26 Jan 12
Hello,
is there an estimate of the maximum and minimum possible elo rating?
If you think about the ideal chess program it should have the highest possible elo. It should be above 3200 which is for present best programs.
If you think of a program that moves randomly than you can easily beat it with very low elo. If you have program that moves intentionally in the worst way it should always loose even against the random program. Elo for this should be largely negative.
Did anyone try to estimate these extremes?
Miro
27 Jan 12
Originally posted by kundasIn theory that's just a function of the number of people in the group you estimate the elo from.
Hello,
is there an estimate of the maximum and minimum possible elo rating?
If you think about the ideal chess program it should have the highest possible elo. It should be above 3200 which is for present best programs.
If you think of a program that moves randomly than you can easily beat it with very low elo. If you have program that moves intent ...[text shortened]... . Elo for this should be largely negative.
Did anyone try to estimate these extremes?
Miro
* Of course you can't go below zero.
But if we consider a group which has a number of people for which we have the strict rule player 1 is better than (wins always against) player 2 written as 1>2
So we can have a group of people with n members and the boundary condition pn-1>pn. Then we start with all people with a ceratin elo number, say 1200. Then we make a complete matrix of games (every player plays n-1 games) and repeat that ad infinitum. We then get player n with an ever decreasing rating number until player n-1 reahes about 400. And player one reaching (in theory) something like n times 400.
This is of course gray theory. In real live people win, draw or lose according to many factors and the elo number is a convininet means to see relative strength in a given plenum.