Hi there! The recent Good Bye 2019 was very important event for us, and thanks for all the positive feedback we got, especially from Radewoosh! As a follow-up, we would like to share some background on its preparation.

Problems:

A: This one was the hardest to come up with. We wanted it to be super easy but still to include some idea, and this is like the $10$th version of the problem.

B: Initially there was another problem at this position. However, it turned out to be too hard for B (one orange couldn’t do it during the testing), and people didn’t like it in general, so we came up with another task.

C: Funny, but we didn’t know that it’s always possible to make array good with adding at most $$$1$$$ element before the contest.

D: This problem was created just 3 days before the contest to make it more balanced :O. We are glad that the contest did turn out to be well-balanced.

E: Initially we wanted this problem to be at position C, but the coordinator told us it was too hard for it and we moved it to D. However, testing showed that we have to move it even further!

F: This problem caused a lot of troubles because we wanted to cut $$$n^2$$$ brute force solution which was working very fast. Luckily, nobody managed to pass it in $$$n^2$$$ during the contest, but in upsolving savinov managed to!

G: Before there was another problem at this position, but this problem by a recent round by hugopm turned out to be too similar to it :(, so we had to replace this position.

Initially, I wanted to send the current problem G to IMO (asking to prove that there exists a nonempty set with zero-sum), and I think it would make a great P2, but we wanted to make this contest the best we could make, so I decided to use it here. We are glad you also found it beautiful!

I think it may be interesting how this problem was created. I wanted to come up with some problem with permutations, and the first (and obvious) version was:

Prove that for any permutation $$$p$$$ of $$$n$$$ elements there exists some set $$$S$$$ such that

Of course, this is a very bad problem, as even $$$S = {1, 2, \dots, n}$$$ would suffice. So I decided that permutation wasn't necessary, and the next version of the problem was:

Prove that for any array $$$a$$$ of $$$n$$$ elements such that $$$1\le a_i \le n$$$ there exists some set $$$S$$$ such that

But this one was too obvious and too well known...

I just decided to regroup terms, so that the new version became:

Prove that for any array $$$a$$$ of $$$n$$$ elements such that $$$1\le a_i \le n$$$ there exists some set $$$S$$$ such that

It was left to just set $$$b_i = i - a_i$$$ and to ask to find subset with zero sum! Funny, that this simple process led to such problem O_O

Creating tests was really hard for this problem, you can read about one testcase here, but we are glad that not so many heuristics solutions managed to pass!

H: We needed some programming task (or otherwise some people would kill us). Initially we came up with the $$$q = 0$$$ version of that problem and considered it as Div2B — Div1A level, but then we thought about the queries and Mediocrity managed to find $$$n\log{n}$$$ solution!

We had problems with TL with this problem as well, wanting to cur $$$n\sqrt{n\log{n}}$$$ solutions, but didn’t manage too, unfortunately.

Fun fact: initially we considered G and H to be of about the same difficulty. The contest proved that we were very wrong :D

I: I loved this task a lot (this is my favorite problem in general) and wanted to create something with a similar idea (Idea: count some function for every cell such that some operation changes values of functions nicely).

Firstly I wanted to send it to IOI, but after I decided that it's really hard to come up with meaningful subtasks for this problem, so I decided to use it here.

I would like to point out, that problems E, G, I were created in direction solution -> problem. For some reason people keep saying that this way gives bad problems too often, and I want to persuade others that this way of creating problems is great :D

During the contest

We were really surprised when the first 2 accepted solutions of problem G came from people with ratings 1900 and 1700 :o

At the end of the contest, we thought that the decision to make it $$$3$$$ hours was wrong, as the number of attempts to pass some heuristics in G and some $$$nsqrt{nlog(n)}$$$ in H was increasing in last $$$30$$$ minutes :(.

Post-contest impressions

We expected people to complain about bad TL in F, H, about bad tests in G, and about the contest being too Mathforces in general (yes, that’s why I wrote Mathforces Thinkforces in the announcement. We were really surprised by the amount of positive feedback though, thank you a lot!

There was a very interesting post-contest discussion involving Radewoosh and Swistakk. We intended to make the problems heavily thinking oriented and less coding oriented, and we think that we succeeded. However, we would like people to share their opinions: do you think a lot of ad-hoc problems in one contest are bad for Codeforces?

Thank you for the amazing year though, we hope to return with other good contests!

Happy New Year!

(You will see my cyan hair in the next few days)