Perfect play

[corres]

Is there any play in chess what would be called "perfect" ?
Botvinnik and his colleague have searched for "truth of chess" but they did not find it.
Naturally I talk about playing chess from starting position.

[yurikvelo]

Is there any play in chess what would be called "perfect" ?
Botvinnik and his colleague have searched for "truth of chess" but they did not find it.
Naturally I talk about playing chess from starting position.

There is perfect play for TB7-Lomo positions, but Botvinnik won't come anywhere close to perfection even in 7-men position

[corres]

If 32 men (Lomonosov?) database will be then we shell talk about "perfect play".
Play of chess engines are far from perfect.

[megamau]

Some replies.

Use the tool "eloGap" in 40H-PGN tools

Norm, I can't seem to access your site from Singapore. Is there a mirror or could you send by email ?

100% draws does not necessarily mean perfect play. It only means that the difference of the imperfections is not large enough to push closely matched players out of the draw zone.

100% draws does not implies perfect play, but perfect play implies 100% draws (I assume the starting position is a draw. Otherwise perfect play will reach 100% wins for one color).
When talking about approaching perfect play, there is no "small imperfection" which "stays in the draw zone".
A move is either correct and keeps the game theoretic value (a draw) or is a mistake and translates into a win for one player. The only way it reverts to a draw is by another mistake by the winning player.
So if two players draw every game, they are always making the exact same number of mistakes. For high numbers of draws (let's say 200k in a row) it is much more probable that they are making 0 mistakes each rather than 1,2 or 15 mistakes each every game.

In TCEC superfinal, engines gets 90% of Draw easy even on imbalanced opening. They play with a strenght around 3500, so I don't think the graph here is accurate

Is the time/hardware advantage of TCEC worth 150 elo ?
Also, please consider that this is an extrapolation. Perfect play could be nearer than we think or farther. I think 3800 to 5000 elo is a good guess.

[Henk]

Maybe after some years some engines only play draws against each other. But then we have the problem who plays better for playing more games doesn't work.

[Norm Pollock]

Some replies.

Use the tool "eloGap" in 40H-PGN tools

Norm, I can't seem to access your site from Singapore. Is there a mirror or could you send by email ?

Hi Maurizio,

First time with this problem. I suppose something doesn't trust my site. Try this: Here is a link to the html code of the my links page (40H.html). Download (save) it, then open it as a file (File/OpenFile) in your browser. It "should" be the same as opening it from the "www" below.

http://www.mediafire.com/file/4be5t525b5u2l21/40H.html

-Norm

[corres]

[quote="megamau"]Some replies.

100% draws does not implies perfect play, but perfect play implies 100% draws (I assume the starting position is a draw. Otherwise perfect play will reach 100% wins for one color).

Is there anybody who proves that chess is a game what leads to draw theoretically? I do not know that man. The state "..perfect play implies 100%
draws" is a hypothesis only.

[hgm]

No one really doubts that hypothesis, however. A strong indication is that the white advantage is only some 53-54%, while Pawn odds is a 70% advantage, and most simple end-games are very difficult to win if you are not more than a Pawn ahead.

[corres]

Undoubtedly, chess play tends to draw.
However true proof may not be derived from games played imperfectly.
But what is the perfect play?
As you have stated well the draw result is not an evidence for perfection.

[megamau]

Thanks to Norm Pollock utilities I managed to better filter the data.
Now I have the draw rate of all engines when facing opposition within 20 elo.

The results are similar, although the linear extrapolation is a bit nearer, at 4400 elo.