17-04-2013, 01:32 PM
We use FIDE ratings as start grade CS ratings for visitors.
The calculations are pretty much as Hamish mentioned.
Players more than 400 different are assumed to be exactly 400 different so the 1830 player is worth 1849 in the calculations of both TPR and expected scores.
The reason why TPR can be flat but actual-expected positive is because the expected score data is a curve rather than linear. So taking average opposition and taking each game individually gives a different result.
The calculations are pretty much as Hamish mentioned.
Players more than 400 different are assumed to be exactly 400 different so the 1830 player is worth 1849 in the calculations of both TPR and expected scores.
The reason why TPR can be flat but actual-expected positive is because the expected score data is a curve rather than linear. So taking average opposition and taking each game individually gives a different result.