Will computers ever solve chess?

It doesn't matter, because computer chess programs are no longer an opponent, but only a reference and instructional tool. No one can memorize all the lines involved, but CAN build knowledge from them. Then we can play our favorite opponents--other human beings. 

And oh, by the way, I think your numbers are basically correct for "strongly solved" and his are basically correct for "weakly solved" and either way neither one is going to happen, but there is no point getting into name calling or dismissing others -- no one here has demonstrated a really deep understanding of this problem -- the sort that would qualify any of us to author a paper on it, so there isn't any place for assumed superiority except, maybe, if you do have people talking about "an increase of a million times" or "99% of positions," then yes they are missing the point, but even then you can be nice about it.

Sorry, meant to edit the previous one.

Yes, I've used Google. I've also read and understood academic papers on the topic of solving chess.

There is no peer-reviewed academic paper that says chess cannot be solved.

For this reason, people should be free to express their views without others treating them in a condescending way, or people saying things that appear to be personal attacks. This thread is more interesting and informative when people are free to express their interesting, innovative, inspirational, and creative ideas.

Btw, if there is a peer-reviewed academic paper that does say chess cannot be solved, please feel free to include a link. I would like to read it.

And thanks to all the people who have visited this thread, many very recently, who have been polite and help make this topic interesting.

Your chess 101 and my chess 101 are different.

1. Nf3 does temporairly block an early f4 but in the great majority of 1. d4 openings an early f4 is not best or desired. After opening 1. d4 the best follow up is usually or often 2. c4  and pushing both the c pawn and f pawn forward 2 squares in the early opening is usually not best play.

1. d4 commits to playing that pawn forward 2 squares early in the opening and 1. Nf3 does not.

Similarily 1. c4 has some merits that 1. d4 does not.

And 1. d4 has some merits that 1. e4 does not.

and 1. e4 has some merits that 1. d4 does not.

So all 4 moves 1.d4   1. e4  1. c4  1. Nf3 have merit and differing merits so one cannot say all 4 moves are exactly the same or that 2 of them are exactly the same or have the same merit.

They would all have to be investigated to see if any one of them [or two of them or three of them or all four of them] lead to a forced win or a forced draw. [or loss but that is very unlikely]

There is also 1. g3 to be considered.

If you've been reading this for years, then you already know how patient I've been .  This used to be not even the largest thread on this topic, and there's nothing new under the sun here.

Anyone that wants to discuss this topic is perfectly free to do so.  I'm not attacking every new poster...far from it.  I am disputing with people that have repeatedly and knowingly kept spouting BS after it's already been shown to be BS, and not just by me, either.

If you want to say "I think it's solved in 50 years", then show your work.  I've shown mine.  I used the speed of light as an example before.  If this thread were about creating warp drive, people could go on and on about theories all they want to...what they cannot do is dismiss people bringing up the speed of light by dismissing Einstein (unless you show your work...and good on ya' if you prove special relativity wrong), or by simply saying "yeah, whatever, I still think warp drive is coming in 50 years and I have nothing to back it up".

In the earliest threads I did some numbers on carbon (diamond) storage arrays (which are a potentially viable technology with a foreseeable future), then atoms/universe calculations, etc. and I eventually proved to myself that it won't happen.  Anyone playing with realistic numbers is not going to hear anything from me.  It's the people that drop a tree graph that's entirely inapplicable, or the people saying "we just need a new paradigm...I don't have the foggiest what it should be, but I am confident in my 50 year number anyway", or worst of all, Patroit-types with their "nobody can imagine what will happen in the future...except ME!  50 years." that are the ones I take issue with.  

As for the dismissiveness, it is coming from both sides.  I'll agree that I can be kinder, and I am to pretty much everyone else with the exception of trolls/sockpuppets.

I would like to thank all the posters that come in to ground people back in reality when the fairy dust and maypoles come out in this thread, which happens over and over.  There's nothing interesting in this thread that wasn't posted long ago.  We might as well be discussing reality TV.

Actually, I think we've covered a lot lately, which I'd summarize as follows:

1)  The size of the problem is 10^123 games, but only 10^46 positions, roughly.  It is the number of positions which matters, not the number of games.  Likely this will be forgotten again, but it's a good starting point.

2)  There are several commonly used definitions for solved, but two main ones.  "Weakly solved" suggests that we know the opening position is a win for White, a win for Black, or a draw, and we can show the path to that result.  "Strongly solved" suggests that we can say the result from any legally reachable position, and show the path.  This is important since using different definitions for "solved" will result in people talking past one another.

3)  Weakly solved is much easier than strongly solved.  It requires a tree of only about 10^23 positions This is because the side trying to resist the forced result (whatever it turns out to be) must explore every possible move, but the side trying to confirm the forced result only needs one move that gets to that result, so it isn't necessary to look at other moves once one is found that works.

4)  Strongly solved is much harder than weakly solved.  It requires a tree of almost the full 10^46 possibilities.  A few possible positions (actually very many, but few relative to the problem size) may be effectively identical, and identifiable as such.  For example, a machine in the early 1900s could do K&R vs K, with the algorithm encoded in gears, entirely mechanically, and there would be at least some number of other similar positions.  Or, even more trivially, there is no need to separately record all instances of K vs K.  I mention these few possible positions not because they change the result, but only because otherwise someone else will bring them up as if they have not been considered.  They are too few to change the result.

5)  Neither strong no weak solutions are in reach of current computers, or reasonable extrapolations of current computers.  In particular, there are limits to how far Moore's Law can go, in terms of the speed of light across a chip and also in terms of the size of structures on a chip which must be made of atoms.  In another particular, this may or may not include quantum computers, even if successfully implemented on a practical scale, since it is not clear whether the advantages of quantum computers are especially useful for this task.

6)  But anyone is free to essentially, "expect the unexpected," on grounds there have been unforeseen developments in the past and something will come along that can do this, even if we don't know what it is, although we should be careful not to have faith in things which are demonstrably not possible, as opposed to merely unknown at this time.

So, hopefully that is a good summary.

What kind of systems have you worked on post graduation?  Anything with high transaction volume, server farms, etc.?  I'm asking because I see a lot of textbook theory in this thread, and not much of the practical or anything actually implementation-minded, and I'm not sure why nobody else with large scale systems analysis/design background is chipping in on this thread; surely there are many such people that play Chess.

Yes, it's all theoretical, the more so since I think we both agree no real system can do it.  You said it's a systems problem, but I think it's more one of theoretical math.  But anyway, look at my summary.  Hopefully you will be satisfied with it.

(1) was already discussed long ago, was forgotten, and has been "re-discovered".  That is not progress on this thread.

(5) is the only point that really matters.  Also, Moore's (diminishing returns) Law has also already been debunked previously as a factor in predicting future processing capabilities re: solving chess.  Not progress.

The summary is fine, in terms of gathering up what has gone on recently...but this info is not really new, or refined, or changed much.  If there were new nuggets to be gleaned among the fanciful musings, I would not really be harping on this thread.

My point I guess would be...yes, impossible at the theoretical level, but double-dog-diddly impossible squared at the implementation level.  So impossible that the 50 year estimates are just laughably absurd.

I think it's interesting that some people on this thread seem to have concluded that chess can never be solved. Showing the entire game-tree of course is impossible (a strong solution) - that was demonstrated in 1949 by Shannon, so that's nothing new.

But there is no peer-reviewed scholarly article that says chess cannot be solved in terms of who wins from perfect play. Chess is a widely played game with very strong interest by mathematicians, game theorists, and program developers. One would think that if solving chess was impossible, some credible academic paper would have been published saying so. It would surely be a landmark achievement.

I guess everyone is afraid to be the next Kershner. He believed that no new shape could be discovered that tiles a plane, but was proven wrong a few decades later.

It is interesting to ponder what "better" means for moves that in all likelihood have the same evaluation. The answer is that a better position is one from which you can achieve a 50% result using less computing/thinking time, with the players being of equal strength.

For example, the opening position scores about 54%. This can be turned into 50% by giving black a time handicap (either for a computer or a human).

 All opinions are welcome here as long as they are not abusive. Disagreements  are ok as long as they are not abusive. 

However if you have  disagreement with a posting or some postings--it is best to try and give a reason why you disagree.

A solution of chess is of far less theoretical interest than Kershner's contribution to the topic of pentagonal tiling of the plane, even if he guessed incorrectly about the lack of any other solutions!  A fascinating fact is that although the first pentagonal tilings were discovered in 1918 and Kershner added more in 1968, it was not until 2015 that the last one was discovered. And now it is proven there are no more! The latter point is the nearest to the topic of this forum. See Pentagonal Tiling of the Plane

Thanks for that information.

The paper by Michael Rao, just written May of this year, appears to be well-written, has good diagrams, is based on sound theory, and has good references. When I see a paper of that quality that says "Chess cannot be solved" then I will believe it.

Btw, I don't think solving chess is less interesting than tiling a plane, but that's just my opinion.

Kudos to Mann, McLoud, and Von Derau who discovered a new shape when others said it couldn't be done.

It's very easy in the sense that flapping your arms and jumping over Everest is easy. All you need to do is flap your arms fast enough.

Your argument is like saying since it is possible to jump a foot in the air, it is possible to jump over the Moon.

These are two separate things. As we should all know, the value of a position can only be +1, 0 or -1 with perfect play. "Advantage" is more commonly used to denote practical advantage for real imperfect players (human or computer), giving greater chance of a win.

bb bum are you saying chess can be solved because we now have endgame tablebases?