Disclaimer:
I'm not suggesting this should be added to codeforces or any other competitive programming platform, I just wanted to speculate on what would happen.
I came up with this idea myself but it's likely that someone else might have thought of it already.
I just had a fun(?) idea for the rating system:
- There is a button that you can click to switch between being rated and unrated for a contest. You can press it an unlimited amount of time in the contest, but once it ends, you can no longer change it.
- You can't see whether someone is currently rated or unrated (honestly it might not matter that much since they can choose to switch later when you're not looking).
- When the contest ends, only people who choose to be rated by the end of the contest would have their rating recalculated.
What I think this would do:
- A contestant can choose to be rated or unrated depending on whether they feel happy with how they performed.
- For example, if you had some emergency to handle during the contest, then you can happily be unrated and not have to suffer because of something out of your control. On the other hand, even if you had something else to do, if you can comeback soon enough and are happy with the performance, you can still choose to be rated.
- So this would also encourage participation since unless your computer or network explode, you can choose to not be rated based on performance.
Why I think this wouldn't result in people choosing to be unrated all the time to only get positive delta:
- Each contestant won't know their rating delta in advance, since the standing doesn't tell them who would choose to be rated. There are some obvious cases where a person would choose to be un/rated (unless they are actively trying to lose rating), but for most contestant who performed their level, it's hard to tell.
- Sometimes, even if your performance is not good enough, it might still gain you rating. And vice versa, even if you are happy with your performance, you might lose rating anyway.
So what do you think about this system? If there is enough interest maybe we can try it out using a group standing or something.
Auto comment: topic has been updated by mir (previous revision, new revision, compare).
Rating predictors still exist.
Well, they can't predict much since they can't know which contestants are rated and which are not
Rating predictor won't be correct, because it won't have access to whether each person is rated or not.
Assume people use a predictor to make decision, then the decisions from the predictor would be an input of how the rating change would actually be, and that would invalidate the predictor's output.
For example, just assume everyone is rated and calculate a rating delta from that. Then let's say for each person with negative delta, they choose to not be rated, so only positive delta are rated against each others. So the actual delta change would be become negative for some of them. In fact, even if your original prediction is negative, you can still get positive delta if enough people who beat you become unrated.