I wanted to bring up an issue I’ve noticed regarding rating increments after cheaters are disqualified. Recently, my rank improved from 935 to 821 after some cheaters got caught, but my rating didn’t change at all. This doesn’t seem fair, as my performance was directly affected by the disqualification of those cheaters.
Additionally, in a previous contest, my rank jumped from around 450 to 231, and despite that significant improvement, my rating only increased by +1, which felt disheartening.
I believe it would be fairer if ratings were recalculated when cheaters are removed, so that our actual performance is reflected accurately. Has anyone else faced this? Could this be something CodeChef considers improving?
We recalculate ratings after removing cheaters. But there are a lot of factors which will determine whether your rating will increase after recomputation or not.
Presence of cheaters do skew ratings for people, but for some it affects negatively and for others it can affect positively as well.
If the person with rating x performed better than what the skewed rankings (due to cheaters) reflected, their rating might increase after the recomputation. This is because their performance is now being compared against a more accurate representation of the field.
Conversely, if the person benefited from the presence of cheaters (e.g., if their ranking was artificially inflated), their rating might decrease after the recomputation.
I remember that in Codeforces, the ratings are updated the sooner its system testing is complete and plag check as well. Then, many times, the ratings’ delta were scaled to some change.
I think that this fact is concerning the contenders here for ratings’ delta.
But, here, in Codechef, that phenomenon is not observed. It may be due to a non-comprehensive Plag Check or fewer cheaters being caught.
I hope our community will help you to tackle this issue. Please let us know how to help.
Just in case if you can provide data next time. I have a request for you - let me know the top 500 rankers’ delta changes before the plag check and after it. (because I haven’t seen a great change among these, and I will better understand the actual issue & help others to…)