The rating of a perfect player

Understood, though my point is that there are lots of possibilities.

Supposing two perfect players play, how long do you think the game will be?

However long, they will think that many moves ahead (and perfectly), so if there is any relation between chess rating and moves thinking ahead, perhaps that can be a guide.

Suppose computer scientists 100 years from now finally solved chess. They have a 32 piece table base. They calculate how to win or at least draw from any position from the start. They combine this engine with a random number generator. The RNG selects from any of the winning lines. Suppose after the first 10 moves the computer calculates it has a win in 20 a win in 19 and a win in 18 if it plays one of 3 different moves. It will pick one of the moves at random so as to not play consistantly and to play multiple lines.

Trying to use current rating systems to calculate it's strength won't work accurately. It is simply to strong of a player to rate. Imagine if superman decided to take up the sport of baseball. He would hit a homerun at every at bat. He would pitch a perfect game every game. How would you give him an era and batting average? 0 era batting avg 1000? That sytem is not designed for a perfect player. The Chess rating system is not designed for a perfect player either.

Not designed for a perfect player? Rating systems don't discriminate! The system doesn't care how good you are. And sure, if Superman hits a home run every time, then that'll simply show up on the statistics; nothing wrong there.

"It is simply to strong of a player to rate."

How so? A perfect player can be expected to get a certain result against a given player over many games, and that performance translates into a rating.

It depends on the solution to chess, and on what equation is used to determine rating.

A rating system could probably be made so that the perfect player's rating would be infinite, but I don't think current rating systems are set up appropriately.

Because every draw would be a perfectly played game. Why should rating be lost on perfectly played games? When your rating goes down that is a sign you made some sort of mistake, for this computer it did not make any mistake and played perfecty there for the rating system is flawed.

Frankdawg, you are overlooking that the rating system is comparing you to other people, not to some absoulte standard.  The perfect player loses rating points for a draw because someone ELSE played a perfect game, indicating that the perfect player was not quite so far ahead of that person as the former rating may have indicated.

If everyone in the world started playing perfect chess tomorrow, some people's ratings would go up and some would go down.  That's not a flaw in the ratings system.

Except they wouldn't win all games.  There's a finite number of moves possible for the opponent.  Even a random move generator would eventually stumble into a draw... and the average opponent would do significantly better than a random move generator.

I know!!! OVER 9000! LOL

The perfect player would be disqualified due to telepathic harrassment

You're forgetting chess is a puzzle

What if two perfect players play each other?

Answer this question and prove it.

No matter what the outcome, a perfect player wouldn't win every game because they would have to be White or Black.

You're forgetting chess is a puzzle

My point is that you can't assume a perfect player will always win. Suppose the solution to chess is white wins. A perfect player plays a near-perfect player, with the perfect player black. The game could result in a draw. Never assume.

 At some point, they no longer gain anything for defeating players rated greatly lower than them.  And that all depends on the population size.  Not exactly sure about the formula, but some of you Math whizzes can probably figure out the general form for a population "N", at what point the amount of change for a win becomes negligible...

It's an infinite regression, but it eventually bottoms out, and when the change becomes less than a full point, rating won't increase.

I honestly believe that a perfect player could beat Houdini 100% of the time, even as black. This might sound unbelievable, considering how often draws are amongst computers, but there is a huge difference between manually calculating things and having a table that plays completely perfect. It may only take a slight inaccuracy to put a "MATE in 168!" or something spectacular on the table.

If Houdini is estimated to be 3400 rating, I think there might be another computer in about 100 years time that can beat Houdini 99% of the time (about 4000 rating) and a perfect human player could beat that computer 99% of the time.

So the perfect human player could get about 5000 potential rating (though would probably never go above 3500 in this day and age, due to a lack of competition).

No psychology, just advanced inside information on how the opponent will react to different schemes

The perfect player would play like Botvinnik: http://www.chessbase.com/newsdetail.asp?newsid=2833

This would depend on Chess being "solved", which, by-itself, is a matter of debate:

http://en.wikipedia.org/wiki/Solving_chess

Firstly, it should be 13^64 because of empty squares-- you're right, this is an upper bound on the number of positions but a game can have up to 12000 positions, so an upper bound on possible games is (13^64)^12000 = 13^768000 >> 10^10000. Obviously there are no where near as many as 13^768000 games, so the 10^10000 figure is probably more useful.