Idea:

We can access HTML pages like https://mirror.codeforces.com/contestRegistrants/*/page/*, it has registrants and the rating. We can loop all the registrant pages, exclude out of competition users and store them to cache. After done, we can predict the expected rank, from it, we can predict delta from our rank! That can reduce API access and improve speed

I tried to implement it but couldn't, it would be very helpful if any of you are able to implement the idea :)