Should Long Challenges be combined rounds?

yes, I Think it is very good idea but make sure add one more rated contest.

1 Like

Agreed. Long challenges should be combined.

1 Like

Don’t need to combine. But don’t give the first 3rd-4th problem very easy problem and then 4th-5th problem suddenly very very hard problem in a long challenge…

So is this going to be implemented or nah?

In my experience, when alei launches a survey is because he is paving the way :grin:

4 Likes

Please implement this in next long

Hey @alei,
It’s a great idea to share problems between divisions, the problem you stated of Div. 1 users not being able to solve the first problem in Div.1 is because getting to Div. 1 is quite easy at the moment.
Making 2000 rating as the lower bound for Div. 1 or sharing the problems between divisions and making it more challenging to reach Div. 1 might help.

1 Like

This. 4* isn’t really div1. Its mostly just people who did good in the last contest. I believe having no divisions in long contest is the way to go forward.

I liked the concept of combining long contest, but i would also want to give one suggestion that like codeforces make some rule about trusted participants so that even if few people make fake id just for testing the solution, they shouldn’t be placed in final rankings.

@alei I don’t understand what you mean by this; Aren’t there already 5 problems shared between both divisions? Do you mean that 1/2 extra simpler problems get added to div1?

Challenging users on their own level

I definitely think there is a big advantage in the division system, the reason why the division system was introduced in the first place:

I think this is one of the features we need to try to guarantee in any system. Simply combining div1 and div2 for the long challenge would mean that higher-rank users have to waste time on trivial problems. Personally I like problem solving, but coding less so. So for me it would be really annoying to have to implement trivial solutions.

The division boundary

Let me start by saying that I have never personally experienced the division boundary, I crossed it when it didn’t yet exist. That being said, I think I have a theory why the division boundary represents such a spike in difficulty. I believe this is because the ELO rating system doesn’t fare well with the divisions. It’s essentially having two islands of chess players. Because players mainly play againts players of the same island it is easy to compare players of the same island. But comparing players of different islands becomes really difficult. This is also the case on Codechef. The islands represent div1 and div2 which only play against eachother. The only way of comparing the two is by looking at those that moved between them, which is only a small fraction of the population of either island. I think the result can be summarized in the following graph:


Note: this is only a sketch of what I think is happening

We see a bunching up at the boundary as ELO decides that some of div1 are not good enough and some of div2 are too good. But then in the next competition the ELO decides that those demoted are actually a lot better than others in div2; and those promoted are actually a lot worse than others in div1.

Resolving the issue

I don’t think abandoning the divisions is a good idea, at least not yet. I think the best way of resolving the issue is by building a bridge between the islands: let the different divisions compete. I suggest the following method to do so:

  • compare two competitors of the same division using the total score they got on the 8 tasks in their division.
  • compare two competitors of different divisions using the total score they got on the 5 shared tasks.

This method does however imply a major change in how the rating system is currently implemented; as that uses a modified ELO for use in ranked contests.

7 Likes

So why not just make the contest common and ranking made by divisions. It’s not hard implement as it would just mean make available all questions in div1 page as well as div2 page and at the same time making them compete in there own divisions.

Because then you don’t get challenged to do hard problems. To get 100 extra points you need to choose between solving an extra trivial problem or an extra really challenging problem. Everyone will choose to try to solve the trivial problem because it is the easiest way to get extra points. Meaning that to get a good score compared to other div1 competitors you need to solve the trivial exercises; thus wasting time you could have spent on challenging yourself. Furthermore the boundary hump exists because people are only compared within their own division.

1 Like

You misunderstood me, what i mean is to take score of all 11 problems not like choose any 8 question whichever someone like.
Anyway in long challange especially, problem made available only for div 2 are very easy to implement and so 5star and above won’t even need to spend half-hour to solve them. And so everyone will do their best.

4* should be in div 2. Long Challenge inflated my rating, but I cant’t even code some elementary algorithms.

1 Like

That’s definitely true for me, and i think it won’t be making any difference because whenever any1 changes division it’s difficult to solve questions, i have seen post and questions on codeforces and youtube when people asking for help when they reached div 1 of codeforces they were not doing good. So i think it’s normal for any1 who is progressing from beginner.

2 Likes

Not to demean, but I have friends who are 4 stars here, but newbie/lower pupil in codefoces.

This is true.
I have seen people having 5 star on Codechef but stuck on Specialist on Codeforces since long.
On the other hand, I have friends having 4 star here and been expert there since the beginning.

Hi karan

2 Likes

No offence but he is Pupil there. I have seen his profile.
But why does it matter? Making efforts to improve is all that matters. And he is sure doing that, unlike you hiding under an anonymous account.

1 Like

I think it should be combined as Global Long challenge. The problem should be set by 2-3 coders in group because that will create a more balanced contest.

1 Like