Every attempt must be made to play with and force emotion on your opponent. If not humans are little more then very sad little computers. IMHO The enormity of possible chess positions if solved would make trying to remember them a weakness. I would assume that is why chess has not been "solved". the slightest deviations of position would create multitudes of now "solved" results. Deep breath
Can Chess be Solved by computers?
Sort:
gisleyt wrote: Nytik wrote: The solving of chess is done retroactively, that is, we are solving from the back, not from the front! The first moves will be the very last to be solved.
This is very interesting, do you have any pages I might read for current initiatives in this area? I didnt know there were any serious attempts on this right now, since it clearly lies beyond the scope of what current computers are capable off.
Endgame tablebases. All positions with 6 and fewer pieces (including the kings), and some 7 piece positions are completely solved.
However, they take a huge amount of disk space and time to compute, making it pretty unlikely that we'll see all 8-piece positions any time soon (I think every step is about a factor 100). So let alone all 32-pieces positions...
TheGrobe wrote:
Not classical computers -- if it's solved it will likely be with the use of a quantum computer. I doubt it will occur within my lifetime.
How long you think you are going to live? Under 200 years? ;)
I think that even those supercomputers are ill equipped for the complexity which is still multiple orders of magnitude outside of their capabilities. Even if they did reach a point where the capability was there you'd never see the processor time dedicated to such a trivial problem when other more practical problems such as climate modelling and simulating the big bang are competing for it.
- I agree with one of the above posters that if both sides play a "perfect" game then the result would be a draw. You may have "solved" chess, i.e. discovered how to play the game perfectly, but in an engine vs. engine match using the same hardware and software, no one would win.
- I think this "solving" of chess is many years off, and very unlikely in our lifetimes.
- Chess is vastly more complex than Checkers/Draughts (and Go, in turn, is vastly more complex than chess!)
- Even if computers learnt to play a perfect game, humans would still play their imperfect brand of chess (athletes didn't stop running after the automobile was invented...)
Hi guys,
Chess cannot be solved by todays computers. But, the future is ripe with promise.
Watch your backrank.
Why do the engines have to play different openings? That's not in the rules of chess.
Bear with the following hypothetical: In a world where chess is solved and it's determined the outcome is a draw and only one move in each position leads to the draw, computers programmed with the solution would always play the same drawn game. Insisting they play different openings every game would be insisting they play below perfection.
A perfect game isn't a relative term -- it is a game in which both sides made the best possible move at every turn. There is at least one best move for each position but becuase there may be more than one, you are correct that there may be more than one perfect game.
The only way to know what the best possible move in each position is, however, is to solve chess -- the solution of which would effectively be a 32 peice tablebase (i.e. a list of all legal positions, and the best move from each).
Schachgeek wrote:
"perfect game" is a relative term.
How would you define a perfect game, anyway?
In baseball, a pitcher can have a perfect game and still lose in the 10th inning.
Just because it's drawn does not mean it was a perfect game, and a series of games between two identical engines would not always end in a draw since one side plays White and one side plays Black.
Of course if there is some software flaw with the engines that causes them to play the same moves in each game leading up to some 3-fold repetition then that doesn't count. The engines have to play different openings in each game for this to be a valid test.
"Perfect Play" is a mathematical term which is a fancy way of saying that every move retains the ability to WIN the game. Both sides would play perfect moves, regardless of the game-tree complexity. In such a case, the outcome of every tree is a draw, which is what we want to know. At that point, Chess becomes "Solved", since you always know what move to make in order to "not lose".
Wikipedia has some good articles on game theory to help understand what is meant by "solving Chess":
http://en.wikipedia.org/wiki/Perfect_information
No -- in order to know that that one particular variation is perfect, you have to prove that all the other possible variations are worse. So you have to solve the rest of chess as well.
Schachgeek wrote:
In baseball, a pitcher can have a perfect game and still lose in the 10th inning.
How's that? A scoreless nine inning game in which one pitcher has pitched the entirety without giving up a safety is not a perfect game. They have to win it. They can be perfect for nine innings and lose it in the tenth, thus losing the perfect game.
Schachgeek wrote:
Computers programmed with the solution would always play the same drawn game, that's my point. You can't solve chess that way, you've only solved one particular variation.
Can't solve chess what way? My post made no reference to a methodology for finding the solution. I was merely pointing out that your requirement that the computers play a different opening every game is artificial.
Dec 1, 2009
0
#35
Keep tracking the Tablebase sizes each year ... As an earlier poster indicated, we're barely at 8 pieces in the year 2009 so I think we're a ways off.
Nobody is suggesting to solve chess by letting engines play each other; that's not how you solve a game.
Shivsky wrote:
Keep tracking the Tablebase sizes each year ... As an earlier poster indicated, we're barely at 8 pieces in the year 2009 so I think we're a ways off.
Each additinal piece adds an order of magnitude of complexity to the problem as well (granted, once you get above 10 you can start to cull illegal peice combinations like King and ten Queens etc) which is why the currently available collective computational capabilites are simply still not sufficient for this problem (and likely never will be using classical computing technology). Our progress will slow substantially -- it already has.
Schachgeek, this is how the solution will be arrived at -- by working backwards to create tablebases for an increasing number of pieces until we're at 32, not by blindly trying to find best play for each move starting from the setup position.
TheGrobe, as long as we're speculating on things that are impossible, I'll suggest the possibility of solving chess in a way other than generating the 32 piece table bases.
Suppose you could prove the size of the upper limit on the dynamic complexity of a chess position. By this I mean you could definitively say "For every chess position, there exists a change in static evaluation within X ply or the position is equal." The statement is true for some X (because chess is finite) and I personally believe (without proof) that it is true for some "small" X. By "small" I mean something like 100-200.
If you could prove the dynamic complexity of chess is 100 ply or less, you could do an exhaustive search forward 100 ply from the starting position to solve the game.
Given that some endgames that we didn't understand until table bases were produced require longer than the 50 move rule to win, perhaps the above idea is a bit naive. But it is an idea.
Dec 1, 2009
0
#39
tryst wrote:
First, we don't know if d4 is a forced win for either side, or a draw.
Well, I do. Feel free to ask me any time 
. Denying the most obvious fact that the perfect game would be a draw with the premise that chess isn't solved yet, is like denying that the world isn't ruled by pink jello with the premise that we haven't solved the universe yet.
i don't think so as even the most powerful computer in the world has a point in which it won't work even though it can do hundreds of sums at the same time. one day maybe they will finish working it all out but then again it could be 5009 until it works every thing out. any one got a time machine just to check when they work it all out?
If solving chess from the starting position is many years away, have computers solved the first 20 moves yet? If so studying Chess openings at the top level will become irrelevant.
The solving of chess is done retroactively, that is, we are solving from the back, not from the front! The first moves will be the very last to be solved.
This is very interesting, do you have any pages I might read for current initiatives in this area? I didnt know there were any serious attempts on this right now, since it clearly lies beyond the scope of what current computers are capable off.