alanback: That's a pretty radical change. It sounds more like a different formula than a bug fix! I can see how you expect a large difference in your rating when it's applied to the BKR history.

Though it's an improvement, I'd still like the ELO Backgammon formula to be used (like at Vog, Fibs and elsewhere) as it knows about the luck of the dice.
Your 2700 would reduce to 2200 or under with such a formula because of the diminishing returns that you mentioned.

It's even more pessimistic in its judgement as it predicts 30% loss of single-point matches between a top flight (2200) and an average player (1600). That's a big luck factor and it's entirely missing from the chess formula.

And I agree about the differential being too punishing. In our game I stand to gain 10 and lose only 5 and you vice versa. The proper formula would award us nearly equal shares of the match value.