Chess will never be solved, here's why

10,000 is a tiny number. Not even enough for 14 binary choices! A drop in the ocean.

@6010
"10,000 is a tiny number"
++ 10,000 games is almost a million positions.

I SERIOUSLY doubt they are "forgetting" that. 

Strange. 

User @tygxc says here

I agree with

"Provide an algorithm that secures a win for one player, or a draw for either, against any possible moves by the opponent, from the beginning of the game. That is, produce at least one complete ideal game (all moves start to end) with proof that each move is optimal for the player making it."

In case of chess that would be: 

"Provide an algorithm that secures a draw for black, against any possible moves by the opponent, from the beginning of the game." (my highlight.)

Are there two users with id tygxc?

Incidentally when are you going to post your calculations to determine the theoretical results of these positions and the number of errors in the games. If you do that we can all forget about your proposals and talk about sensible things.

You needlessly insisted that I produce a drawn KRPP vs KRP set of games for you while declining on spurious grounds to produce any yourself. (It's to assess the calculation that you have proposed).

I spent a week doing that. The least you can now do is fulfil your promise and post your calculations.

Yes, a TINY number. There are about 10^38 times as many basic chess positions. To put it another way, the number of basic chess positions is much more than the 7th power of this number.

[No need to repeat the mantra about weak humans deciding most of the positions can be ignored based on vague and unreliable heuristics - we've heard it almost as many times as the size of your dataset].

@6014
"Yes, a TINY number. There are about 10^38 times as many basic chess positions."
++ There are 10^44 legal positions and of these 10^17 are relevant.

@6007

"The famous and challenging solution of checkers was a weak solution that required years of computation." ++ Checkers also used only 19 relevant openings of the 300 imposed openings.

"there's a .0000000000000000000001 chance we're wrong"
++ As there are 10^17 relevant positions, 1 error in 10^20 positions is OK. 

"why it took years to solve checkers" ++ He had to write his own engine Chinook, he has to generate his own endgame table base, his computers were less powerful. It also takes 5 years to weakly solve chess, with Stockfish available, with a 7-men endgame table base available and with engines available that calculate a billion positions per second.

"I advocate strict rigour - the same as the peer-reviewed literature on solving other games."
++ The peer-reviewed literature shows Connect Four solved with knowledge only.
So weakly solving Chess can use knowledge too.
The peer-reviewed literature shows Checkers solved with only 19 relevant openings of the 300 imposed openings. So weakly solving Chess can discard irrelevant lines too.
There is no reason why solving Chess should be subject to more stringent requirements than Connect Four, Checkers, Losing Chess, or Nine Men's Morris.

Apart from the fact that you can't explain why 10^17 and what your positions are relevant to, the figure of 10^20 depends on your "calculation" of error rates. You can check that by applying your "calculation" to my games here where we can check your results with Syzygy.

When are you going to do that as you promised?

Are you then asserting that Checkers is not in fact solved?

Which peer reviewed literature would that be?

I think @tygxc does not understand that it was proved rigorously that the said set of rules worked. The rules define a strategy, the verification of the value of that strategy involves the same exhaustive analysis of all opponent responses.

@Elroch

You must have read the same one I read. @tygxc apparently knows of a paper solving it without any exhaustive analysis. 

@6017
"why 10^17" ++ Calculated in 2 ways:

1. top down from 10^44 legal, then 10^37 (Gourion), then 10^32 sensible, then 10^17 relevant, exponent in analogy to solving Checkers and Losing Chess.
2. bottom up from width 4 and depth 37: upper bound without transpositions,
   lower bound with full transpositions, the geometric mean as estimate.

"what your positions are relevant to" ++ To weakly solve Chess.

"the figure of 10^20 depends on your calculation of error rates."
++ Yes, I have calculated the 1 error in 10^20 positions from the AlphaZero autoplay paper.

"my games" ++ Your games are not relevant. I wait for a drawn KRPP vs. KRP you promised.

@6018
"Are you then asserting that Checkers is not in fact solved?"
++ No, I accept Checkers as weakly solved: the 19 relevant openings are enough
and the 300 imposed competition openings are not necessary.

Likewise weakly solving Chess does not need all possible openings either, only the relevant.

@6019

"Which peer reviewed literature would that be?"

++ The MSc thesis of Allis
http://www.informatik.uni-trier.de/~fernau/DSL0607/Masterthesis-Viergewinnt.pdf 

Well that's the same one I read.

In the sense that we have knowledge that an exhaustive search is a logically valid way of arriving at a solution I could possibly agree that 

The peer-reviewed literature shows Connect Four solved with knowledge only.

with the important proviso that the nature of the knowledge involved is deductive as distinct from that which you apparently want to include in your own proposals.

8.0658 x 10^67 = 52! (number of arrangements in a deck of cards)

10^120 = Shannon's number (number of games of chess)

@6035

"To weakly solve chess how?" ++ I have explained ths several times: calculate from the opening towards a 7-men endgame table base draw or a prior 3-fold repetition. A good way is to utilise ICCF WC drawn games and to investigate 3 alternatives for each white move. 

"Are basic rules positions even relevant to your proposal?" ++ An x-fold repetition rule is vital. The solution without the 50-moves rule also applies with the 50-moves rule. If black can draw without the 50-moves rule, then black can draw just the same with the 50-moves rule.

"it's been asked for many times." ++ I have explained many times.

"check whether your calculation works"
++ I have checked, also to 2 of your irrelevant games above. It does work as shown above. You did something wrong with your version of Stockfish.

"1. Your calculation purports to determine if the position is won or drawn"
++ No, the calculation purports to determine how black can draw
i.e. achieve the game-theoretic value against any opposition from white.

"Why ask for a drawn position?"
++ Because weakly solving Chess is hopping from one drawn position to the other.

"Why ask for a KRPP vs. KRP  position?" ++ Because it counts 7 men, inside the 7-men endgame table base and just outside the domain of 32-8 men for weakly solving, because it is the most prevalent type of endgame, because it can be drawn despite a pawn down.

"ONE OF THEM WAS A DRAWN KRPP vs. KRP POSITION" ++ No, all 4 of your positions are wins.

I think you may have missed my response to the first post where you quoted that second figure as the number of games of chess.