Solving chess? With no BS. (moderated)

And I think they generally still suck at basic endgames because they don't get the practice. At least Leela does on a basic PC.

Maybe I should just make this the title of the thread .

Humans suck at chess?

Let's be clear about what knowledge and skill means. Quantum improved on Einstein who improved on Newton... but in the year 2021 the way we understand and calculate a falling apple is the same. There can be a mountain of things left to discover, but each new discovery does not invalidate the things we currently take as true... in fact it's usually the opposite. New discoveries confirm what we knew was right all along while also solving some tricky edge cases like how apples might fall when traveling near the speed of light.

So I'll argue that humans "suck at chess" in the same way humans "suck at physics" or "suck at math." We're not gods, but as far as simple things go, we actually understand a hell of a lot... as in basically everything.

---

I'm not saying the result of a perfectly played game is as established as the acceleration of gravity... but I am saying it's disingenuous to argue we know nothing just because the frontier of our knowledge is mobile.

I think that perfect chess ends in a draw is a safe conjecture.

I feel like that's giving him an awful lot of credit... I assume he just doesn't like you or something.

"He knows that..."

The man doesn't even know his own definition of the word proof and evidence. It changes every few posts.

And look, I understand, I'm old enough to see myself declining here and there. I have aged relatives... we're all headed the same way, prince and pauper. I don't like making fun of Ponz... but some of my discussions with him in the past have been extremely annoying.

The 50 moves rule and/or the 3 fold repetition rule are essential. Without at least one of them chess would not be a finite game and thus it would be impossible to determine if it were a draw, a win or a loss. It is only possible to solve chess because at least one off both rules. The 50 moves rule guarantees every chess game ends in at most 5,949 moves. The 3 fold repetition rule guarantees a finite ending as well, but in a much higher maximum number of moves.
As said ICCF and TCEC allow win claims based on the endgame table base without the 50 moves rule.

From a solver's perspective that's not true. A game with infinite moves will repeat a set of positions infinite times. The solver only needs to know the outcome of these positions.

Or more humorously, an infinite move game, by definition of the rules, has not ended in checkmate, and so is equivalent to a draw

Actually it's 5898.5 moves.

Although I have to say, arriving at the number is pretty tricky. Most people (myself included) aren't able to figure out how to get that extra 0.5 of a move. (I know now, after I saw someone else's solution.)

This.

I guess the fact that tablebases, moving backwards from mate, can not possibly run into a game of infinite repetitions has just escaped some people.

From the perspective of a 3600 player, yes, all humans suck at chess, to the same degree a 1200 players sucks when they play you.

I'm not throwing out the value of human play, I'm saying that it's too small to affect the discussion of solving chess.  

You made a comparison about calculating falling apples...but you're comparing playing chess to the effort to solve chess.  Playing chess games is like a rain of several billion falling apples when we need 10^40+ apples.  The *calculation* of the falling apples, that's tablebases, and yes, we have made it to 7 man tablebases and *that* work cannot be discounted, but it is also an infinitesimally small slice of the problem space.

Oh ok, so you're fine with me saying chess is probably a draw, but you're arguing that in terms of solving we're still very far away.

I'm fine with that.

You're preaching to the choir ...I'll leave it at that.

That might be true in game theory, but it's not true in chess played according to FIDE rules (or any other rules I've seen).

Art. 5.1.2
The game is won by the player whose opponent declares he resigns. This immediately ends the game.

This implies a player my resign and there are no other articles restricting when he may resign.

If two players resign simultaneously then both players win. If either player resigns simultaneously with a piece being released into a dead position then the game is drawn, but his opponent wins.

Interesting that according to 

Art. 1.4
The objective of each player is to place the opponent’s king ‘under attack’ in such a way that the opponent has no legal move.

and Art. 5.1.2 gives either player an infallible method of ensuring his opponent cannot achieve his objective.

Probably why @btickler suggests art 5.1.2 is ignored in the thread, together with

Art. 5.2.3
The game is drawn upon agreement between the two players during the game , provided both players have made at least one move. This immediately ends the game.

which also has no restriction on when the event may occur.

     The only way to solve chess to everyone's satisfaction is to analyze every possible opening move through every possible variation to reach a win or draw. The amount of computation is beyond the capabilities of any present-day method. The only hope is a revolutionary technological advance. This may happen 1000 years in the future, or someone may have already thought of it but isn't able to produce practical demonstration. Short of a breakthrough in computational abilities we will remain stuck with opinions.

Not really definition of the rules, more definition of ordinal infinity together with the rules defining only successor positions. The result of an infinite game is not a draw according to the FIDE laws, only in game theory.

You have no justification for saying that. It's possible that there is a forced mate from the starting position in say 30 moves that has not yet been seen, in which case full analysis of only the tree associated with that mate would be sufficient. Or if not a mate in 30 possibly a mate in 3000.

And in any case mates (or draws) can often be proved without fully analysing the associated tree. You can probably mate with rook and king against king from any winning position on any size of board. You don't need to analyse the infinite number of variations move by move.

There's brute force and there's ignorance, but there's nothing mandates you have to use both.

You need only play two mats, A and B of Nim where the players take from piles on the mats in the order A,A,B,B,A,A,... . Two players unaware of the theory would probably draw one mat each about half the time.

But you need look no further than chess. If you restrict the pieces to KQ for one side and KNN for the other in the starting position, then from a moderately deep winning position (which most are) the result is usually a draw under the 50 move rule. (Moderately deep would possibly need to be around 45 moves or more for SF v SF.)

Opinion by the late GM Sveshnikov:

"Chess is an exact mathematical problem. The solution comes from two sides: the opening and the endgame. The middlegame does not exist. The middlegame is a well-studied opening. An opening should result in an endgame.... Give me five years, good assistants and modern computers, and I will trace all variations from the opening towards tablebases and 'close' chess. I feel that power."

So he suffers from the same malady as you ...not understanding at a fundamental level what large numbers are and how fast chess exponentially increases in viable move variations.

Having never even attempted to begin his asserted premise, it seems he was a tad early to make the claim.

One thing is interesting though...like all GMs I have ever seen pontificating on this topic, he does not guarantee a forced draw result.  

Two updates of note:

- Tygxc has finally given up, apparently. No sign of him in quite a while on the forums.

- Quantum computers may be taking a leap:

https://pubchem.ncbi.nlm.nih.gov/compound/Uranium-telluride-_UTe2

...but this still won't solve chess. I did want to mention it though because it *starts* to address the fragile nature of quantum computers and the "destructive read" that has been discussed in this topic. So, more qubits are going to become possible.