Gary:

"... but does NOT mean that it is an advantage or a disadvantage to play a high or low rated player. It also means that the range of ratings from the top player to the lowest player will be much smaller than in 100% skill-based games."

Sorry I don't agree with this. The problem with playing lower rated players at backgammon with the USCF rating system is the reward/penalty favours the lower-rated player. You risk around 30 points to gain 2 when his chances of winning are say 40% - it just is not worth the risk. We were in this situation at Gold Token and in the end the top players refused to play the bottom players for this reason.

The number to the left is the minimum, the one on the right the maximum. Ignore everything else!

Following on from the comments below and having seen how the ratings system has worked at GoldToken and other places I would like to offer some observations:

1) As others have said, the USCF ELO system does not work at all well for Backgammon and its variants. There is too much luck involved in these games and soon the higher-rated players will refuse to play the lower-rated ones because they lose too many points if they lose the game. FIBS is a much better system for these games and its introduction at GT has resulted in the top players lists becoming much more representative of ability. FIBS is probably better also for the other games which involve luck such as the Battleship-type games.

2) The Provisional Ratings formulae used by the USCF ELO system are designed to facilitate a small number of new players reaching their correct rating in a large pool of established players. On a site such as this the opposite applies - a relatively large number of new players with no real established base. I don't think they are of any benefit here and I would simply get rid of them. ie just use the established formulae. They are particularly inappropriate for Backgammon etc as they quickly take you to an (almost) arbitrary rating (see 1 above) from which it can take around 300 games to approach your true rating.

3) The treatment of multi-game matches is I think a little simplistic. I think common sense suggests you should receive more credit for winning a long match than for a short one. Fortunately multi-point matches are common in Backgammon and FIBS copes with them simply by multiplying the rating adjustment by the square root of the match length. So you get 3 times as many points for winning a 9-point match as you would for a single point one. An n-point Backgammon match is equivalent to an n-wins match here. There is no such thing as a draw in Backgammon but the same formula could probably be applied to an n-points match. I think an n-games match can be considered as an (n+1)/2 - wins match. I'm not sure how you would treat a tied n-games match - again there is no such thing in Backgammon.

4) Ratings based on all matches played are easily manipulated by the unscrupulous. They would be more meaningful if based on tournament matches only.

I tried England, Scotland and Wales. They work fine.