Solving chess? With no BS. (moderated)

I wrote up a short scenario for this in the "Will computers solve chess?" thread, let me see if I can find it.  It's based on the best supercomputer from the time, maybe 2015-ish? 

Edit:  Here it is...

100 PetaFLOPS is 10^17 floating point operations/sec. Evaluating a chess position is not 1 operation, mind you, nor is it 10, so let's be kind and say it falls in the 100s order of magnitude, which knocks 10^17 back down to 10^15 positions/second, which is 8.64^19 positions/day, 3.15^22 positions/year.

At that processing rate (assuming infinite memory/storage and ignoring all the issues thereof already laid out) you would solve chess in...3.175^24 years. I guess you could amortize a loan for the duration on the $273 million for the supercomputer...

Adding in storage, you solve chess...never (in this scenario).

Which I expanded upon later:

The fastest supercomputer could solve checkers in a matter of seconds (10^17 FLOPS vs. calculating 10^14 positions), but it will take 3.175^24 years to solve chess. If you spend the entire wealth of the planet (approx. $80 Trillion in currency) to build an array of these supercomputers, approx. 290,000 of them, and use them all for solving chess leaving the human race to starve and die (so we'll also leave as aside the question of who would run the giant computer array), it still would take 3.8 million years to get your answer.

I make it 20 and at least 19 to each of those 20.

By "the starting point" most people mean the one at the beginning. 

You're right, that's why tablebases are built going backwards from mate.

Chess is solvable, but it's a near certainty it would take a quantum computer (or some even more advanced computer not yet invented) and holographic memory to brute force every possible position allowed in chess, including variations in which en passant and castling is or is not allowed.

I can give you a host of "starting points" that satisfy your goal, but if you're to solve chess it has to be the one I posted (but it doesn't necessarily have to end in mate).

Noughts and crosses (tick tock toe in American) is considered solved. There is no forced win.

Solved in the context of games just means you can predict the outcome given perfect play, win or draw.

Theoretically possible. Practically impossible.

Philosophically pointless w/r/t chess itself as it is an engineering problem that will have zero effect on how chess is played because to a human it would be an infinitesimal improvement on even today's engines. 

Understood and thank you, OP. 

How do you know there is no forcing move possible from the setup position?

There is definitely a single answer to the question, "Is there (a) a forced win for White, (b) a forced win for Black or (c) a forced win for neither side?". It is very unlikely indeed that unique moves are required from both sides for the remainder of any game to maintain the status.

You don't actually know that, though.  When you are playing a chess game against an engine and walk into a tablebase's "mate in 153", you don't know yet that there's a forcing sequence of 153 moves that will result in you losing, but it's there.  The same is true when extending the tablebase all the way back to the starting position, it just hasn't been done yet.

"All first moves require choice" is ultimately irrelevant.  When you are in a losing position, there's a forced sequence of moves that will produce the fastest win for your opponent even with perfect play form that point forward on your part, whether either of you realize it or not...and you do have a choice, to make a move that loses faster (and your opponent also has the opportunity to misplay and release you from the 153 move mate) .  The only question is whether the opening position is losing for either side, or drawn.

Most chess players lean towards drawn, mainly because it is possible for one player to be up by 2 minor pieces in material and still not have the ability to mate.

A draw can be perfect play from a position. So can a loss. Neither is relevant to the issue of whether chess can be mathematically solved. Solving it for all possible first moves is a requirement for a solution. 

It could be that chess is always a draw for every possible first move. In that case, the lack of an optimal first move does not mean that chess wouldn't be solved.

If a handheld computer had been made prior to the big bang that was fed the initial state of the universe, and it had been programmed with perfect knowledge of physics, particle dynamics, etc. and it had infinite memory and instant computations, it would be able to tell you everything you are ever going to do in your life, and could track an eternity of movement for every particle in the universe.  Because there's isn't any choice, only a lack of knowledge that approximates choice and random occurrences.  The same is true in the logical sandbox of chess, just on a much smaller scale.

Perfect in this context means making optimal decisions. There is no reason optimal decisions can not lead to a draw. There is also no reason that there can be multiple decisions that are perfect. 

I

"Forced" means that only one perfect move exists. This does not have to be the case. The choice of first move for white could very well be meaningless in that they all lead to draws with best play. 

This is my own point of view.

With no supporting evidence at all, I believe (call it "faith", if you like) that with perfect play chess is a draw, and that the path to that draw is fairly broad... in the sense that the draw can be reached by a multitude of different lines, not just a single unique continuation.

@ExploringWA

What do you mean by a forcing move? The following position is won for White.

If Black plays 1...Kh1 then in order to keep the win White is forced to play one of the moves 2.Ra1(a3-a8,b2-f2) or 2.Kd1(d2-f2). Would you then say Black's 1...Kh1 is a forcing move?

A draw under the 50 move rule would certainly be considered perfect play from any position where neither side had a forced win and that was the case for all positions up to the draw claim.

That would include play from the starting position if the starting position is a position where neither side has a forced win (but I don't know whether it is).

I'm not as technically aware as many of the posters in this thread, but I think a few things should be considered:

1) Since chess is already partially solved, with 8-piece tablebase(s) likely to arrive in the near future ...

2) ... One can, for practical reasons, discount "bad" or inferior openings, if we're only in the hunt for solving chess at the "best play" level. There are some openings that regularly lose at the engine level, when our current top engines play against each other. We can reasonably assume that, if that trend continues, that it "proves" that those openings are lost with best play.

1. e4 d5, for example, seems to, time and again, lose for Black, at the engine level. Unless this somehow changes in the future, it's probably safe to say that any move possibility after 1. e4 d5 is likely a forced loss for Black.

So, for practical reasons, we could scratch 1...d5 off the list of positions needed to be "solved", along with every possible move combination after it, to save whatever computing power is needed.

This kind of selective pruning could be done more and more, as engines begin demonstrating, more and more, which openings predictably do worse at the engine level.

And the games don't need to reach completion, either ... they just need to reach a tablebase ending. Soon it will be 8-pieces. Perhaps 9 pieces won't be far behind.

Less and less of the game needs to be solved, if we're pruning from both ends. Eventually we'll just be trying to solve middlegame positions from only the most promising openings.

On technological advances:

I don't know much about quantum computing, but Neven's Law predicts that quantum supremacy is fast approaching.

And it's not unreasonable to expect that technology would, eventually, advance even past quantum computing, to something that makes even quantum computing seem primitive.

With such (admittedly hypothetical) advances, it seems entirely within the realm of possibility that solving chess isn't as out of reach as we currently think. It just likely won't happen anytime in our lives.

Kf2 is mate in 3.

The problem with pruning is that it assumes without proof. To prove something loses it has to be completely analyzed. There could be some super deep variation on the bongcloud that is a win for white that engines just aren't powerful enough to see. 

Computing power is only one practical problem. Storage is also an issue. Calling them problems is minimizing it. We'd have to be profoundly wrong about our knowledge of the universe to get past these problems.  

I suspect the interest around this is the idea that a full solution makes chess pointless. All it would do is create a reference slightly more accurate than today's engines. 

If you mean "from the initial position" then this is a completely unsupported statement. 

Also (IMO) a false one.