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

@126

"A random legal move generator will generally produce 100% draws."
++ I do not know that.

SF8 is simulating a random legal move generator in the examples below if you look through them. 100% draws.

It is disrespectful to compare 17 ICCF (grand)masters with engines at 5 days/move in the ICCF World Championship Finals to a random move generator.

Then stop doing it. I made no such comparison.

"They both arrive at 100% draws"
++ The ICCF (grand)masters did not arrive at 100% draws in previous years.
They now reached 100% (except if the ongoing games would produce a decisive game).

Relevance?

"some agreed draws"
++ Yes, they agree on a draw when neither side has no hope of winning. You cannot expect them to play on for months until a 7-men endgame table base draw or a 3-fold repetition.

I already said I don't. Again - relevance?

"A proof that the starting position is a draw that starts by assuming the starting position is a draw could be regarded as circular."
++ No, I start with the hypothesis that chess is not a draw.
Then I demonstrate this hypothesis contradicts the observed facts and thus the hypothesis is false and the contrary of the hypothesis is true.

Indeed you do, so you have no grounds for dismissing the games I posted on page 1 (starting from a winning position) as counterexamples to your "proof". My comment was in response to your objection

++ 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.

Which suggests your poor befuddled brain is assuming that your argument depends on the initial position being drawn. 

"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."
++ But at what time per move?

Your "proof" involves no reference to time per move. Your comments at the start regarding time are just padding; never referred to in you proof. So the question is irrelevant. (But in fact you can run games from the starting position in the games I post here with any feasible time per move - the result will be the same.)

Some previous year ICCF WC Finals at 5 days/move but with less powerful engines and less knowledgeable ICCF (grand)masters ended decisively.
There probably were games with two errors that undid each other.

"In previous years they came closer and closer, but did not yet reach it."
++ That is a fact: the number of decisive games became smaller and smaller and is now zero.

How very interesting. Are you responding to my post?

"The games I'll post will verifiably reach the game theoretic value."
++ If you reach a draw and you have an even number of mistakes that cancel out, then the same procedure will also generate games with 1, 3... errors, i.e. games that do not reach the game theoretic value.

Well, I tried it and they didn't. So yah boo.

On previous years' ICCF WC Finals I used a Poisson distribution to estimate the number of errors from the observed results. Now I no longer need that: they have reached 100% draws.

And it was explained to you at length in previous threads exactly what was wrong with that.

@143

"There is also another problem with your use of "error" in that sense" ++ I defined it.

"chess players think in human terms regarding error"
++ No. 'I have attached question marks to the moves which change a winning position into a drawn game, or a drawn position into a losing one, according to my judgment; a move which changes a winning game into a losing one deserves two question marks ... I have distributed question marks in brackets to moves which are obviously inaccurate and significantly increase the difficulty of the player's task ... There are no exclamation marks, as they serve no useful purpose. The best move should be mentioned in the analysis in any case; an exclamation mark can only serve to indicate the personal excitement of the commentator.' - GM Hübner

"An error is a move which significantly makes the desired result harder to obtain"
++ No, harder to obtain is subjective.

Sorry but that is nonsense. I think you should forget everything that Hubner or Svesnikov propose and start to try to communicate more effectively. That is what nearly all good chessplayers would think an error to be. They simply make a mistake. It need not change the game assessment. You seem to have become hidebound, using phraseology that's out of date and ineffective for clear communication. 

"A blunder is a move that changes the game assessment." ++ No, blunder is win to loss.

"An error is what Chess.com calls an inaccuracy."
++ No. Inaccuracies do not exist. Either a move changes the game state or it does not.
If it changes the game state, then it is an error. If it does not, then it is not inaccurate.

For Heaven's Sake, grow up!!!!

"the engine can't assess what is accurate play and what isn't"
++ The engine cannot, but the final outcome win/draw/loss can in retrospect.

"The engines make a lot of mistakes in deeply positional games"
++ Yes, but less so with more time and better human guidance.
With unlimited time they would make no error at all, as chess is a finite game.

There's no such thing as unlimited time.

I agree on all this.

I think @tygxc was probably referring to finite but unbounded time (or maybe five years on a few cloud cuckoo computers).

Everyone agrees that unassailable proof (analysis of every possible move from the starting position to the ending tablebase) is impossible to achieve, so we are all nit-picking the amount of evidence each person feels is sufficient to satisfy them that the game is a draw.

That is not chess, it is guess the miss!