Will computers ever solve chess?

So with all this being said its safe to say that were not ever likely to solve chess even a thousand years from now and even taking into account how vastly improved technology will be then. So next: is chess a draw with both sides playing ABSOLUTE? Logically the facts point to yes. So bottom line we will never reach solvability with computers and if we were able to hypothetically it would result in every game being a draw.

Interesting...very interesting. How can a game that is a draw by its own essence be so compelling to us. What does that say about our intellect?

Ok, that's not unimaginable, but it doesn't seem practically possible.  Sure for extremely simple positions like forced mates you can prove a win without exploring every position.  But finding some strategy or algorithm to be used on complex positions seems as unlikely as other big breakthroughs... even more unlikely than say quantum computing when no one is interested or working on such a thing.

Although I do think it's much more interesting than the brute force method that quickly became the focus.

look at all the articles, we're not even close yet

Here is an example of a very simple infinite game that can be completely solved:

The game board is an infinite line of discrete "places" (think a number line of integers) - say laid out left to right.  Two players start at random places on this line (any random distribution over the integers may be used).  One player is randomly chosen (50/50) to move first, and the available moves in this game are: move one place left or move one place right.  If a player moves onto the place occupied by the other player, they "take" that player and win.

This game has an infinite number of potential positions.  Not just some really big number.  This game may never end.  There is no theoretical limit to the number of moves that may be played.  No matter how many subatomic particles exist in the universe, you can never store all of the possible positions.

And yet it has a simple solution.  If you start the game next to the other player, then with best play the first player will win (on the first move, by taking their opponent).  Any other position will lead, with best play, to a game that never ends.  This is easy to prove.

Intellect or goals / interests / priorities?  Do we play chess to discover something about the initial position, or (without getting too cheesy here) something about ourselves?

When in 100 years you and everyone you know will be dead, yet you have hopes, dreams, and e.g. follow social norms, what does that say about you as an intelligent being?

It's the journey not the destination.

Yeah that one was easy.

Ok, now syllogistically solve tic tac toe 

It may be that we've focused on the brute force method for chess because it's actually easier as well as faster (?)

There is a well-known strategy for Tic-Tac-Toe.

The only way to force a win in Tic-Tac-Toe is to make a move that reveals 2 potential wins on the next move.  If you only play moves that have a single square threatening to win, then your player can always play that square and block you. 

It turns out that knowing this, you only need to look one move ahead to play perfectly (and perfect play both sides is a draw).  There are only a handful of piece patterns that threaten wins in two squares, and the number is even smaller if you exclude positions where you should have just played the win instead of adding a new direction.  In fact, the number of patterns when you've removed symmetries is just 5. 

Perfect play then is to win if you can, else block your opponent if they are threatening a win, else block opponent from obtaining one of those patterns next move, else try to obtain one of those patterns yourself.  That can be proven optimal and doesn't require brute force (since you only look one move ahead).

Note, this is generally how chess engines work already.  They evaluate positions by some invariant rules to give a value to a position, and they only look some fixed number ahead.  Here, though, the invariants are positional heuristics on the piece patterns, and the fixed lookahead is some adjustable number which we use to give some notion of strength. 

It is known (since chess is a finite game) that there exist invariants that determine perfect play on every position with a fixed look ahead.  For very small lookahead, the complexity of calculating those invariants may be astronomical (the numbers discussed earlier in thread may apply).  Yet, as the look ahead is allowed to grow, it is known that theoretical complexity of the invariants will decrease.  And we know that this is not a zero-sum exchange because there do exist real tactics and strategy that work and can be proven in positions.  So the proposition that chess can be solved is really that we can find positional invariants for some N lookahead that we can prove perfect. 

I do suspect automated provers will be able to reach that ability.  My reason for that suspicion is that GMs regularly play the same moves that engines calculate as best with their 30+ move searches today.  That doesn't mean there are GMs that can calculate 30 moves ahead.  It means that there are patterns there that the GMs have learned that may serve as invariants in the engines of tomorrow, and the engines looking ahead 30+ moves with the new invariants will be even better.  We already can prove many mates and draws, but this will improve with new invariants.  Eventually, I suspect some of those invariants will be proven perfect for more complicated piece patterns, and slowly the game of chess will begin to be solved.

That's an interesting argument, I haven't heard it said like that before.  I certainly think it's possible, but I'm still not sure how likely it is.  Other than the endgame tablebases I don't know that computer chess developers are interested in anything related to these invariants... but I don't know for sure.  I've heard that one method they use is making a small change, then having the program play many thousands of games and comparing the results.  As long as there's a net positive change in the results, they keep the new version.  The fact that some positions are now played worse is acceptable.

The other difficulty I see is in proving these positional invariants.  Grandmasters must, as you said, have very good strategies and patterns memorized, but this is also their weakness.  Brute force searches find exceptions and the humans lose, so I'm hesitant to call these invariants.

ICCF competitors already do their best to combine the look ahead of the computer and the invariants known by the human.  They've claimed that some (major) openings are already known to be draws with best play.  I certainly can't prove them wrong, but I doubt they can prove these claims rigorously.  Even computers ignore best moves due to pruning.  That's probably my main disagreement with a strong solution as you describe.  How can you prove the moves maintain the evaluation (win or draw).

Anyway, I do think we can determine (even today) with a very high probability, and for most positions, what the best move and/or what the correct evaluation is.  But I'm not nearly as optimistic about achieving a strong solution.

I wanted to add one more description to illustrate how chess could be solved, because I think it shows really well why the talk about the complexity of the game is completely misguided.  This is a "what if" game, but that's what we're playing here.

What if we find out that there is a way to force the early exchange of queens with white in one of the 1.e4, ... 2. Qf3 openings (like the Napolean).  These are oddballs, certainly, but lets say there is a strategy that can force an equal queen exchange pretty much whatever black plays.  Black doesn't have to play best play, and neither does white (white can skip forced mates, for instance).  But even if black plays best, the queens will come off equally.  This isn't an outrageous scenario, clearly some openings like the Berlin have common queen exchanges.

And then say that the pawn structure after the exchange can be in one of 23 different patterns, and it is found how to force off at least 4 pawns from both sides in the course of the opening using a certain pawn advancement strategy.  The strategy ensures that this is at least an equal exchange for white as well.

There are already some well-known ways to force piece exchanges (pins, forks, skewers, etc.).  Let's say this is plugged in and we can always ensure at least 3 of the knights and bishops be taken off and at least one rook.

The hypothetical is that this can be pretty much forced equal, but if black doesn't want to take any pieces or put itself into forced mate, that is fine.  It doesn't have to be equal.  Even forced stalemates are fine here.  But with the strategy, there is no way that black can checkmate white while these exchanges are happening.

Where this is leading to is what happens if we can prove a trade down to a very simplified scenario like maybe a rook, a piece, and 4 pawns, and we know the 4 pawns can be made to be on some collection of possible files and suddenly the position is calculable with a good computer.  And now we show that all possible games are at worst a draw for white with best play.

And let's say sometime shortly after this last part is proved, someone shows that black can force the trade down of pieces to with his own strategy and can also force a draw at worst.

The point here is that all of this has now proven that chess is a forced draw.  And at no point were all moves ever possible needed for the calculation.  What was needed was only a particular move strategy that one could prove traded down pieces at least equally to the point where a brute force could prove forced draw.  The move strategy determines one line for white that depends only on the black moves and doesn't need to figure out even every possible move black could play - it could be purely something like "if black moves into one of these pawn structures, counter with this pawn structure" and "in pawn structures X, Y, and Z, if one of blacks pieces protect f4, play the same type of piece to f4".  General rules that someone proves are sufficient to force some trades or draw in 50 moves.

This kind of proof ignores nearly all possible lines of chess.  What if white doesn't want to play 2. Qf3?  Who cares!  What if there are some forced wins in some lines for white?  Who cares!  What is shown is that white can force a draw (or an incidental win) if they make these specific kinds of moves and we'll ignore all the alternates.  And so can black, with these other specific moves, even if white isn't playing this strategy themselves.  So chess is a forced draw with what could now legitimately be called "best" play.

I think this is unsatisfactory for some because playing for a draw (at least as white) is usually seen as bad sportsmanship.  Also, using a strategy and proving only whether that can force equality instead of looking at all the sidelines that might force a win seems to ignore the whole point of the game.

But if that kind of proof existed, for both sides, then chess would be solved.  One could always force at worst a draw in every game, no matter which side you played.

I hope this hypothetical shows how the full complexity really is irrelevant to the solution problem.  Proofs on strategies and the piece patterns they create do not need to calculate every possibility.  Strategies might be shown to simplify the problem space to a handful of possible outcomes which then could be easily calculated or possibly proven independently.  A universe of complexity may be ignored and only a manageable grain of sand remain.

Wafflemaster...I agree with part of what you are saying, but why would it be NECESSARY to store everything. My understanding is that checkers also has a ridiculous number of possibilities as well, but the solution still exists. I am basing this on the following article, regarding the checkers solution.

http://sciencenetlinks.com/science-news/science-updates/checkers-solved/

The main point of their technique: (quoted from above article)

"Getting to this point required not one, but dozens of computers, working around the clock for eighteen years, running through all sorts of hypothetical moves and counter-moves. Even so, Schaeffer says that he wouldn't have lived to see the solution if he hadn't figured out a shortcut. As you heard, his team programmed the computers to stop every time it found a winning move from a hypothetical board position.

Here's an example: Suppose that starting from a certain board position, the computer's opponent makes a dumb move. And suppose that the computer has ten possible responses—let's call them options 1 through 10—based on the board spaces that are open at the time. And suppose, further, that it figures out that it can eventually win the game if it chooses option 1. If so, the computer won't even bother analyzing the other nine choices; it just decides that if it ever finds itself in this board position, it will choose option 1. As a result, Chinook will never have to consider any board position that results only from choosing options 2 through 10, since that board position will never occur as long as Chinook is playing."

Let me ask you this, even if someone announced tomorrow that chess was finally and completely solved....Would you seriously give chess up ?...If you answer yes, your too easily impressed...and if you answer no, ask yourself this: Why did you bother to leave a post?....and why are you tracking this?...and just what is the point of solving a game anyway? It is like pulling the wings of a dying butterfly you find...You do it because you can to make yourself feel omnipotent while everyone else thinks your just one sad tosser with way to much time on your hands.

I would still play chess and why because I am interested in it in a scientific matter: How much is computer able to do.

you don't have all the answers as it sounds like you think you have.

I calculated once we will solve chess exactly 13 minutes before our sun goes nova. So enjoy te solution while it lasts.

Robots will overtake humanity.

This article is incorrect. The author has been misinformed. The sun is not going to explode. Our sun is not big enough to super nova. It will expand into a red giant and then will collapse into white dwarf. It will then fade away quietly into a black dwarf over an incredible amount of time.

I once was a little kid too, no wories my friend it will be ok....chess solution or no chess solution

Son, I didn't say supernova, I said nova.

Kid, I feel you, my bad, I thougth you was propagating nonesens out here in these streets. Play on Playa

You have an assumption that all the non-queen exchange paths are explored. In order for that to happen, chess would have to be completely explored. You cannot just assume best play up to the queen exchanges because best play is precisely whhat we mean when we say we have chess solved. How do you know that a move leading away from a queen exhange is not "best play"?