A perfect game of chess is always a draw. Discuss.

@112

"It seems we just have different standards of "proof"."
++ Here is proof by reductio ad absurdum.

Fact: 97 games all draws in the ICCF World Championship Finals,
by ICCF (grand)masters with engines.
10 exceeded the 50 days / 10 move time limit in otherwise drawn positions.

Hypothesis: chess is not a draw.
Then all 97 games must contain an odd number of errors (?).
An error (?) is a move that changes the game state from draw to loss or from won to drawn,
and a blunder (??) changes the game state from won to lost and counts as a double error.
So all 97 games would contain 1, 3, 5 ... errors and none would contain 0, 2, 4... errors.
It is absurd that some games would contain 1 or 3 errors and none would contain 2.
It is absurd that some games would contain 1 error and none would contain 0 or 2.
It is absurd that the ICCF (grand)masters and their engines would conspire to always make an odd number of errors and never ever an even number of errors.
The hypothesis was false.
Thus chess is a draw.

Q.E.D.

Hypothesis: chess is a draw and the 97 games are not perfect.
Thus all 97 games contain an even number of errors: 0, 2, 4...
It is absurd that games would contain 0, 2, 4 errors and none would contain 1 error.
It is absurd that the ICCF (grand)masters and their engines would conspire to always make an even number of errors and never ever an odd number of errors and never make 1 error.
Thus all 97 games must contain 0 errors.
The hypothesis was false.
Thus all 97 games are perfect games.

Q.E.D.

The inferential reasoning is that the game of chess starts off with a slight initiative to white, which some people consider to be worth about a fifth of a pawn to a third of a pawn, depending on the opening. It is considered that this slight initiative gradually lessens if both sides continue to play good moves, until at a particular stage in a game, the initiative has dwindled to nothing. It would take a inexplicable reversal of that trend to make a game won by one side or the other. There's no reason that there should be such a reversal, unless one side or the other blunders. Therefore it's drawn with best play and that is an inferential proof. There's no deductive proof possible. Consequently, that which I've just given is the best we have to go from.

I don't know if it's available in any books or theoretical papers anywhere, because I just put it together in my head: but it's correct.

@119

"but this fact will nowhere be used, it's just thrown in for padding"
++ Of course the time per move and the engine speed are relevant.

It's not relevant to your argument because nowhere do you use it.

You cannot draw valid conclusions from letting an engine blitz against itself.

I can draw the valid conclusion that your argument doesn't hold water. It contradicts the conclusion when applied to that case.

If you are serious about your experiment, then take your 2048 second/move game,
then go forward until the error in respect to the 7-men endgame table base,
and for that move increase the time/move to 5 days. Is the error still there? Yes/no?

No point. Your argument nowhere takes the time per move into account. As I said you just throw it in for padding. If I'd made them all 1s per move it would still prove your argument invalid.

"the initial position under RCCF rules, which differ from FIDE rules"
++ Nobody but you is interested in RCCF rules, but in FIDE rules and/or ICCF rules.

Your argument nowhere takes into account the rules used in the sample. As I said I could have used a game of draughts to discredit it. 

"Indeed your conclusion is that chess is a draw, not that ICCF chess is a draw."
++ What is discussed here is 'A perfect game of chess is always a draw.'

Which is ambiguous, because there are multiple versions of chess with different rules and possibly different theoretical outcomes. No statistical argument purporting to show the result for all from a sample taken from only one will be valid. Your particular argument is invalid even for the same version as I just demonstrated.

++ A game of chess starts from the initial position.

Which fact again is nowhere used in your "proof". So if it were valid it would be valid for any starting position.

And it's not.

@124

"nowhere do you use it." ++ Of course I use it. If I were to take the Blitz World Championship or the Rapid World Championship, then I would not have 100% draws to start with.

You use it implicitly in your choice of sample, but nowhere in your argument from that sample. Ir's the argument that is invalid.

"Your argument nowhere takes the time per move into account."
++ It does, as only long time per move and strong masters arrive at 100% draws.

A random legal move generator will generally produce 100% draws. 

"If I'd made them all 1s per move it would still prove your argument invalid."
++ No. You take a won position and your engine fails to win at its set time/move.
I take the drawn initial position and the ICCF (grand)masters with their engines now arrive at reaching 100% draws. Your setup fails to reach the game-theoretical value.

They both arrive at 100% draws (some agreed draws). Only your big red telephone tells you yours reach the game-theoretic value. A proof that the starting position is a draw that starts by assuming the starting position is a draw could be regarded as circular.

But I will post a set of SF v SF games that do reach the game-theoretic value of draw from a Syzygy verifiable drawn position and also contain blunders.

ICCF WC Finals now succeed in reaching the game-theoretical value.
In previous years they came closer and closer, but did not yet reach it.

According to your big red telephone. The games I'll post will verifiably reach the game theoretic value.

@121

"rely on the assumption that chess is a draw"
++ No.
Hypothesis: chess is not a draw.
The hypothesis is inconsistent with the observed facts 97 draws in 97 ICCF WC Finals games and thus the hypothesis is false and thus chess is a draw.

@132

"I noticed some faults in your argument."
++ I will try again.
Hypothesis: Chess is not a draw.
Look at the ongoing ICCF World Championship Finals, the strongest chess on the planet, 5 days/move average, engines allowed, 17 ICCF (grand)masters that qualified.
All 95 games are draws (or 10 by 1 player exceeded time limit in drawn positions).
If chess is not a draw and if all 95 games end in draws, then all 95 games must contain an odd number of errors.
It is absurd, that games would contain 1, 3, 5... errors and none would contain 0 or 2 errors.
Thus the hypothesis was wrong.
Thus Chess is a draw.