Chess Will Never Be Solved. Why?

#392

Not disputing that. Each cloud engine will start with a reduced set of reachable positions.

That doesn't change the total number of reachable (i.e. legal) positions.

Any progress with the pseudocode I mentioned here, by the way?

#393
It goes further than that.
Each pawn move, each capture, and each king move reduce the set of reachable positions.
It does not change the number of legal positions or the number of sensible positions, but it reduces the number of positions reachable in the course of solving chess.

This is an essential difference between weakly and strongly solving chess.

To solve chess no programming or code is needed, only use of existing programs like Stockfish on powerful cloud engines. It is the same what Sveshnikov did up to 1988 without any help, engines, or table bases, but now with powerful help.

If you post your pseudocode we can examine that.

The number of positions reached in the course of a solution depends on the exact method used. E.g. if tablebase construction is used only winning FENS need to be examined.

Stockfish will not give a solution. It's usually out of it's depth with mates of depth more than 40 moves even with only a few men on the board. You prove nothing that way.

I do think it's possible to use elimination processes to shorten the work load.. and with the help of neural engines such as Alpha zero for example running 24/7 it should theoritacally be possible. But memorization of these lines will be the impossible part

#395
No code or programming is required, only use of existing code e.g. Stockfish.
Construction of a 32-men table base to strongly solving chess is not feasible with present technology.
Weakly solving chess is feasible as Sveshnikov indicated: calculating from the opening towards the 7-men endgame table base.
Suppose we start analysing a position of C67:

From this position
no position
with white pawns on e2 or d2
with black pawns on e7 or d7
with 8 white pawns
with 8 black pawns
with 2 black knights
with a white light square bishop
with white or black castling rights
with 27 men or more
will ever be reached in the course of the analysis.

#396
"it's possible to use elimination processes to shorten the work load"
++ No elimination process needed: the non reachable positions just do not turn up

"with the help of neural engines such as Alpha zero" ++ The engine should have a simple evaluation function to calculate deeply towards the 7-men endgame table base

"But memorization of these lines will be the impossible part"
++ It may be possible to memorize a selection of 10,000 perfect games with a certain repertoire.

That is so obvious it really wasn't worth posting.

This ply count 0 position is the one you're supposed to be solving.

No legal positions are ruled out on the grounds of being unreachable from that position.

#399
As soon as you start analysing you move pawns, you trade pieces and you move kings.
Each time such an irreversible event happens the number of reachable positions dwindles.

Obviously. What's the relevance?

Perhaps if you posted a definite plan of action we could decide.

He is as you say just saying that as the game progresses, because some moves are irreversible, the reachable positions dwindle after such moves. Blatantly obvious.

That's why I asked for a pseudocode or a complete description of his proposal.

He originally proposed a sytem of takebacks and extra tries whenever an analysis from some position found a forced win, in which case you don't have a game of a set length it can grow indefinitely.

And you can't a priori rule out any reachable positions. If you take back irreversible moves the positions that were ruled out get ruled back in again.

The reachable positions from the starting position, which is the position to be solved, are exactly the legal positions, not a subset 1/10000000000000 th. the size.

Notice the dead silence on my point, and Llamas' about Sveshnikov's claim that all chess openings would have a "score" about a decade ago.

Those are hitting the soft underbelly here.

Also this one by Maratiggan:

"The reachable positions from the starting position, which is the position to be solved, are exactly the legal positions, not a subset 1/10000000000000 th. the size."

If Tygxc can't explain his way past these in simple terms, his game is resignable.  He already knew this, which is why he remains quiet about certain aspects of the inner workings of his bridge construction .

As for the list of openings that *are* "scored" according to Tygxc...they are "scored" exactly the way the Berlin was scored when it was thought to be bad for black.  That is, these openings have been analyzed with the current best means available, but imperfect means that change over time and can be proven wrong.

#404
"The reachable positions from the starting position, which is the position to be solved,
are exactly the legal positions"
++ No. Losing Chess starts from the exact same starting position and weakly solving Losing Chess only took 900 million positions.
http://magma.maths.usyd.edu.au/~watkins/LOSING_CHESS/ICGA2016.pdf 
If you consider all legal positions then you strongly solve chess, i.e. a 32-men table base.
That is not feasible as it would require 10^44 nanoseconds of time and 10^44 bits of storage.
Key to weakly solving chess is to consider only the sensible, reachable, and relevant positions.
Weakly solved means that for the initial position a strategy has been determined
to achieve the game-theoretic value against any opposition.
Strongly solved is being used for a game for which such a strategy has been determined
for all legal positions.

@tygxc

#406

The reachable positions from the starting position, which is the position to be solved,
are exactly the legal positions.

That is obvious if you accept that a position is legal if it's not illegal under FIDE art. 3.10.3

A position is illegal when it cannot have been reached by any series of legal moves

which means exactly what I just said. 

You say here

++ No, the number of legal positions is 10^44.
About 10^32 of these are sensible: can result from a game with > 50% accuracy.
About 10^19 of these are reachable: each pawn move and each capture is irreversible and renders huge numbers of positions unreachable.
About 10^17 of these is relevant.

So you claim that if you consider only reachable positions out of the set of positions you call legal and sensible the number reduces by a factor of 10^19 / 10^32 or 1/10000000000000. That is the number I included in the part of the sentence you quote that you chose to omit.

In fact there is obviously no reduction at all.

Actually I believe the number of legal positions you start with is vastly underestimated in competition rules chess and that your point about sensible positions is misguided, but we can come to that when I answer your points in #354, as I offered to do in #361; but before that you should post a description of what you propose to do that is detailed enough for me to precisely explain where it goes wrong and will not be subject to substantial change as we progress.

That assumes, of course, that you know what you propose to do.

#407

"In fact there is obviously no reduction at all."
++ In fact there obviously are massive reductions. The vast majority of the positions Tromp found legal are not sensible: cannot arive from a reasonable game with > 50% accuracy.
The vast majority of these cannot be reached in the course of the solving process. See e.g. the papers on how Checkers and Losing Chess have been solved with far less positions than are legal. The vast majority of these are not relevant: not necessary for weakly solving chess.
If all legal positions are considered then it is strongly solving and that is not feasible.

"I believe the number of legal positions you start with is vastly underestimated"
++ No, the 10^44 legal positions are calculated by Tromp. The 50 moves rule plays no role. The 3-fold repetition rule does not affect the positions, only the evaluation process should terminate once a position is repeated 3 times (or 2 times): the same FEN.

"your point about sensible positions is misguided"
++ No, it is important. The vast majority of the Tromp positions is not sensible e.g. because of the multiple underpromotions to pieces not previously captured which can never happen in a game of > 50% accuracy. Even the vast majority of the 10^37 positions calculated by Gourion cannot be reached by a game of > 50% accuracy and are thus not sensible: play no role in weakly solving chess with optimal play.

@playerafar

You say, "Positions that are legal on the board -  some of them 'couldn't have been reached'".

Legal positions are precisely those that can be reached. Whal do you mean by "legal on the board"?

Positions that are legal on the board - but  some of them 'couldn't have been reached'.
But the ratio of such positions could be far under half.
Instead of a crazy lopsided majority that only leaves less than a trillionth of the previous number.
Positions otherwise legal on the board but 'unreachable' usually have some exotic or unusual feature that makes them unreachable.
Such exotic features would be in a Minority of positions - not a majority.
And hardly an 'almost total' majority
and there'd even have to be very inefficient 'game/moves' analysis to find them as opposed to 'positions' analysis that gets linked to future mate or to tablebases.  

So.  A new approach to the subject begins to emerge.
1) Chess has not been solved and might never be solved in trillions of years or ever.
But - other things within chess have also 'Not Been Solved'.
2) Like the number of positions that are legal on the board - but legally unreachable.  Unsolved.  Unknown.
3) Approximate ratios of such - Unsolved.  Unknown.
4) Like the number of positions that are 'sensible and relevant'.  
Also Unsolved and Unknown.  Not even Defined !!  
5) And the Ratio of that category Unsolved and Unknown too.

6) Is even the Ratio of positions where there's lopsided material disadvantage but the other side has enough 'compensation' even known?
No.  Unsolved. Unknown.
7) Is there even a proper way to define that ?
No. 
8) Because among other things - of the weakness of the computers in that they cannot even correctly identify many draws that a D player can or could easily identify.
Also Unsolved !!

So there's a list of at least eight things that are actually unsolved.
A much more comprehensive description of the nature of the task and tasks.

Further breaking up the Snake Oil of ridiculous reductions of the number of positions in the first and overall task - and the crazy cosmetic semantic manoeuvering to try to make the hardware speeds of the computers irrelevant while that's obviously extremely central and relevant.

When a film of oils is messing something up - what does one do?
Chemicals are used that are 'caustic' to those oils and burn them off. 

But more Snake Oil will then be poured on the subjects.
Semantically-based repetitive snake oil rather than mathematically-logically based oil-free contributions.

Is this a legal position?

dxe6

I'm finally not 100% terrible at these... only 99%

#411
Yes legal: d4-d5+ e7-e5 d5xe6 e.p. +