Chess will never be solved, here's why

unlike this meme

#1123
"None of them are draws. They're all wins for Stockfish."

That is not true. Stockfish versus Stockfish is a draw for all four 1 e4, 1 d4, 1 c4, 1 Nf3.

#1120
"a simple calculation shows that even a partially full solution of a 60 move game would take thousands of years"
Most ideal games end long before move 60, see ICCF World Championship games.
https://www.iccf.com/event?id=85042 
Even with 60 moves 4^60 = 10^36.
We only have 10^36 positions.
Weakly solving chess does not need to visit all 10^36 positions, but only about 10^17.
That takes 10^8 seconds = 3.16 years on a cloud engine of 10^9 nodes / second.
On a desktop of 10^6 nodes / second it would take 3160 years.

In a solution it's the long wins that take up virtually all of the complexity. Draws are likely to be longer, and the plausibility of a short forced win is extremely low.

It's worth bearing in mind that while real games get agreed drawn, to prove a draw, your strategy needs to deal with ever possible way to play on that is not disastrous. The opponent can aim to make it as long as possible before a draw is proven.

4^60 is a number that makes no sense at all. This would correspond to 2 options for each player at each move! Every option for one side needs to be dealt with (typically around 40). So, thinking of games, you get more like 40^60 even with the assumption that games are remarkably short. So let's make that 10^63 years and realise it is not helping.

"I'm only trying to be polite"  - calling people fools and trolls isn't polite.
No need for me to read that post further.

"So let's make that 10^63 years and realise it is not helping."
Good point.  Well summarized.

A way to go at the topic in general - is to talk about what's easy first.
What's been done as far as 'totally solved' in chess is concerned.
By both players - and by computers.
Not only what's been done - but what can be very quickly re-done.  Fast.
To start that off - one can start from opposite ends of the task.

Whether for players/computers - two Kings only.  Instantaneous Solve.  Draw.
Whether for players/computers - 32 piece original position.  Instantaneous - Unsolved. 
And from there - only 20 positions possible in one move -
and only 400 positions possible after two moves. 
Half moves that is.  Known as ply.

But whether adding pieces to endings - or adding moves to openings -
it starts getting much more difficult when you go to three pieces - or 3-ply for openings.
Add a pawn to two Kings - not an instantaneous solve for most humans.
But for computers yes.
Add a ply to 2-ply from the opening - difficult for humans to count up the possible positions. 
But still instantaneous for computers to count/itemize positions but continues Unsolved. 

Objections:  Hey !  The Duras and Englund gambits Lose !!
Even if that could be thoroughly proven - there's other 3 ply positions that are 'thoroughly unsolved'. 
So 3 ply continues unsolved - whether with 32 or 31 pieces onboard.

Easy: - realizing it is 'not helping' to add more 'ply' to openings except for computers to count positions - not solve them.
Now easy: - for computers to 'solve' endings with less than 7 pieces it seems.
Because the tablebases have been made already.
But even at just 7 pieces onboard - already there's struggling - castling and en passant considerations not factored in yet.  

Never said either.
But - logging off soon.
You will find somebody to play ping pong with perchance.

Regarding 'here's why' in the forum topic -
that is being made abundantly clear ...
supercomputers cannot even solve for 7 pieces thoroughly.
(castling and en passant possibilities had to be skipped)
per the Wiki article Here:

https://en.wikipedia.org/wiki/Endgame_tablebase

With material on the board going up to as many as 32 pieces -
the daunting nature of the task is being made clearer and clearer.

The blindness of chess engines (and NN engines) to accurate play in many such positions is consistent with current best engines being very weak rather than close to perfect. An optimal chess player has some quantifiable rating by virtue of the results it achieves against existing engines, and it is not impossible that it expected score is close to 100% and the rating difference is therefore huge. The depth needed to ascertain best moves is crucial to this, because there is little doubt that increasing depth demands more computing power at a new constant (logarithmic rate) and that the rating achieved by increasing depth goes up at a decreasing rate. The key thing is the way in which the rate of increase of rating with additional depth decreases (it used to be 40 points for a doubling in CPU speed, providing one ply of extra depth).

The relationship needs to be some function that is 0.0 at additional depth 0 and whose slope is always positive and always decreasing.

I haven't seem a paper on the relationship between CPU speed / depth of analysis and rating. I would observe that it is definitely engine-dependent. Even for conventional engines, how clever the algorithm is at being selected affects the relationship between speed and depth and also affects the relationship between each and rating. Also AlphaZero was achieving a larger increase in rating for a doubling in computing time than Stockfish in the DeepMind research.

It's amusing how every time you misremember something you blame it on people deleting their posts instead of realizing the truth...you don't find some of the things you go digging for because they were never said.  You've been asked to produce such things many times now, and you have failed on every occasion to produce something you've attributed as an exact quote.

You still carry around some fantasies about things I have told you in PMs (PMs that we've never had ...).  You tried to claim Elroch said things he did not even come close to saying right here in this thread...so it's not like you are forgetting things because they are slowly fading away, either.

Your high opinions of other posters who are primarily known for trolling the same threads endlessly is misplaced as well.  But I suppose that isn't surprising...being unable to discern the difference between weakly solving a game and strongly solving a game is pretty much the same as being unable to discern the difference between weather and climate.  Neither lack of understanding is particularly "clever".

You tell me .

Not really.  You're the one with the "record" here.  I've never been muted for trolling...so I should reasonably defer to your expertise on the subject...

P.S. Playerafar and I do not get along, which I believe has been mentioned twice already.  It's only in your head that these dotted lines of secret trolling conspiracies exist.  If you were talking about Elroch, we have communicated barely at all outside public threads, and I would say have at most a mutual respect for the amount of BS each puts up with (and perhaps a shared love of accuracy, precision, and reputable sources in discussions), at present. 

I'm not precluding future possibilities of friendship, mind you .  But at present it would be a big stretch to call us friends.  Who has the time, really?  These forums are a busy place for people that like to stop the spread of misinformation...

That just goes to show we really don't know what the result is.

I only need two words to solve chess.

Git gud

Correct.  Stockfish versus Stockfish proves nothing.

@Optimissed, the extreme example among computer chess players of selectivity are the NNs like AlphaZero. It was looking at about 0.1% as many variations as Stockfish in their match, I believe.  Humans, of course are way more selective still.

that refers to this match in 2019?
https://www.chess.com/news/view/updated-alphazero-crushes-stockfish-in-new-1-000-game-match

I can understand a program that looks at less calculations winning - because its programmed to understand what matters more.
As opposed to being 'selective'.
With human players - its very pronounced that stronger players are much more efficient in choosing and performing their calculations.  
Idea:  start with the two Kings' and  two Queens' situations as opposed to 'what they can do' and then work down through the rooks/bishops/knights/pawns for both sides.
Stronger players also know better how to look at squares (whether occupied or not) as opposed to pieces. 
Priorities versus 'brute force' calculations and crunching.  
To do this right - observation needs to be distinguished adequately from 'calculation'.  

#1145
"4^60 is a number that makes no sense at all. This would correspond to 2 options for each player at each move!"
++ No.
It corresponds to 4 options for the white player of a drawn game. If the candidate ideal game ends in a draw, then that retroactively validates all black moves as good enough to draw.
As calculated before 4 options for the white player correspond to less than 1 error over the number of positions. 4 options for the white player also corresponds to the number of positions needed for weakly solving chess: 4^30 = 10^18

"Every option for one side needs to be dealt with (typically around 40)."
++ That is not true. In the initial position white has 20 possibilities, but most of these are clearly no ideal play. Just two options 1 e4 or 1 d4 are probably enough. Two more options 1 c4 and 1 Nf3 may complement it.

"So, thinking of games, you get more like 40^60 even with the assumption that games are remarkably short."
++ No.
Thinking of games is wrong. There are many, many transpositions that reach the same position with a different game.  There are e.g. maybe 10^40 ways to just arrive at the position after 1 e4 e5. They can hop around with knights, then white can play e3, then they can hop around with knights, bishop and queen, then black can play ...e6, then they can hop around with knights, bishops and queens, then white plays e4, then they hop around with knights, bishops and queens returning them all to their initial squares and then black plays e5.
60 moves is not remarkably short, it is extraordinary long. Most games are either finished or within the 7-men endgame tablebase before move 40.

"So let's make that 10^63 years and realise it is not helping."
++ No.
10^63 years = 10^63 years * 365.25 days/year * 24 hours/day * 3600 seconds/hour = 3*10^70 seconds. 3*10^70 seconds * 10^9 nodes/second = 3*10^79 positions. Chess does not even have that many positions. Even for strongly solving chess to a 32-men table base only about 10^36 positions need visiting. For weakly solving chess i.e. providing an ideal game with proof that all moves are optimal only about 10^17 positions suffice.

#1166
"Alpha Zero is therefore a great improvement in programming strategy."
No. AlphaZero defeated a weak version of Stockfish.
In a later TCEC superfinals Stockfish with a simpler evaluation function defeated LC0 with a more elaborate evaluation function. The reason was that Stockfish could hit the 7-men endgame table base earlier and more often than LC0.

#1102
"Not very ideal for white though."
How do you know it is not ideal?
Here is another candidate ideal game starting with 1 d4.