I have been noticing this for quite some time now, but I always just glossed over. Just wanted to share, so I'll make a short post about it.
I have seen quite a few problems that seem to have a very high difficulty(sometimes same as Div2C / Div2D) even though they are Div2A. For example: https://codeforces.me/problemset/problem/518/A
This is a pretty easy problem and I just can't fathom why it would be the same difficulty (1600) as for example: https://codeforces.me/problemset/problem/570/C or even a Div2D!!: https://codeforces.me/problemset/problem/735/D
I have seen this happen more often for older contests, but I may be mistaken. I figured, maybe I could find a reasoning if I read more about how problem difficulties are calculated: https://codeforces.me/blog/entry/62865 but from this it seems like that you have to be an expert to have a >= 50% chance at solving the mentioned Div2A problem. What??
It doesn't really bother me all that much, I just wanted to share and am wondering if anyone stumbled upon such "high difficulty" problems and if you have any idea why their difficulty is so high.
For the particular problem you mention, the reason of its rating being high is because only 795/3044 official participants solved it in contest, and although that was most probably because of an edge case which people missed and led to a lot of hacks/system test fails, for Codeforces Difficulty Evaluation, it's as good as 795/3044 people solving it in contest, which is comparable to a Div2C difficulty generally.
Thanks for the comment! :D That makes sense. It kinda sucks that easy problems can be evaluated with a high difficulty because of edge cases but I guess it can't be helped. There are a lot of problems on the site so it only makes sense that a few will have their difficulty skewed.
I think there are two main reason: Firstly it depends on the porpotion of the participants who got AC on the problem and secondly, there are just a whole lot of people who are more skillful than you. For example, you found that the Div2D problem is hard but actually, Goldbach's conjecture instantly kill it
I would say this is harder than average div2-a problem and hence the difficulty 1600. It's hard for organizers to always estimate the difficulty correctly. I would say this one was more like div2-b.