out to you.

How do you expect to be taken seriously when you claim you will solve 32 man chess, if you don't even understand the first thing about the solutions we already have for 7 and fewer men?

(And it's still not 10^44; not even under basic rules.) 

I think that not having a clue causes most of it.