Re: [pbmserv-dev] ratings

["Claude Chaunier" <claude.chaunier@wanadoo.fr>]

Randall Bart interestingly wrote:

> In games with randomness, the best player in the world 
> will occasionally lose to barely competent player.  OTOH, in a game without 
> randomness, this does not happen.

It's more like the gamut of levels in Backgammon is narrower, isn't
it? Say, a Go game is like a match of one million backgammon games.

Even one backgammon game wouldn't be lost 20% or 5% of the time in
the conditions you state, I'm sure. More like 0.1% ?

And one Go game out of 10^10 could well be lost too - if only by
a random player.

But what you write then makes sense:

> tweaking for how random the game is.  A game where a great player loses to 
> a mediocre player 20% of the time should be rated differently than a game 
> where he only loses 5% of the time.  Taking a backgammon rating system and 
> applying it to other games is not necessarily a good thing.

I find this very interesting. Maybe it's only a quantitative
problem, a matter of different scales, and we don't really
care about how much stronger it means to be rated 100 points
higher in a game compared to another, but I don't really know.

Claude Chaunier