Chess will never be solved, here's why

We don't know all the laws of physics, nor the exact structure of the universe, so theories about it are just more or less accurate models. They are in fact approximations and scientists have understood that long ago. You know that QM and GR do not explain everything and likely they will be superseded, as others theories before.

But solving chess is another thing. We know all the laws (the rules) of the game and its structure (64 squares, 32 pieces...), and the number of positions and possible games is finite; thus an exact solution must exist. When a game is announced weakly solved (in the game-theoretic meaning), nobody really expects that some day that solution will be falsified. The only way to be sure of that is to test a strategy (which provides a move every turn) against all the possible opponent's replies, or to provide a general theorem (even by mathematical but only mathematical induction) that encompasses all the possible situations a player can face along the path, no exception. The latter seems impossible, considering the game complexity, so a solution has to rely on computations to provide the first type of proof. Heuristics provide strategies to be tested in an ordered fashion (so e.g. if a strategy wins from a position, no need to search for something better), but if they cut off lines according to a non exact evaluation (that is, not a tablebase hit or a 3-fold repetition), the strategy is not 100% reliable. The greater the number of non-exact cutoffs, the greater the risk the "solution" will be falsified later.

But solving chess is another thing. We know all the laws (the rules) of the game and its structure (64 squares, 32 pieces...), and the number of positions and possible games is finite; thus an exact solution must exist.>>

No, I don't think it exists and any proof it exists is also an inductive proof.

What does "no, I don't think it exists" even mean?  That you do not believe a solution can exist?  That you disagree that a full 32-man tablebase would be a valid solution if it were ever achieved?

The point is valid.  Every discrete piece of knowledge needed to solve chess and a valid method for solving it already exist, the solution is just not achievable until some pretty large breakthroughs that will likely not occur in our lifetimes are made.  What you are saying is like saying that is is impossible to count the grains of sand on a beach...but it *is* quite obviously possible, just entirely infeasible at present.

I don't think it can necessarily be achieved.

No method for "solving" it exists as yet. It may be that it will be developed but it can't be one of humanitiy's priorities. My opinion is that chess may reach a full solution but probebly won't, ever.

I also disagree with the idea that <<an exact solution must exist>>. It may be that one hypothetically exists. That is, a solution might be "waiting to be found" but I'm not convinced of it. Even if it existed, it would be difficult to recognise it for what it was.

Hello, that's been discussed in this and other threads. I would say that no agreement has been found as to what is meant by solving chess. Some people imagine a full solution of all possible lines. Others understand that's completely silly. Therefore, much of the discussion is about how best to define the lines that need to be followed; and people being what they are, no agreement has been reached there, either. Neither is there any progress being made towards agreement, because most of them have allowed various meaningless diversions and they just follow them rather blindly. But of course, humanity waits in hope, upon the outcome.

I would say that the method of constructing the tablebases is already a solution. The tablebases don't actually need to be constructed. We have in fact always had a solution of chess.

That's why I inserted the word "timely" in my definition of a solution. 

For example the checkers solution needs about two minutes to return a move, so it wouldn't be much use in a ten minute blitz game (if they have such things in draughts). 

Asking someone to go through the 32 man tablebase construction process similarly wouldn't be much use in a ten minute blitz game of chess.

A solution is only a solution if it's timely for the purpose for which it's required.

It's Zermelo's theorem in game theory. So if the game value is a win, a win can be forced, otherwise both players can force a draw.

Of course its valid.  
The logic of the point is valid. 
The number of possible chess positions is finite - therefore a solution exists.
Where whoever might go wrong on that - is to believe that since a solution hasn't been found - 'it doesn't exist' ...  that is illogical.
A solution can exist - before it is found.
Claiming its not so is similiar to the error claiming a tree falling in the forest 'doesn't make sound' unless someone is around to hear it.
Of Course the falling tree makes sound Anyway !   

"Mathematics"  doesn't care if something hasn't been found yet.
The ratio pi is the same too - whether there's people around or not.
That ratio doesn't need people around - to be the ratio of the circumference of a circle to its diameter.

Regarding those who don't understand - they're ignoring the significance of the 'finite' part. 
Not grasping the nature of that fact.
But that's their choice. 
There are no 'Thought Police'   

Can anybody do somebody else's thinking for them?
The short answer is no -
but to understand the nature of 'finite' there ...
consider hypothically if there were 'infinite' chess positions - then in that imaginary case it could be argued that chess could never be solved because there would always be more positions to solve.
By itself would that be enough to prove the converse?
Not quite.
But the point about finite - (this is a hint - since there's no 'Vulcan mind meld') ...  is that eventually - given enough time - positions to solve would be 'run out of'.   
Can semantics of words like 'inductive' substitute for mathematical reasoning and logic ?
People who try to use semantics alone to reason - might run into errors of illogic.
Often it occurs that people who excel in subjects like English and History - do poorly in math. 
English and history would be considered 'humanities' I believe.
Does it matter?  Chess is a subject of leisure I'd say - not of necessity.

Invalid claims do get attention though.
That part works.  
And illogic is more entertaining than logic.
Hollywood exploits that all the time !

Incidentally, haiaku, I'm sure you will understand that if something (any hypothesis or theory) can't be falsified, then it isn't a strong hypothesis in the sense that it relates to the real World: to real situations. It's one reason I keep using the Big Bang as an example: because it's rather difficult to see how it can be falsified: that is, proven to be false. An omnipotent, supernatural entity is similar. It can't be recognised as a scientifically valid proposition because it's difficult to see how it can be falsified. Many would say it can't be proven to be false and so it can't be real. I don't like using such an example but the problem is that it's probably about the best example we have, of the issue of falsification and how necessary it is. If something can't be proven not to exist then it can't be considered real, because it cannot be related to evidence in any way.

That point aside, I can see what you're saying.

I haven't gone into any detail in this thread and don't really intend to but in Ponz's thread, I attempted to explain my own conception of how chess might be "solved". My conception was much more heuristically based than what was tending to be proposed. I preferred to analyse games rather than positions, because it seems to me that analysing positions is a false shortcut, since all positions have to be analysed as games in any case, to see where they end up. I also distrusted the idea that games can be analysed effectively in a chronological direction, yet similarly distrusted the idea that a tablebase should be identified in order to analyse backwards from it. That also is a false conception because there's simply no reason why such an arbitrary division ought to be made. All it does is point to the weakness of present analytical methods.

My idea was that in order to "solve" chess, firstly, chess has to be mathematically depicted. Such a depiction would have to involve a considerable amount of heuristics and therefore wouldn't be accurate in a theoretical sense. Therefore, the essence of "solving" is to find a set of heuristics which works in a practical sense. The game is simply too complex for any other approach.

In particular, it's necessary to analyse situations where a "strategic turning point" is encountered, to find what they have in common, in a mathematical sense. That in itself is an enormous project and it hasn't really been started as yet. I've mentioned my son ... well ... he thinks it's impossible. That means, to him, that it couldn't happen within the next few generations because the mathematics for it doesn't exist. There was a situation regarding his dissertation, where he felt that although he had managed one half of a proof, the other half necessitated using his variables as constants and vice versa. He told me he'd done the easy bit and that the maths doesn't exist as yet for the "difficult proof" but he thought it would exist within a generation. This means that he considers the chess problem far more difficult. That is, "not for the foreseeable future". I think he and I think alike on that one, which is why I think a solution won't be found.

#2925

'"Towards" tablebases literally means towards a full solution.'
++ I interpret is as starting from the opening and working towards i.e. in the direction of the 7-men endgame table base, as opposed to strongly solving chess by generating 32-men table bases starting from the 7-men endgame table base and working towards i.e. in the direction of the opening.

'I'm trying to be generous to Sveshnikov but I must admit "and 'close' chess" sounds very much like your interpretation.' ++ Yes, I think so too.

"In that case it was clearly an empty claim, which he would never have been capable of carrying out: and so you've chosen to believe a completely empty claim."
++ I do not think it is an empty claim and certainly not clearly. I believe he could have carried it out given the modern computers and the good assistants. 

"I wonder why you imagine he could have done it."
++ I admit when reading it first I found it hard to believe, but after thinking more and checking some facts & figures I believe he was spot on.

"Bobby Fischer made various claims, which tended towards subjectivity, such as the "best by test" garbage."
++ Most of Fischer's claims turned out right too. Apparently the participants of the Yekaterinburg Candidates' also believed his "1 e4 best by test": 1 e4 was both played more and with better results than 1 d4, 1 c4, and 1 Nf3 combined. Also 'the King's Gambit loses by force' was vindicated by AlphaZero and Kramnik.

"As well as holding various political opinions, which were rather off the rails."
++ Political and religious beliefs do not disqualify opinions on other subjects. A good example was Nobel prize winner Schockley. Alekhine and Karjakin are other examples.

"Sveshnikov's claim is not a chess-related claim at all."
++ It is a chess-related claim, it is about chess analysis, that was his job.

"It's a computing-related one."
++ No, it is not computing-related. His 'good assistants' should be grandmasters or even better ICCF grandmasters, skilled in chess analysis with engines. They must use the engines, not create them. The good assistants must understand chess so as to prepare the 26-men tabiya for calculation and so as to halt calculations when a known drawn endgame of >7 men is reached, like the 2 ICCF WC example games I posted.

"He is/was not a particular expert in computing"
++ He was an expert in chess analysis. He was so before engines existed, and he still was so after engines emerged. We won the world senior chess championship. He wrote the ultimate books on the Sveshnikov Variation, the Kasashnikov Variation, the French Advance Variation, and the Alapin Variation.

"otherwise he would not have made the false claim"
++ This is a circular reasoning. You think his claim is false. Because you think so you consider the claim false is true. Because of that you conclude Sveshnikov was no expert. I turn this the other way. Sveshnikov was an expert at chess analysis with and without engines. Therefore his claim is true. Therefore you are wrong dismissing his claim. Therefore you are no expert at chess analysis.

"You see, you have managed to persuade me that he probably did make the false claim. I had assumed he didn't."
++ Now comes the second part: managing to persuade you that his claim is not false and thus you were right assuming he did not make a false claim.

I have great respect for Sveshnikov as a chess player and researcher. I have no fewer than three full length books on the Sicilian variation he breathed new life into. As for the ageing Sveshnikov as computational game theory researcher, well, he just doesn't cut it. He has no publications and no background of the sort of formal discipline needed. He is 100% focussed on the competitive game and practical chess to the extent that he seems even to be unable to realise that more rigour is required in game theory.

My understanding of the character of his viewpoint is that he reckons that past experience of chess, consisting of large numbers of games not a single one of which is KNOWN to be optimal is sufficient to ignore the move 1. a4 as inferior.  This is ok for practical chess but entirely inadequate for game theory. There is no room for compromise. To prove chess is a draw it is NECESSARY to PROVE black can draw against 1. a4, not just assume this without anything close to an adequate argument.

[Note: sadly Sveshnikov died in August 2021 from complications of Covid-19. He will be long remembered and respected as a chess player and writer].

#2930
'the problem is how and how much you rely on that "knowledge" to solve chess.'
++ IMHO the knowledge cannot be incorporated into an evaluation function: we know from Stockfish vs. LC0 that thin nodes i.e. deep calculation with a lean evaluation function beats thick nodes i.e. shallow calculation with an elaborate evaluation function. 
IMHO knowledge should be incorporated by the 'good assistants' to prune the search tree. They should prepare the relevant 26-men tabiya for calculation. They should prune known inferior possibilities like 1 a4 or 1 e4 e5 2 Ba6 and handle transpositions. They should also stop the calculation when an known drawn endgame of > 8 men is reached, like with opposite colored bishops or rook endings 4-4 or 4-3 on one wing.

"you are the only one who thinks you have demonstrated such a thing."
++ OK, let me be the only one. Let all others then come up with another way to explain the available data.

"If you lived in Europe before 1790 you would have concluded that black swans did not exist." ++ It is impossible to prove non-existence of unicorns, Martians, the Loch Ness monster etc. However statistics is extensively used in many sciences and in industry to draw conclusions about a larger set by sampling a subset. That is also what Tromp did to arrive at his 10^44 legal chess positions.

As for your theory as a whole, science is based on agreement: if someone says s/he is scientific, while everybody else thinks the opposite, s/he can be a misunderstood genius, a fool, or someone who tries to deceive people to gain some kind of advantage (or even just for the hell of it).
++ The everybody else (or rather the other 5) may also be unscientific, fools, or people who try to deceive people to gain some kind of advantage (or even just for the hell of it) or people unable or unwilling to accept certain facts & figures. E.g. while that would lead to a conclusion they dislike. 'I do not want chess to be solved.' -> 'Chess cannot be solved.' -> 'Whoever says differently must be unscientific / a fool / a deceptive person / a liar / an autist.' -> 'Whatever he says must be obviously / clearly / blatantly wrong / a lie / ridiculous...'

"it is not even our responsibility to prove you wrong"
++ It is not my responsibility to prove anything either. I try to the best of my abilities. People can claim without proof. "Chess is a draw", "1 e4 best by test", "King's Gambit loses by force" are Fischer's claims without proof which I believe to be true.

"The burden of proof is on the claimant"
++ The claimant GM Sveshnikov is dead. I kind of load his burden on my shoulders.

"what you call proofs are not proofs for our standards, nor you can prove they are for scientists' standards outside this thread" ++ Assessing the feasibility does not require any proof in the purist sense: plausibility is enough.

"you have not published anything on the subject" ++ No, I have not and I do not intend to. I have never worked in the field of game theory and do not intend to.

"There cannot be with you any real discussion about a theory to provide a proven solution for chess, if there is not even agreement on the criteria to prove something."
++ In this thread there is no agreement on anything: not on definitions, not on the meaning of words in a dictionary, not on facts, not on figures, not on the Laws of Chess. This is a public forum free for all, it requires no scientific degree or chess rating.

@MARattigan
Your post #2949 is apparently a response to my post #2945
a few before.
This one:  https://www.chess.com/forum/view/general/chess-will-never-be-solved-heres-why?cid=68837185&page=148#comment-68837185
A main point about that post of mine was about the relevance of 'strong' to weak.  Strong mathematics.  Exact mathematics.
I brought up the relevance of 13 to the 64th power much earlier in the forum.
That and factorials for ways to have 32 squares empty got a good reception.
The relevance of 'strongly solving' is not just as part of the whole process.
Its also relevant for the purpose of contrast.
To understand what is being done with 'weak solving' by comparing with thorough solving.  Obvious relevance.  'Meaningfulness'.
Also called 'the big picture'.  'View from the balcony'.  

Yes - but there is reason to believe that @tygxc will not concede that error nor any other error he's making.
But its a paradox because he keeps it civil.

Not really a paradox. Just often not so in such circumstances!

Incidentally, the Chinese chap who commented the other day was right to question whether anyone even understands what "solve" means, in the context of chess.

It can't be a "strong solution" since what does that mean? Every possible position or every possible game? And evaluated or not evaluated? To evaluate every possible position is not relevant to a solution for chess in any meaningful sense. We're only interested in meaningful positions: that is, positions reached in the context of at least a strong attempt to play the best moves on either side.

Yet, how are the best moves to be played or identified, without an evaluation function? Such an evaluation function must be heuristic. The only possible way forward is via successive approximations, since a "solution", created by using known values for positions, can't exist before the "solution" and those "known values" are found: and yet it's necessary. So it can only be achieved via successive approximations, using evaluation functions.

Now, the problem with Sveshnikov's idea is that present evaluation functions aren't accurate. That can be seen by the horizon effect, where an engine evaluation changes as it moves forward and yet, quite often, it wouldn't go in the right direction of its own accord. They are just too weak, because the techniques the evaluation functions use are too primitive, as yet. There's no doubt about it. I think everyone here would agree on that.