We know that currently rating change is based on Elo, so when there are lots of cheaters, normal participants would lose points. My new proposal is this:

let each tester rate each problem (or a subset, based on the rating of the tester), take average. Then after competition, each participant has their rating change calculated based on the problem's rating. The cheaters would still be there but at least people can just compare with own historical rating.

Thoughts?