Currently, the system is that each wrong submission leads to a fixed penalty of 50 points irrespective of the maximum points available for a problem. This means that for each wrong submission, one can lose as much as 20% of the points for a problem notionally rated 250 and only 2% of the points for a problem rated 2500. I am not sure if this fixed penalty system is a good system.
Perhaphs a proportional penalty system would be better where each wrong submission leads to a 4% penalty. This would mean that a 250-rated problem generates a 10 point penalty for a wrong submission and a 2500-rated problem generates a 100 point penalty.
Such a proportional penalty system would also be consistent with the current time penalty system which for most competitions works as 0.4% penalty per minute. Therefore a 4% penalty for a wrong submission will translate to a 10-minute time penalty, which is also what we use for ICPC style contests on codeforces.
If there is a way to do a poll, I could add that here. I am only gathering views and don't know if downvoting the blog entry is a disapproval for seeking views or expression the opinion that fixed penalty system is better.
Auto comment: topic has been updated by ag._. (previous revision, new revision, compare).
If all people are equally affected, then net no effect on ranking table.
But then the same logic applies for time penalty. The proposed system brings consistency between the two.
I think it's bad idea since getting accepted from a 800 rated problem is very easy with at most a few submissions. On the other hand, it's harder to get accepted a 2500 problem, and mostly more attempts are required. Getting a wa on a 2500 problem != gettin a wa on a 800.
I agree with what you are saying. Getting a WA on a 2500 problem is not the same. If you can solve a 2500-rated problem then the error on 250-rated problem is likely due to misreading the problem etc and should not be penalised with 20% cost. If you are a newbie, a 250-rated problem may be the only problem you solve and 20% penalty is again too harsh.
Auto comment: topic has been updated by ag._. (previous revision, new revision, compare).