Looks good. A couple thoughts:
- I missed the 16+ part before the bracket in your formula for K. Maybe it would be better with only a bracket (16, 16+stuff, 32).
- I think you don't want to say "proportional to." X is proportional to Y when you can write X = cY for some constant c. That would look like a line through the origin (a ray). Also, even the basic idea of the sentence — that the change is greater when you're further apart — is not true. When you lose, your loss is smaller when Delta is larger.
Except for that sentence, I really like the first paragraph.
A couple things that might be worth adding, but aren't strictly needed:
Elo is designed so that when you reach your true ranking, you should stay there. That means that the thing being subtracted from S in formula (2) is the expected score, your probability of winning. $E[ R_i^{NEW}|R_i,R_j] = R_i$. I just realized that I was using delta differently from you. I mean it as the difference between the two original scores, which is on your x-axis (not as the change in R) when I mentioned it above.
And maybe it would be useful to talk about the wagering analogy (which only applies in the special, though quite common, case where K_i = K_j). Each player wagers/gives up some of their Elo with
- the favorite to win having to put in more,
- and the winner taking the pot (which is the sum of the wagers, which is always K).