Will computers ever solve chess?

Here's my take on the current status and key points of solving chess:

First, contrary to what is sometimes implied, it is not necessary to store the entire game tree, nor the total catalog of all chess positions to solve chess. If chess is a forced win for White, it is necessary to show how White will move for any move by Black. Vice-versa if chess is a forced win for Black. If both sides can force a draw, then it is necessary to show the move to be played by each side to avoid losing, for any move by the opponent.

Such a "perfect game" will reside within the small dark green area of this diagram (other areas can be ignored for the purpose of solving chess).

So some key parameters related to the problem of solving chess are:

1) Number of nodes in the game tree that need to be examined to solve chess: Unknown.

2) Number of mathematical operations required to solve chess: Unknown.

3) Computer performance achieved by humans: known - at least 500 Teraflops (example: SETI)

4) Size of computer algorithms manageable by humans: known - at least 100 million lines of code (example: 5ESS switching system).

So in my view, the big unknowns related to solving chess are (1) and (2) and not (3), and (4).

You can get reasonable estimates and bounds for  (1) and (2). There is no reason to believe big algorithms help much, until they contain a large part of the solution because generalisations are very rarely of much use in chess: absolute precision can only be achieved by examining every possibility. (It is very easy, conceptually, to write a program that can solve chess in an instant, by containing the solution hard-wired in every detail. It is also easy to write another small program to produce this stupendously huge program, but it takes a lot of trillions of years to run, and an inordinate amount of space to store the result.  )

Actually there are some important technical details here: communication being limited by the speed of light, if you need to get from a part of the program that is in one galaxy to one that is in another, it is going to take a while.

It's sufficient to be able to generate parts of the tree as needed, which indeed is all that's possible for any usefully-sized search space. 

If you can only handle perfect games, you haven't solved chess, though. 

It doesn't matter what the 'possible' moves are because there is one, maybe two, correct moves. The possible moves don't count because they are wrong moves and lead to a loss. 

How reliable is a 2000-rated player commenting on 4000-rated players?

I would agree with you about the likely truth, but it is appropriate not to be sure.

Computers can defeat humans in chess.

 As usual, it is feasible to ignore most moves for the side for which you are designing a strategy, but chess does not provide shortcuts that allow you to ignore legal moves by the opponent.

This is how checkers was solved.

If anyone knows of a good academic paper with these results and calculations please let me know the link.

In 1949 Shannon produced an estimate for the pink area in the diagram (10^120), but that has no known relationship with the green area.

Hmm.  I was going to say Schaeffer's checkers papers (I did actually say that, I'm editing this) but they don't really give this.  I guess other than...what makes you think you can find the small triangle without searching a much larger area to find it?

.. If, Anything ; The recent developments, in the 'Alpha-Zero', chess, 'Go' and 'Shogi' software ; Show, that the 'brute force', move algorithm ; Is as unnecessary ; As it is, a crude, methodology !

https://www.technologyreview.com/s/609736/alpha-zeros-alien-chess-shows-the-power-and-the-peculiarity-of-ai/

It shows brute force is unnecessary to play very strong chess.  It is extremely far from being a solution.

Exactly. 

AlphaZero's (dubious) match results with Stockfish have no significant bearing on this thread/solving chess.  Further developments in AI may have some bearing, but the AlphaZero event as it stands now is meaningless to this debate. 

I will also state well in advance that even if AlphaZero reaches a point where it wins every single game against every other engine and human player, that is not "solving chess" either.  

You are correct in every respect except for the bracketed "dubious". The win was entirely above board. It is possible that Stockfish could play a little stronger with more powerful hardware, but the published results suggest its standard of play would not reach that of AlphaZero.

No.

It is disingenuous to put out press/stories/Wikipedia entries laying out the results of a private match, one where Stockfish was set up and run by the AlphaZero team.  They had a vested interest in seeing Stockfish lose badly, and they made sure it happened...

This match was a private test at best, a shady publicity stunt at worst.

If they hold a real match with the Stockfish TCEC team involved, then I will give them some credit.

P.S.  Parentheses (sample text), not brackets [sample text].  

I don't know if it can or cannot be done. The question has been posed to chess theorists, programmers, and mathematicians, and nobody knows the answer - and so this thread never dies.😄

The problem with your premise is that, failing some other method put forth, traversing the entire area searching for the solution is the default and only accepted method for "solving" a game (see:  Wikipedia's entry on "solved games").

So, much like "innocent until proven guilty", chess cannot be solved given current hardware/storage/universe size constraints ...and that *is* the answer, until someone can prove otherwise.

You, sbog, et al can continue to speculate that chess can be solved in our lifetimes, but it holds no more weight than claiming you think we will discover FTL travel in our lifetimes...

Scientists are welcome to posit theories for FTL travel, just not to publish papers, etc. unless they can back it up.  Similarly, people can speculate about methods for solving chess, but those people are just living in a fantasy world unless they can back it up.

You can't solve chess but maybe someone else can.😛