Rnew=2/3Rold+1000((N-i)/(N-1)+((N-i)(N-1))^3)(1+SumRold/6000) and divide by 1000 or 100 or 10 to get a not-so-large number, and add (+ kGw-k1Gl), where k and k1 are coefficients that will depend on the rating. If the rating is low, say up to 500, then k will be 3, and k1 = 1. If the rating is from 500 to 1500, then k = 1.5, k1 = 2, and if the rating is from 1500 to 3000, then k = 0.75, k1 = 3. If it's too low, you can divide it into more gradations. (THIS IS JUST MY EXAMPLE)
What are your thoughts?
When are we going to implement this?:p
(I haven't read the whole thread, it's hard to understand the notations)
What is the purpose of the gradation and the dependence of the coefficients K and K1?
Does it turn out that we derive it empirically?
