As you know because the Div 1 + 2 had relatively difficult questions the weak ones when reading problem A decide to not participate anymore and result in a decrease or rating for the moderate people (expert to master).
I want to suggest the system of participation become like atcoder which is even if you don't send any submissions you will be rated.
Megaplus
Why expert to master? What about low rated people?
Anyways, I support the change.
Agreed, and I also think that if A problem is a bit easier than today's contest, this issue will be much less
weak samples would also work
+1
MikeMirzayanov Please try to implement this feature on Codeforces(just like Atcoder)
I'm more interested in how the testing was done and influenced this last round. There's absolutely no way that testing showed that C was div2C difficulty. This actually extends into how testing is done, as most testers come into a phase where the problems have already been set and are in a hidden contest, so they usually do a virtual contest. This format isn't ideal for testing in my opinion because people are given the problem "order" and they have that bias going into the virtual contest. I also think that it's unlikely for people to upsolve the later problems as testers, and usually when it happens it's someone with higher rating.
In my experience, a common part of tester's feedback is "D is easier than C" or "to me, the difficulty order was "B<(A=C)<E<D", perhaps with reasoning.
But that's for people that try up until those problems right? That means that: since rarely a cyan tester goes that far, they give no opinion on that. On the other side, I feel that if a red says that it gets ignored.
Edit: copy pasting some feedback here from a round that I tested:
C: Didn't solve. Not sure why it being balanced helps.
D: Didn't solve. I thought it was approachable for a D and is pretty interesting
C:what
D:i think this is a really good problem,although i couldnt solve it.
that makes it seem that these people judged the problem not by themselves, but maybe by looking at the editorial...?
Hmm, but if people stumble e.g. on C, they usually still read D, and can solve it and say "wtf". Also, my understanding is that problems often get a minor shift (e.g. C<->D) between testing contest and final contest, so a tester has reason to expect that too.
Yeah, I stumbled on C. Gave up after an hour. Stumbled in D. Gave up after an hour. Stumbled on E, solved it but was too late :(. Personally I feel like C was weighted too low probably because the writers have god intuition and immediately see the construction. Therefore, they probably thought the solution was obvious. Also, dp problems tend to be very hit or miss, or at least for people like me who don’t really know dp. I feel like E should have been C, it is significantly easier than C and D and is a lot more accessible to people who don’t know advanced algorithms like dp or have god intuition like on C. For E, you could use a constructive algorithm to find where the end points of the two cities are and use a constructive algorithm to connect them since n and m are small. Perhaps also I am horrible at string problems that deal with lexicographical order and I haven’t done enough dp problems.
Even with bias, it should've been clear to anyone that tested that C wasn't fit to be C...
Basically there are two situations that a registered users is not rated in the current situation.
A: The user is not able to attend the contest due to oversleep or have some time conflict with other affairs. This user does not even read any of the contest problems.
B: This user read any of the contest problem, and decide to gave up the contest.
I think these two situations should be distinguished and users in situation B should be rated.
In Atcoder, you can change to rated/unrated before the contest starts or in the first 5 minutes of the contest, I think that works for these situations. https://atcoder.jp/posts/745
Many people register but at the time of the contest maybe they can't join.
For that case, I suggest a "pay to see model", where the round only becomes rated once the competitor views the problems (after clicking on a confirmation button).
But honestly, I think the rating difference due to this problem dissipates in the long run.
OR have a rated/unrated registration system like AtCoder
Div 1+2 == Div 30
Many people register early but may be because of some issues they can't participate in the contest. I think there should be a system that if anyone see at least one of the problems of any rated rounds then he/she should be rated for the rounds
What if you registered but forgot about the contest.
This guy shared a solution to this problem, he seems somewhat dumb though, so I would take everything with a grain of salt.