True or false? Chess will never be solved! why?

Solved...The problem being what exactly?

tmodel, you are right, but that's not the definition of "solve" that was used in the context of that thread... Chess will be "solved" by computers within -say- ten years in that definition. But the question is whether we can generate a database of all legal positions (including very weird stuff like 1.a3 a6 2.a4 a5), and obviously the answer is different.

Well this is the crux of your misunderstanding right here.  You're not clear on what solving chess actually is, and that's because you're not clear on what "best play" actually is.

You must be 100% certain that the move selected is best play.  Without an exhaustive analysis through all of the subtrees you may be able to get close, but your certainty will always be something less than 100%.

I'd refer you back (again) to my post #82 for the objective definitions of different strengths of solution.

Have you done the math? The reason I ask is that not all "exponentially"'s are equivalent. Do we need to ask:

1) How quickly does the size (or calculation time) of a tablebase grow with number of pieces?

2) Assuming we can extrapolate, how fast will mankind's computing power (available to apply to this problem) grow in the next 100 years?

No, not at all.  This seems a really common misconception.

Solved means that we know what the outcome of the game is with perfect play (ultra-weak), and have an algorithm that can produce perfet play to obtain the most favourable outcome from the starting position even against perfect play from the opponent (weak) or from any other position (strong).

The problem is really, really strong play, inhuman strength even, is not necessarily perfect play.  In order to know that you've achieived perfect play you must have 100% certainty that no variations or lines you might go down are suboptimal, meaning you must inspect them all.

Quite! It's the sort of calculation that would be better run in the quantum-superdupercomputer screensaver, when there's nothing more important to calculate.

The evidence is believed to be overwhelming. I know it's a draw and I'm not even a good player!

Especially since the galaxies are rushing away from each other. Communication between different parts of a super-galaxy computing system is going to be a problem.

What a bizarre definition of solved.  ???

I read (and acknowledged) your previous references to an acceptable solution - it's just that I disagree with your definition of "solved".

There are "laws" in chess (e.g. pawns cannot move backward) and there are "theories" in chess (e.g. it is best not to expose your king in most situations) that we all accept.  It would be a waste of resources, for instance, to prove that with best play white would beat black if black only moved Nf6 on odd moves and Ng8 on even moves from the starting position.  While this is an unproved theory, we know it to be true.  You can test it in Live Chess if you feel otherwise.

Just like I am confident that gravitational theory is true although I cannot test every situation, I am confident that a solution to chess - a device that could only lose to a stronger (non-human) device will be developed and probably already exists.

My problem with your argument is this:  You want to demand an exhaustive solution to a finite problem while citing the limitations of a finite (possibly infinite) universe and time continuum, but you will not acknowledge that an effectively exhaustive solution already exists.

No doubt, you will argue that my solution is "ultra weak" or less.  Who cares?  You would still lose to a strong computer as would I.  Other threads on this site talk about the FM who was banned purely under suspicion that he had computer assistance in obtaining a GM norm.

Ergo, chess is already solved.

chess is already solved, god is all knowing so he already has the solution to chess. we just need to wait till he tells someone.

If chess is theoretically a win (which I think is unlikely), it would be easier to weakly or ultra-weakly (but not strongly) solve it than if it is a theoretical draw. Even if it is a theoretical win, I don't think an ultra-weak solution would be coming any time soon.

Actually, the math holds, it's the assumption that doesn't:

Some pedestrian calculations:

The Titan supercomputer performs at 10 PetaFLOPs (10^16 FLOPs):

http://en.wikipedia.org/wiki/Titan_(supercomputer)

Assuming one FLOP per position (10^43 in total, and a gross understatement of the calculation requirements because you must evaluate each of the leagl moves in any given position, and each evaluation will take many FLOPs, but let's play along for argument's sake since it won't change the order of magnitude appreciably), it will take 10^ 27 (10^[43-16]) seconds or 3.1*10^19 years (That's 31 million-trillion!) with today's strongest supercomputer.

OK, that's celarly more time than we have.  So let's extrapolate out 100 years, assuming Moore's law holds and computational power doubles every 18 months.  100 years from now, we should have computers that run 10^18 times as fast as the ones we have today (10^34 FLOPs in fact: 10^[16+18]), and should be able to perform 10^43 calculations in 10^9 seconds -- 32 years.

So where's the problem, this actually seems quite acheivable by comparison (and in fact, waiting a few years should bring the problem down to seconds or even fractions thereof).

The problem is in the assumption that Moore's law will continue to hold for any appreciable amount of time.  We hit the physicial limit for the size of transistors some time back (the metric Moore's law actually cited), and have only been buying ourselves additional computational power by throwing other tricks at the problem (larger, faster caches, pipelining, multi-core processors, multiple processors etc.).  Our new bag of tricks is going to run out fairly quickly here, and we will hit a plateau in Moore's law long, long before 2113.  The physics dictate it.

A fundamentally new type of computing (quantum computing) may hold some answers, but at best it will probably only be able to offer us an Ultra-Weak solution.

This simply makes you wrong.  This is not a subjective set of definitions up for debate, these are the objective mathematical definitions of what it means to solve a game.

Your new circuits will be organic

It's not the definition of anyone here, it's the definition from game theory. I don't want to take the time to find a reliable source, so I'll just quote Wikipedia: "A solved game is a game whose outcome (win, lose, or draw) can be correctly predicted from any position, given that both players play perfectly." (Actually the "from any position" makes this the definition of a strong solution, an ultra-weak solution just means proving whether the first player wins, draws, or loses from the opening position with perfect play.)

the grobe but that is only if moores law is correct. I mean the house prices in america also didnt rise forever.

Yes, but a good centaur will still beat a strong computer - does that not mean it's not solved, even by your definition?

What you're saying is it's a practical solution... i.e. for every position a satisfactorily strong move can be suggested by a computer such that over a course of a game the computer would always win.

Even this isn't a very good definition in my opinion because a computer can win without finding a move superior to a human's selection for every position.  It simply has to make less errors to win.  So let's say the practical definition is for any given position the computer finds an equal or superior move to what a human can find.

I can appreciate the practical approach, but this makes the title question trivial, of course this will happen... it's nearly happened already.  The more interesting question is if chess will ever be solved in a truer definition of the word.

Yeah, that was the point of my post -- you can do the math and get to something that seems reasonable based on some assumptions, but those assumptions are fundamentally flawed.