Chess will never be solved, here's why

"Maybe maybe not its impossible to try and guess at what we will be capable of in 1000 years"
My point is that it doesn't have to be somebody 'cleverer'.
If the hardware and software become strong enough and the money and will are there because society survives well enough and long enough - then yes the very unimportant project of completely 'solving' all of chess might be completed.

long after we die.

and ur point ?...and its negative ?

A quantum computer can perform this operation in parallel: given an array of positions P(k) in one operation gives all arrays of positions that could reach P(k) in one legal move.

Thx Ty !...its gonna be beautiful . were well on our way and we must face & embrace.

'long after we die' looks like an adverbial phrase qualifying a comment.
many comments don't have a 'point' so to speak.
They're comments rather than 'arguments'.

completely 'solving' all of chess might be completed.

not might...will. trust me. only cuz chess is a very teensie microslice of a very much larger thingy. a thingy we cant see right now cuz were staring at 64 squares. but its panoramic. more like 64000 acres. a 100 sq miles. am i right ?...maybe not. but then ima embracer. its one a the few things i like abt myself.

summa my friends like cookie & babber are statists. so convincing them is like......nvm.

well...MY point is whats so bad abt dying. were all gonna do it right ? i mean feces stops making contact w/da propeller & one makes a clean break. and fyi ?...i told april to bury me in april. lemme winter on dis gawt4saken rock. only cuz im dying one a these 10.31's (halloween nite).

No need to get stuck with definitions, keep the discussion practical. If the move is trash, no need to have a calculated draw or win in that position.

It's a good approach but the problem is where to draw the line, how do we define a definitely losing move to count them out.

For example I don't believe that you disregard theoretically bad opening moves like 1. A4 - Simply because without a calculated (solved) line, we can't count out that strong but imperfect engine play might have a chance to result in a different result from that position.

I agree regarding 1) a4.
But no white move of white's opening 20 options can be discarded.
Try a4 b5 axb.
Black's first move is horrendous. He is now down a pawn and white has an extra b-pawn that will probably be easy to protect.
That extra pawn is also cramping black's development.
Plus black has allowed his a-pawn to be isolated on a half-open file ...
plus black's horrible b5 reply allows white to half-open the a-file for his a1 rook.
White is already winning. But does he have a forced win?
The short answer is No.
Can we discard the terrible b5 reply for black?
Depends on the degree to which 'solving' is to be approximated.
Solving from the 'front' is kind of doomed. Because of the enormous permutations of 'game trees' when attempting that kind of 'solving'.

Yes, it's a "safe bet" that 1. a4 b5 is not winning for black. But it cannot be deduced that this is so without a great deal of computation, absurd as the alternative seems. Humans are not good at dealing intuitively with the difference between impossibility and extremely unlikely. Maybe there is 1 in 10^50 chance that 1. a4 b5 wins for black. It is conceivable in a pedantic sense but all our experience says to to reject it based on uncertain inductive reasoning.

Not ignoring such extremes seems more a matter of rigor and discipline than genuine doubt. But it is no more so than observing that a uniformly random natural number in the range [1, N] might be 1, regardless of how big N is. The probability is (obviously) never zero.

There is no indication that the most efficient solution of chess would be much different to that of checkers - computation split fairly evenly between a big tablebase constructed with retrograde analysis and a proof tree constructed using forward analysis to verify two explicit drawing strategies, the two meeting in the middle.

Interestingly, two minimal strategies (each just giving one optimal move at each choice) would only give one game up to where they hit the table base. The tablebase, being a strong solution of smaller subgames, would then provide some number of optimal ways the game could be finished. The number of ways depends only on the single point where the combination of both strategies hits the tablebase!

It's an interesting question how big the tablebase would be. It could be, say, 20 pieces.

The notion that 1) a4 b5 2) axb could lead to a 'forced win for BLACK' is hard to stomach. Yes.
But I've known about mathematical rigor for decades.
You're right Elroch. We can't just dismiss it 'computationally'.
Intuition Screams 'Black force a win from There?? That's got to be Nuts!!'
And chess opening books probably won't even mention such a notion nor that variation itself.
--------------------------
We won't find a lot of tomes titled 'How to win at chess with 1) a4!'
How should black respond?
Chess instructors would probably say - black just proceeds on principle. He is now essentially playing with first move and could almost play as white would.
He can 'open' with any of the four most popular moves that white plays?
e5 or d5 or Nf6 or c5.
I don't like c5 there. Not with white's a-pawn at a4.
I like Nf6 there.

I was curious so I ran 1) a4 on Stockfish just now.
'Fish' liked e5 as best reply for black with advantage of 0.13 for black.
Nf6 got 0.04 adv for black.
And 'Fish' found 0.00 if black replies d5.
That's the 0.00 you were talking about Elroch.
Fish also found 0.00 if black replied c5 and white then played e4.
a4 b5 axb5 got 0.68 for white advantage.

I think there could be greater effort still to "solve chess". I think in computer vs computer chess there could be something like pairs of opening books and calculation.

tygxc is claiming the mathematical definition of solving a game, which requires even the trash moves of at least one side to be fully calculated out. tygxc syas we can just reject moves based on if they seem bad.

player !...imagine mining for bitcoins using a Q-com ?...or for some typa proof of work/SHA-256 guesser.

TGL
I hopeth thou dost findeth El Dorado by such Quest.
And liveth longeth and doth Prosper.

@11864

"Maybe there is 1 in 10^50 chance that 1. a4 b5 wins for black."
++ This shows you do not understand probability.
1 a4 b5 like any position is either a white win, or a draw, or a black win.
Probability is always tied to an experiment, either a real experiment or a thought_experiment. You cannot look into 10^50 parallel universes and find one where 1 a4 b5 is a black win.

Likewise chess is either a draw, or a white win, or a black win.
It cannot be 'a high probability to be a draw'. You have no related experiment.

Likewise a cat is either dead or alive.
Schrödinger's cat can be 33% alive, 67% dead,
meaning if you open 100 boxes you find 33 cats alive and 67 dead.

Figure 2a shows 10,000 AlphaZero autoplay games at 1 s/move
yield 88.2% draws, 7.72% white wins and 4.09% black wins.
Likewise Figure 2b shows 1,000 AlphaZero autoplay games at 1 min/move yield
97.9% draws, 1.8% white wins and 0.3% black wins.

From that we can extrapolate that AlphaZero autoplay games at unlimited time yield 100% draws, 0% white wins and 0% black wins.
Chess being a finite game, unlimited time yields perfect play, independent of the AlphaZero provisional heuristic evaluation. At unlimited time only the legal move generator counts.

The present 112 draws out of 112 games in the WC33/final, World Championship 33 Final are also perfect games with optimal play from both sides.
This is at least part of a weak solution of Chess.
It is redundant, but not yet complete. Redundancy makes it fail safe.
Even if double errors were found in a few games, there remain perfect games.

The most plausible distribution of errors is 112 - 0 - 0 - 0.
It could be that it is 110 - 0 - 2 - 0, but redundancy allows to ditch the 2 erroneous games.
It is not plausible 76 - 0 - 46 - 0: all errors would have to come in pairs.
It is not plausible 0 - 112 - 0 - 0: all 17 players would have to collude to make exactly 1 error.
As long as we have even 1 perfect game with 0 error, chess is ultra-weakly solved and a draw.

On the contrary, with all due respect, you reveal that you lack a basic understanding of the hugely useful technique of Bayesian probability. Part of the blame for this might be that you have never learnt anything about it but have learnt something about frequentist probability and that you think that is the only type of probability. Both forms of probability are useful, but the Bayesian form is more general and philosophically important.

Bayesian probability is the quantification of belief. It quantifies uncertain knowledge.

Let me give you a simple example which shows how inadequate your expressed view is. Suppose a robot flips a fair coin into a box, puts the lid on and gives it to you. What is your belief about the state of the coin flip? Is it a head? Is it a tail?

Personally, my belief would be that there is a 50:50 chance that the coin is heads. My Bayesian belief state is a probability of 0.5 that the coin is heads and 0.5 that the coin is tails. It doesn't matter a damn to my belief that the coin flip has already taken place: what matters is my state of knowledge.

You might say: "the state of the coin is fixed, so the only valid beliefs are to be sure it is heads or to be sure it is tails". This would be wrong. Uncertain belief is appropriate for me, and it is appropriate for you in the same situation.

Now suppose instead there is a lottery ticket in the box, for a lottery that has just happened. The numbers have been generated and it's time for people to go cash in their winnings. Someone bought you the ticket as a present and gave it to you in the box saying to open it after the numbers had been called.

You would presumably take the position "the ticket is either definitely a winner or a loser at this point so my belief has to be certain too. Since almost all tickets lose, I can be certain the ticket has lost".

I would instead say something like "1 in a million tickets wins, so my Bayesian belief is that there is a 0.0001% chance that the ticket in the box is a winner"

Note that there is an important point about Bayesian belief that it depends on your state of knowledge even at a single time. With the lottery ticket or the coin, it is perfectly reasonable for someone who has seen the contents of the box to be certain while you are uncertain. Both are correct Bayesian beliefs.

No-one has solved chess. No strong, weak or ultra-weak solution exists. There are merely people who express inappropriate certainty, like someone who says he is certain that a lottery ticket in a box hasn't won despite not knowing what the number is.

One thing that confuses you is that chess is a single specific game. But that is no different to it being a single specific lottery ticket in the box. Bayesian probability works with one off events. You can do useful quantitative reasoning about them. A frequentist would think that you need a population to make a probabilistic statement about chess meaningful. We could construct one by generating a huge number of games rather like chess.

One way to generate such games would be to throw the pieces randomly on the board and then do a computer evaluation to check if the position seems balanced (we could demand an evaluation within 0.2 of 0, say). Each apparently balanced position defines a new game, starting from that position. We could generate say 10^30 such games. All have a similar game theoretic status as chess.

It is highly likely that some of these games are winning for white and some are winning for black (the evaluation was that of the latest fallible chess engine). It is highly likely that in some of these games, a dumb looking sacrifice by one of the players on move 2 turns out to be good. Maybe one in a trillion dumb looking sacrifices (according to some definition) wins. So rare that all experience would tell us dumb looking sacrifices are bad, while it is only usually so.

Difficult point: we don't know for sure that normal chess is not one of these flukes.