Chess will never be solved, here's why

Han solo would NEVER lie...

Vegas isn't that far away.

I used the shortest named well-known star that jumped to mind (no Sol jokes, thanks in advance)...sub in Rigel if you prefer .

Vegas isn't that far away.

@12527

"If @tygxc would be so kind as to produce a flowchart or pseudocode for something that might remotely be a weak solution of competition rules chess in principle"

++Here is how the weak solution is being done. 17 ICCF (grand)masters each with twin servers of 90 million positions/s each, at average 5 days / move.

That's a farce, not a flowchart. 17 grandmasters producing 1 ply each per 5 days means it would take you over 8000 years to produce the number of moves in the partial saved checkers database even assuming there were no overlapping moves. The difference would be you wouldn't have a clue which were right and which were wrong.

"we could say whether the issues of the 50M/3R rules arise or not."

++ No, there are no issues with the 50-moves rule: it never triggers.

That's because if they can't see anything to do they agree a draw instead.

The perfect games draw in 15 to 73 moves, average 40 with standard deviation 11.

Yes, the 3-fold repetition rule is a vital mechanism to achieve the game-theoretic value:
it occurs in 37 of the perfect games. Most of these are sacrifices to deliver a perpetual check.

And you start your "calculations" with Tromp's basic rules positions which take no account of it.

@12560

"17 grandmasters producing 1 ply each per 5 days means it would take you over 8000 years to produce the number of moves in the partial saved checkers database"
++ The 17 ICCF (grand)masters run 3,060 CPU for 2 years to consider 10^17 positions.

So where are they? Stockfish considers a large number of positions but not in any way that will produce a solution. You only get 17 moves (which may or may not be correct) in 5 days to add to your "solution".

Schaeffer ran 50 CPU for 2 years to consider 10^14 positions to weakly solve Checkers.
The storage of 10^7 positions has nothing to do with it.

Well I can agree with that. It leaves 2 minutes waiting for a computer to redo the last bits of the calculation. You would need to keep a much bigger percentage of a VASTLY bigger solution to give the same response time if you had a method for finding a solution.

"which were right and which were wrong"
++ The absense of decisive games implies the absense of errors (?), and that implies the absense of error pairs (?)(?) i.e. error (?) + missed win (?). That the games end in certain draws retroactively justifies all black moves as fit to draw and all white moves as unfit to win.

Complete heap of garbage. You've been shown lots of sets of 100% SFvSF draws from winning tablebased positions and invited to do your Poisson about so we could check it with Syzygy. Conspicuously refused all invitations.

You have no idea whether perfect play from the starting position would produce a decisive result or not.

What you're actually proposing is that if we wait some vast multiple of 8000 years and if you're still alive you can announce that your big red telephone tells you you've got a solution. You might by then have enough moves for a solution but you still won't actually know which moves are correct and which not, which is a bit of drawback in a solution.

Not very impressive.

As soon as all reasonable white moves are exhausted Chess is weakly solved.

Is it Stockfish rings you up to check if its moves are reasonable or the players?

"if they can't see anything to do they agree a draw instead" ++ This is disrespectful.
They play drawn positions a long time, trying to squeeze something out of nothing.

It's not disrespectful; I think SF is a superb programming effort. It's just realistic from comparing what it does in simpler tablebased positions - it can't see how to win in a lot of those either, so quite often it'll just show 0.00 in a winning position.

If a player has no hope to avoid checkmate, then he resigns.
If both players have no hope to checkmate, then they agree on a draw.

Yes - in preference to playing on to a draw under the 50 move rule. But both Stockfishes having no hope to checkmate doesn't mean the position is actually a draw. The same happens, for example, in winning KNNKP or KQKNN positions with just five men on the board. (The same mostly happens with humans of any ELO too.)

"start your calculations with Tromp's basic rules positions which take no account of it."
++ I start estimation from Gourion's positions. A position is defined by 9.2.

I don't mind if you want to use the term that way so long as you don't pretend your positions determine possible forward play under competition rules or correspond to the nodes searched by Stockfish and quoted as nodes/sec (or to your other definition FEN-move#).

Gourion's diagrams with side to play added are an even less appropriate starting point because they omit an unknown percentage of positions with promotions.

When a position is repeated 3 times it is a draw.

With one of your definitions (the one above). With mine and your other definition, when 3 positions that are considered the same according to 9.2.3 have occurred it's a draw. But that's just arguing about words. The salient point is that the nodes/sec figures you quote are not positions/sec with your definition, not by a very long way. The other salient point is that your (cough) flowchart doesn't bear any relation to the nonsense figures you keep cut and pasting.

"Maybe solve it with a scientific proof rather than a mathematical one."
++ Mathematical proofs are much wider than the narrow meaning some here attach to it.

actually they arent lmfao, you just have no idea what a math proof is. statistical inferences/estimations are not math proofs.

A mathematical proof does not need to be deductive. It can be inductive like Tromp's method to establish the number of legal positions or like Monte Carlo Methods.

wow, i see you struggle with basic definitions too. calling something "inductive" doesnt make it inductive. tromp's method isnt a mathematical proof. it's a statistical estimation by definition. 

'Next to brute-force methods it is often beneficial to incorporate knowledge-based
methods in game-solving programs.' page 303 section 5.2Game knowledge ≠ assuming conventions as fact. just because you struggle with basic definitions doesnt mean you can just lie about the ones you dont understand, tygxc.

Example of chess knowledge is this paper.
The only human input are the axioms: Laws of Chess

actually no, the human input includes the lack of full analysis, the statistical preferences, and limited search space. so you are just lying now. this entire claim is just ASSUMING a chess engine is deductive.

By trillions of boolean operations conclusions are reached: theorems.

statistical inferences ≠ theorems. Very basic mistake you continue to make due to your complete lack of education and critical thinking.

@12512

"For comparison, I have a first in maths from Cambridge, and an MMath and a lot of experience of applications of mathematics (as well as physics and computational techniques."
++ So more than you, as I thought.

"I respect those who stuck it out longer than me" ++ Show respect then.

"a mathematical deduction and a statistical inference"
++ Mathematics is not only deductive.

Mathematical truth can only be arrived at deductively. No excuses.

Tromp's method to establish (4.82 +- 0.03) * 10^44 legal positions is inductive.

This is a statistical estimate from a random sample, with uncertainty, not a mathematical fact. Do you not know the difference? Do you think every political poll is a mathematical fact about the population? (They use EXACTLY the same technique Tromp used).

A statistical estimate is not a definite fact. It is not exact, and it can be wrong. There is a low but non-zero probability of it being very wrong.

Monte Carlo methods are inductive.

Monte Carlo methods are likewise used to find statistical estimates. I have used them for Bayesian inference. They never give you a mathematical fact.

@12549

"Maybe solve it with a scientific proof rather than a mathematical one."
++ Mathematical proofs are much wider than the narrow meaning some here attach to it.

No, they are not. You seem never to have learnt the difference between a mathematical theorem and a statistical estimate. The latter is entirely in place in science - fundamentally all of science falls into this category, when examined very precisely - and can never be used in a mathematical proof.

The key point is that a mathematical theorem is certain. A statistical estimate is uncertain. And let me underline the fact that you have no valid statistical estimate relating to the solution of chess.

A mathematical proof does not need to be deductive. It can be inductive like Tromp's method to establish the number of legal positions or like Monte Carlo Methods.

Wrong. See my other post for an explanation of your misunderstanding and learn something.

'Next to brute-force methods it is often beneficial to incorporate knowledge-based
methods in game-solving programs.' page 303 section 5.2

It is. Using a chess/checkers engine to guide the construction of a weak solution is very useful. But your blunder is in thinking it allows you get away with not constructing a weak solution at all. There is no excuse for you not understanding this now. If you construct a supposed weak strategy for white and there is a single legal black move that leads to a position you have not analysed, you don't have a weak strategy.

Example of chess knowledge is this paper.
The only human input are the axioms: Laws of Chess

By trillions of boolean operations conclusions are reached: theorems.

Now read the big text slowly.

Tygxc calling tromp's work a mathematical proof is hilarious when tromp literally writes the probability that his work is wrong LOL

@12570

"the probability that chess is not a draw is XXX"
++ The probability that chess is not a draw is 1/117^116 = 10^-240.

@12570

"the probability that chess is not a draw is XXX"
++ The probability that chess is not a draw is 1/117^116 = 10^-240.