Chess will never be solved, here's why

Sort:
playerafar

If the researchers and promoters wanted to find out more about their computers that they're willing to publish (which could risk funding)
They could play their strongest computers against each other - but make a time difference between them - that's lopsided.
Give one a day a move. Give the other ten days per move.
See how big the difference has to be to get rid of 'all draws'.
Try it other ways too. Give one an hour per move. Give the other a week.
And so on.
-----------
They've probably done lots of experiments like that.
Doesn't mean they've published the results.
Is there some kind of 'magic' at five days per move each?
Maybe in ten years from now - the computers will draw each other at ten days per move each. Or one day per move each.
It doesn't prove they played perfectly.
Never will.
Bottom line. That's right. Coming up.
Chess is not solved.

tygxc

Meanwhile the ICCF WC Finals reached 116 perfect games with optimal play from both sides.
This is at least part of a weak solution of Chess: it shows Chess is a draw and it shows how to achieve the draw.

MEGACHE3SE
tygxc wrote:

Meanwhile the ICCF WC Finals reached 116 perfect games with optimal play from both sides.
This is at least part of a weak solution of Chess: it shows Chess is a draw and it shows how to achieve the draw.

wow, i didnt know random games that dont explore alternative routes were absolute mathematical proof.

tygxc has already admitted that he has no proof that any of the games are perfect.

in addition, tygxc still cant mathematically prove that chess is a draw, despite his claim of being able to.

MEGACHE3SE

tygxc, earlier you claimed that 99% was mathematical certainty. want to elaborate on that?

7zx

There isn't an absolute mathematical proof that human activity is causing global warming, or that vaccines are safe and effective, or that smoking causes cancer. Yet people believe all those things.

Why demand a higher standard of proof for admitting that chess is a draw than for anything else?

Corn-In-Mochi

Cool forum

MEGACHE3SE
7zx wrote:

There isn't an absolute mathematical proof that human activity is causing global warming, or that vaccines are safe and effective, or that smoking causes cancer. Yet people believe all those things.

Why demand a higher standard of proof for admitting that chess is a draw than for anything else?

because a game solution is a mathematical object, its literal definition requires a mathematical proof of its existence. the scientific process of belief and evidence is an entirely different process than the logical deduction of math.

tygxc doesnt really understand that, he claims certain mathematical properties but only justifies them with the circumstantial evidence akin to conventional persuasion (this is tygxc at his best, often times tygxc is infamous for also just making stuff up based on his opinions and referring to them as certain facts).

Elroch
7zx wrote:

There isn't an absolute mathematical proof that human activity is causing global warming, or that vaccines are safe and effective, or that smoking causes cancer. Yet people believe all those things.

You are confusing the entirely separate domains of science - which deals with the natural world and relies on observations that are always incomplete and uncertain (to some extent) - and the mathematical sciences - which deal with abstract truth.

Questions about science are dealt with by applying the scientific method, while mathematical questions are dealt with by rigorous deduction. The solving of checkers and other games are successes in the latter category, while the solution of chess is beyond forseeable computational capability.

Why demand a higher standard of proof for admitting that chess is a draw than for anything else?

As the above explains, it's the SAME standard of proof as similar questions, and all questions of the same very general type. That being said, while it is impossible to truly solve chess using empirical methods (all those referred to by @tygxc) it is in principle possible for such methods to support a level of belief (which always has uncertainty) in the optimum value of any given chess position, including the first. This is most of what playing chess consists of.

But note that chess has been believed to be a draw for a long time. All further "scientific" investigation can do is to strengthen or weaken this belief. It can never turn it into certainty. That is the job of rigorous solution.

tygxc

@12301

"Why demand a higher standard of proof for admitting that chess is a draw?"
++ Ivory tower thinkers like Elroch prefer an agnostic opinion of not knowing by disregarding all evidence and a puristic view about what he accepts or not as proof.

Chess is a draw. White has the initiative, an advantage of +1 tempo = +1/3 pawn,
not enough to win, as each further move dilutes the advantage.

The now 116 perfect games with optimal play by both sides in the ongoing ICCF World Championship Finals of 17 ICCF (grand)masters with twin engines of each 90 million positions/s at average 5 days/move show chess is a draw and show how to achieve the draw.

P(B misses win| A has erred) < P(A has erred) for 2 reasons:

  1. B has more information: the move played by A and thus looks 1 ply deeper
  2. A is more likely to err after a short thinking time e.g. 2 days/move, and B is more likely to find the win using a normal 5 days/move or a long 10 days/move when he suspects an error,
    e.g. while B did not expect the move by A.

Thus P(double error) = P(A errs and B misses win) = P(A has erred)*P(B misses win|A has erred) < P(A has erred)*P(A has erred) = P²(A has erred) = P²(single error)

Assume game 117 will be a clean win, no clerical error or due to illness.
Then P(single error) < 1/117
Thus P(double error) < 1/117² = 0.007%
Thus the 116 ICCF WC Finals games are > 99.993% certain to be perfect games with optimal play by both sides.

The probability that Chess is not a draw and thus all 116 games contain an error is
P(single error)^116 < 1/117^116 = 10^-240 or in Schaeffer's words: vanishingly small.

Elroch
tygxc wrote:

@12301

"Why demand a higher standard of proof for admitting that chess is a draw?"
++ Ivory tower thinkers like [Schaeffer and all others with peer-reviewed publications about solving games] prefer an agnostic [wrong word] opinion of not knowing by disregarding all evidence and a puristic view about what he accepts or not as proof.

The correct statement is that they know what solving a game is and you don't.

It's a simple matter of domains. "Scientifically", chess has been strongly believed to be a draw for a century or so. All you can do is get a little more or less sure. Sveshnikov did not aim to do significantly more. You don't claim to have done any more (apart from not understanding the difference between strong belief and certainty).

Mathematically, while checkers has been proven to be a draw, chess has not, and will not be in the forseeable future.

Your key blunder is in indicating that all those who publish in the field are wrong - with all due respect, a characteristic of a crackpot.

Also, repeating an argument containing specific fatal flaws that have been pointed out to you is, likewise, crackpot behaviour.

tygxc

@12306

"they know what solving a game is and you don't" ++ I know, I follow Prof. van den Herik

"Sveshnikov did not aim to do significantly more."
++ He did: he aimed to trace all openings to technical endgames and solve chess.
The 17 ICCF (grand)masters are the good assistants he asked for,
and their twin servers each 90 million positions/s are the modern computers he asked for.

"understanding the difference between strong belief and certainty"
++ I even quantified the certainty: < 10^-240 probability chess is no draw,
and < 0.007% probability one of the 116 games contains a double error.

"checkers has been proven to be a draw, chess has not"
++ I gave compelling arguments, and the 116 perfect games prove it.

"all those who publish in the field are wrong" ++ That is not what I say. Van den Herik wrote incorporating game knowledge is beneficial in solving a game, which you deny.

"an argument containing specific fatal flaws" ++ None have been pointed out at all. Rather:
I do not understand -> not proven to my satisfaction -> not proven -> not true -> fatal flaw

"crackpot behaviour" ++ Insulting is no accepted way of proving or disproving.

MaetsNori
Kygo_Garrix_Script wrote:

After accessing all your comments. I find the winner to be tygxc.. Thank you

Hmmm ... I have my own issues with the ICCF discussion (a circular debate that seems to go around and around), but aside from those - declaring that chess is a draw because the current ICCF WC is all draws isn't really "solving" chess ...

Solving chess means knowing how many exact moves to win/loss/draw from any possible position ... without any calculation or "engine pondering" needed.

tygxc

@12309

"Solving chess means knowing how many exact moves to win/loss/draw from any possible position" ++ That would be strongly solving Chess,
we are discussing weakly solving chess, as e.g. Schaeffer did for Checkers.

Ultra-weakly solved means that the game-theoretic value of the initial position has been determined,
weakly solved means that for the initial position a strategy has been determined to achieve the game-theoretic value against any opposition,
and strongly solved is being used for a game for which such a strategy has been determined for all legal positions

Elroch
tygxc wrote:

@12309

"Solving chess means knowing how many exact moves to win/loss/draw from any possible position" ++ That would be strongly solving Chess,
we are discussing weakly solving chess, as e.g. Schaeffer did for Checkers.

Ultra-weakly solved means that the game-theoretic value of the initial position has been determined.

Where "determined" means rigorously proven. This is important and, I believe, not understood by you.

weakly solved means that for the initial position a strategy has been determined to achieve the game-theoretic value against any opposition,

No, if the result is not a win, it means a complete strategy has been exhibited for each side (i.e. it will tell you a move to play in any position that can be reached while applying it) with both strategies proven to achieve the optimal result.

and strongly solved is being used [more precisely, defined] for a game for which such a strategy has been determined for all legal positions

This one is ok.

Thee_Ghostess_Lola

"...solving chess mean never having to say you're in checkmate"

"...if you love chess ?...let it go. if it doesnt come back to you ?...it was never meant to be solved"

Elroch
tygxc wrote:

@12306

"they know what solving a game is and you don't" ++ I know, I follow Prof. van den Herik

If he gave a course on the subject, I regret to tell you you would fail it. Your errors of comprehension are very extreme.

Thee_Ghostess_Lola

distance to mate.

one day we might solve chess & discover the ohhh-so-elusive minimum 'distance to mate'...tho "distance" feels clunky.

Elroch

You can see it within any tablebase. We won't see it in general in our lifetimes!

Kotshmot
tygxc wrote:

@12301

P(B misses win| A has erred) < P(A has erred) for 2 reasons:

  1. B has more information: the move played by A and thus looks 1 ply deeper
  2. A is more likely to err after a short thinking time e.g. 2 days/move, and B is more likely to find the win using a normal 5 days/move or a long 10 days/move when he suspects an error,
    e.g. while B did not expect the move by A.

Thus P(double error) = P(A errs and B misses win) = P(A has erred)*P(B misses win|A has erred) < P(A has erred)*P(A has erred) = P²(A has erred) = P²(single error)

Assume game 117 will be a clean win, no clerical error or due to illness.
Then P(single error) < 1/117
Thus P(double error) < 1/117² = 0.007%
Thus the 116 ICCF WC Finals games are > 99.993% certain to be perfect games with optimal play by both sides.

The probability that Chess is not a draw and thus all 116 games contain an error is
P(single error)^116 < 1/117^116 = 10^-240 or in Schaeffer's words: vanishingly small.

The argument structure is flawed, is missing some parts and leads to a conclusion that cannot be reached logically.

"P(B misses win| A has erred) < P(A has erred)"

Better:

P(A has erred) = P(B misses win| A has erred) + P(B plays a flawless game| A has erred)

And

P(B misses win| A has erred) < P(B plays a flawless game| A has erred)

Which conclusion you reach by saying

"B has more information: the move played by A and thus looks 1 ply deeper

A is more likely to err after a short thinking time e.g. 2 days/move, and B is more likely to find the win using a normal 5 days/move or a long 10 days/move when he suspects an error,

e.g. while B did not expect the move by A."

In vast majority of cases where an error were to occur the difference isn't going to be 1 ply but much more than that.

Straight after a move is made, an error suspicion or an unexpected move would almost never happen due to similar analysis tools and same or less amount of analysis on the position. They would have to use more time than their opponent did on their turn, which in reality sometimes would be more and sometimes less. Further analysis would be needed on their time usage habits to support the time factor in this argument.

"Thus P(double error) = P(A errs and B misses win) = P(A has erred)*P(B misses win|A has erred) < P(A has erred)*P(A has erred) = P²(A has erred) = P²(single error)"

Even if you were completely right with the previous points which I counter argumented, this conclusion does not follow logically. More specifically this: "P(single error) < 1/117Thus P(double error)

< 1/117² = 0.007%"

Because even if you're right that P(B misses win| A has erred) < P(B plays a flawless game| A has erred)

All that means is single error game probability is higher than double error game probability. It doesn't mean double error game probability = single error game probability². That is an assumption that isn't even based on your argument.

If the argument you made is right, it would simply mean double error probability < 1/117. For example if game 117 is decisive, a possible probability for a double error game could be 1/130 instead of your supposed 1/117².

You need new better arguments to claim double error probability is 1/117².

mpaetz
7zx wrote:

There isn't an absolute mathematical proof that human activity is causing global warming, or that vaccines are safe and effective, or that smoking causes cancer. Yet people believe all those things.

Why demand a higher standard of proof for admitting that chess is a draw than for anything else?

These other matters you mention are practical problems that can cause severe repercussions for individuals and society. When there is a great deal of evidence that these propositions are likely true, people might choose to alter their behavior to stave off the likely negative results.

The subject of this forum is whether or not we will ever be 100% certain that chess is a win or draw. Yes, that does require a higher standard of proof.