Chess will never be solved, here's why

So how do you explain 5 blunders in 79 positions in the sixth game here? Was it just rotten luck?

(We can look at the difference between 10^17 and 100 x 2.9 x 10^44 x 3^(100 x 2.9 x 10^44) later.)

#1943
"In fact @tygxc appears, on the whole, to propose solving the competition rules game, though he's a little evasive on that question." ++ No, I have said several times: forget the 50 moves rule. The 50 moves rule is a practical rule to make games and tournaments end in a reasonable time, but for the purpose of solving chess it can be considered unwritten.

"The factor "100" here represents positions with the same diagram and side to move but with different ply counts (which can have different win/draw/loss evaluations"
++ No, forget the 50 moves rule. Also there are not twice as much positions as diagrams. For every position with black to move there is a diagram with white to move. Even more: for every position with lost castling rights there is a left / right image. So each diagram represents 2 positions.
Example:
Position = diagram with white to move.

Exactly the same with black to move:

Exactly the same position, white to move:

"If you look at the SF14 v. SF14 KNNKP games in this post SF14 blunders between 1 and 9 times under basic rules in each case."
I like this argument in principle, but there are 2 major arguments against it.
1) You talk about Stockfish 14 on a desktop of say 10^6 nodes/s for what time per move?
I talk about Stockfish on a 10^9 nodes/s cloud engine and at 60 hours/move.
2) You found some instances where the Stockfish top move was a mistake as indicated by the table base, but did you find any incidence where the Stockfish top 1 move, the Stockfish top 2 move, the Stockfish top 3 move as well as the Stockfish top 4 move were all errors?

#1950
++1 error per 10^20 positions means
at 60 h / move on a 10^9 nodes/s cloud engine and after 3 takebacks per white move:
top 1 move, top 2 move, top 3 move, and top 4 move are all errors.

We'll hit 2000 posts soon.  And many of those posts will have been invested in the 50 moves rule -
which is one of the 'competition rules' rules.
Frankly - I think tablebases should ignore it.  And ignore repetitions of positions too.
Completely.  Except for repetitions that are forced or can be if needed.
Like availability of perpetual check.  Because that's Forced Repetition.  
Legally Forced.
There's other kinds of forced repetitions.  Usually by deterrence though.
Is there any way to have a legally forced repetition of position that's not by perpetual check ?
There are other ways but they've got to be found and noted.

Like with white pawns at a4 and d4 and f4 and h4 -
and black pawns at a5 and d5 and f5 and h5.
With the two Kings each stuck behind their pawns.
Repetition of position is forced by legal compulsion - without any checking even possible.  Neither side can make progress and the game is a draw.
One doesn't need a 50 move rule for the computer to find and solve such positions as draws.

Drawn. 
By legal compulsion.

Neither side can make progress.

The next one is better.

Now you can give white a darksquare bishop or even five darksqare bishops and give black up to five lightsquare bishops - and its still dead drawn.  
Could add some doubled pawns behind front pawns too - or instead and its still dead drawn.  What difference would it make if the bishops 'won' those pawns ?
Point:  computer's got to find them - as opposed to being concerned with 50 move rules or unforced repetitions.  

#1947
"Not necessary to consider alternatives to what? The move you know is right?"
++ White tries to win, black tries to draw.
If a series of moves ends in a tablebase draw, then all black moves were good enough to achieve the goal of drawing and it retroactively validates all black moves as optimal by the authority of the table base.
If the series of moves ends in a table base draw, then white failed to achieve his goal of winning, so alternatives for white needs exploration, taking back all white moves in a 1st, 2nd, and 3rd validation step.

If you can reach the 7 man tablebases in 3 moves from the starting position, you're a better man than I. SF14 is usually not far short of 60.

#1956
"If you can reach the 7 man tablebases in 3 moves from the starting position, you're a better man than I."

++ No, you misunderstand.
First I advocated before the 3-prong method like in solving Checkers and Losing Chess.
That is also what Sveshnikov hinted at.
1) Good assistants with ChessBase set up 26-men tabiya as starting points for the computer.
2) Then the modern computers calculate from the 26-men tabiya to the endgame table base.
3) Then the exact evaluation draw / win / lost is retrieved from the 7-men endgame table base.
So I will not start from the 32-men initial position but from a 26-men tabiya.

Second, I did not mean to reach the endgame table base in 3 moves.
I mean to consider 4 candidate white moves.

"and it retroactively validates all black moves as optimal by the authority of the table base."

No it doesn't. 

But we'll keep seeing posts like the one quoted in italics. 
2000 posts mark coming up.
At 3000 posts - there'll probably be at least 100 more like that.

Exactly.

Thank you Martin !  

And - whether the 50 move rule is invoked or not -
it will not change the uniqueness of any chess position.
What Does need to be addressed is Ambiguity in positons.
Ambiguity can't cut it in math and math-related.
Unless you want to have unsolved equations.
So if en passant is possible in 'a position' 
then its no longer 'a position'  - its two positions -
with each one unique to the other and unique to all other chess positions.

Similiar with castling ?  Not quite - but close -
if its an endgame tablebase then obviously it wasn't generated from the opening 32 piece position !  
Which means that anytime either King stands on its original square with at least one of its rooks on its original square - then castling rights must be both preserved And denied.
In other words - with just one rook (and its King) in place - then that must be regarded as two different positions.
With both Kings in place and all four rooks at home - then that must be regarded as sixteen different positions !  
Does that greatly increase the amount of work to be done by the tablebases ??
No No No  !!
Because a very small percentage of simplified tablebase positions would have at least one King at home and at least one of its rooks at home ...
far far under 1% in fact.
It doesn't take a mathematical mind to realize that.
Just a slight slight effort.

Of course it will.

Try mating from this position on the analysis board.

We'll disagree on that one Martin - 
but you might 'get me' on how I worded it ! 
I should could would have said - 'uniqueness on the board'.
Qualification.  
Too late !  

@playerafar

A position with e.p. and a position with not e.p. is still uniqueness on the board if they have the same diagram. Chess works by positions, not diagrams.

@tygxc's proposed "solution" uses SF14 which evaluates positions differently depending on the ply count.

@tyxgc

Just realised why you were asking if I could find a position where Stockfish's first four evaluations were all wrong.

As I already said, it's not relevant, but it is quite easy.

This is an 8 second evaluation (time which gave the lowest blunder rate in the KNNKP position I posted earlier).

The only move to win in either game is ...Qg2+