Chess will never be solved, here's why

Sort:
avram2223
MARattigan wrote:
avram2223 wrote:
Optimissed wrote:
MARattigan wrote:

@Optimissed

Nobody in that subthread appears to be asserting chess is a zugzwang or, apart from yourself, anything else.

Some positions are zugzwangs and most are not. They can't be forced.

Zugzwangs can be forced. If the starting position is a zugzwang, it may not be forced in the conventional sense, but there is no way White can avoid it.

Your argument is false and "for show" only.

This doesn't follow from your preceding statements. If your'e going to assert somebody's argument is false you should be prepared to say where the flaws are. 

In that case it's completely clear that you're attempting to make an argument with the fail-safe that you're not intending to make it. Sorry, I don't believe you. You're known to be dishonest.

Its literally insane the amount of effort and time you guys are wasting on this debate. There are MORE games than we could ever possibly compute in the next 1000 years (as we know). From the STARTING move of Chess it is unknown if White or Black is winning. StockFish 84 might give White a +0.51 advantage, but those are Heuristics, we don't KNOW. The initial position could be a Zugzwang, we don't know, in order to know we would have to parse the ENTIRE tree of Chess games from the starting move. Which we can't do, even with 1000 Quantum Computers. End of Discussion

The initial position could be a Zugzwang, we don't know, in order to know we would have to parse the ENTIRE tree of Chess games from the starting move

Not true. If it's a zugzwang then Black wins. You have to parse the tree only up to the point that you've found a complete half forest (single moves for Black from each node; all moves for White) where each terminal node is a Black mate. If it's a relatively short forced mate, entirely conceivable.

There is also the possiblity of a proof along the lines of the one you first learned for KRvK which doesn't rely on any tree search (but it would obviously be rather more complicated).

I thank you for actually giving a response based in the field this question is in (I am not saying the only discussion has to be scientific, but we shouldn't go around in circles). Now I will be the first to say I am no where near an expert in this field, I am just a comp sci grad, but still in all reality this question is more based on computation than it is any form of Chess "gameplay".

Now to respond, wouldn't finding a "complete half forest" for the game require you to complete the entire tablebase? Essentially the same as generating all the combinations possible akin to my first reply. I could be misunderstanding you because I have not heard the term before, but is that basically saying you only have to parse down the game tree (from start) until you find a string of moves that will lead to a win for either side?

I would have to look more into that, I have not looked into this problem for a while because I am pretty sure it is literally not "worth" discussing (in the sense that the Riemann Hypothesis isn't worth discussing), but I am pretty sure that would not be possible since the only amount of solved games (forced mate without playing BAD moves) are 7-move Tablebases, an entirely minuscule portion of the games possible combination of moves. But Let me know if I am thinking about that incorrectly

MARattigan

in all reality this question is more based on computation than it is any form of Chess "gameplay"

Probably correct, but I don't think the possibility of human mathematical analysis, possibly computer aided should be rejected outright. Not got much further than a few basic endgames so far, but I think that may be lack of application. I agree, "gameplay" is irrelevant because there's no known correlation with theoretically effective play.

wouldn't finding a "complete half forest" for the game require you to complete the entire tablebase?

No. I think SF without any tablebase will find such trees for most mates up to about 6 deep on my old PC in under an hour by selective forward search. (Can't swear to it but that's a different story).

I could be misunderstanding you because I have not heard the term before

Yes, sorry. I made it up.

is that basically saying you only have to parse down the game tree (from start) until you find a string of moves that will lead to a win for either side?

The whole game tree is not searched to the same depth, but you need to find winning responses for every opponent reply.

the only amount of solved games (forced mate without playing BAD moves) are 7-move Tablebases, an entirely minuscule portion of the games possible combination of moves.

Correct so far as publicly available tablebases are concerned. Marc Bourzutschky has produced a small fraction of 8 man DTC tablebases, but not much foreseeable prospect of solving the starting position by tablebases this eternity.

mpaetz
Optimissed wrote:

A zugzwang is a position where the side with the move loses because they must move. The initial position isn't a zugzwang. There's no doubt about that in MY brain. Inferior brains do exist .....

Yet there are also superior brains, or other brains that have more or different information. The "I am superior so I must have the the answer to this question and all those who have other ideas should shut up" argument can elicit nothing but derision.

MEGACHE3SE

i love how tygxc is still acting like basic calculation errors havent been found, multiple times, by multiple people, in his "logic"

and its still funny how he still is refusing to acknowledge the basic errors he makes in his claims that we somehow KNOW that chess is a draw.

he still cant even wrap his head around the basic concept of parity

tygxc

@9184

Meanwhile debaters here still fail to understand the difference between ultra-weakly,
weakly, and strongly solved, and how one is not needed for the other.

Meanwhile 104 perfect games, all draws.
https://www.iccf.com/event?id=100104

tygxc

Prof. Van den Herik was the world's authority on solving games.
GM Sveshnikov was the world's authority on engine analysis.
There is no need for 3 grandmasters:
in the ICCF finals we have 17 ICCF (grand)masters with engines at average 5 days/move.

Elroch

All of them imperfect.

tygxc

@9188

104 games where both participants play optimally.
For the initial position a strategy has been determined to achieve the game theoretic value of a draw against any opposition: follow an ICCF WC draw for as long as possible and then proceed with an engine and an ICCF (grand)master at 5 days per move until a 7-men endgame table base draw or a prior 3-fold repetition is reached.

Elroch
tygxc wrote:

@9188

104 games where both participants play optimally.

They drew, therefore their play was optimal.

Dysfunctional thinking - all you can legitimately justify is some sort of uncertain belief that they played optimally.

There is absolutely no question that both participants are unable to see far enough to avoid the horizon problem. And where both are unreliable, the combination of both of them is unreliable.

In addition, when one game gives you uncertain information, more than one game cannot give you certain information - merely less uncertain information.

Some day you need to learn about uncertainty rather than living your life blind to reality. I recommend E.T. Jaynes - Probability Theory: the Logic of Science.

tygxc

@9190

"both participants are unable to see far enough to avoid the horizon problem"
++ There is no need to see far enough, there is only a need to select among the legal moves a move that does not worsen the game state from draw to loss.

tygxc

@9192

"Chess doesn't work that way" ++ It does. Read Games solved: Now and in the future by Prof. Van den Herik: there is a large section about chess.

"no-one plays the MB in Correspondence chess." ++ For good reason.

"the MB is considered nearly unsound, although it probably doesn't lose."
++ It loses, already Hans Berliner said so.

Elroch
tygxc wrote:

@9190

"both participants are unable to see far enough to avoid the horizon problem"
++ There is no need to see far enough, there is only a need to select among the legal moves a move that does not worsen the game state from draw to loss.

Sloppy thinking. "the game state" assumes a perfect evaluation, which you don't have. Moreover, there is no bound on how badly wrong your evaluation could be. What you do have is an imperfect evaluation which is very useful but not 100% reliable.

At every stage in the history of computer and centaur chess, there have been losses for agents just like those participating now. Next year, one of the participants this year might lose one game to an improved opponent. Such a loss would PROVE the unreliability of that agent.

Elroch
Optimissed wrote:

I disagree with your main thesis, Elroch, that increasing the number of games cannot lead to certainty, since certainty is a judgement, rather than a mathematical depiction. If it were a mathematical depiction, a human judgement concerning it would still have to be formed.

Machines cannot be uncertain or certain of anything, unless they're programmed to depict certainty or uncertainty in a human way and according to criteria defined by human judgement. A computer is a machine.

Of course. You can be certain of anything, without adequate justification. Such certainty can be wrong. Quantified in terms of cross-entropy, this is an infinitely bad mistake. For example, you can be certain that battery acid is a great thing to drink, drink a pint of it based on your certainty and learn that the certainty was unjustified. 

I referred to what could be justified. The relevant class of examples is that that which can be deduced from axioms that are assumed to be true using logical deduction is certain (conditional on the truth of the assumed axioms). In chess, such axioms would be the rules of chess, and a conclusion could be an evaluation of a position as a win, draw or loss. Sometimes this can be achieved, other times not practically. And that is the case for the initial position, as you agree.

MEGACHE3SE
tygxc wrote:

@9184

Meanwhile debaters here still fail to understand the difference between ultra-weakly,
weakly, and strongly solved, and how one is not needed for the other.

Meanwhile 104 perfect games, all draws.
https://www.iccf.com/event?id=100104

wow, funny how you dont address any of what i said in the slightest

none of those games are confirmed to be perfect, what are you smoking?

basically everyone here knows exactly what you are saying, they have just found the holes in your "logic". you are either too stubborn, or too stupid, or both, to admit that you are wrong in so many ways.

MEGACHE3SE

the biggest tragedy in all of this is that tygxc's false claims are now starting to show up on google searches, now people are going to be misled.
dont forget how tygxc's ENTIRE ARGUMENT is based off of a calculation error where he claims that for each of the nodes of a computer that performs 1 move per minute at a supposed 99% accuracy, he claims that EACH of the nodes ALL have a 99% accuracy.

hes literally off in his calcualtions by a factor of over a million.

Ian_Hawke1967

I mean, i have already solved it....

Who do you think created the chess engines that analyse your games on chess.com?

Who do you think made Magnus give up his World Championship title and run away?

Who do you think created this game?

I did!.

MEGACHE3SE

alright Tygxc, how do you strategy steal this position, remember, you claimed that ANY starting position can be strategy stolen.

MEGACHE3SE
MEGACHE3SE wrote:

alright Tygxc, how do you strategy steal this position, remember, you claimed that ANY starting position can be strategy stolen.

btw tygxc, evne your example of a position that could be strategy stolen doesnt work either. you forced black to move the pawn 2 spaces when it doesnt have to.

tygxc

@9204

"If I come across the Herik book"

++ It is no book, it is a scientific article.
https://www.sciencedirect.com/science/article/pii/S0004370201001527

tygxc

@9195

"the game state assumes a perfect evaluation" ++ Yes.

"which you don't have" ++ I have the 7-men endgame table bases with their perfect evaluation.
White tried to win, black tried to draw. They reach a 7-men endgame table base draw or a prior 3-fold repetition. Black succeeded, white failed. All black's moves are justified in retrospect.

"What you do have is an imperfect evaluation which is very useful but not 100% reliable."
++ The perfect evaluation comes from reaching the 7-men endgame table base, or a prior 3-fold repetition.

"there have been losses for agents just like those participating now"
++ Yes, but each year fewer and fewer. Now 100% draws.