Hi!
The main objective of my post is to make most Div. 2 only rounds rated for purle users.
This is beneficial because it would distinguish their rating better. It would also mean a more frequent participation for those people, which is good for the growth of the plataform. It makes no sense to make a purple coder wait for a Div. 1 round, which is rare due to the difficulty to come up with a Div1 D/E problem, when in most cases those users don't even read these harder problems at all.
Another reason is because most Div. 2 rounds are hard enough for people with purple rating. I assume the difficulty of contests should be sufficient to not transform them in a "type race", which means that it is good that few people solve all the problems. I will try to convince you that we can make div2 contests rated for people with higher rating without breaking this principle.
EDIT: Originally the title of this blog was "Please increase div2 upper rating limit" but this was missleading as I only want to allow div. 2 participation from purple guys in Div. 2 only contests. By Div. 2 only I mean the contests that do not have a corresponding Div. 1 rated problem set.
The distribution of participants in div. 2 and div. 1 (when there is a corresponding Div. 1 contest) could and should remain the same.
Some data to back my suggestion:
In Codeforces Round #476 (Div. 2 only) [Thanks, Telegram!] only 1 person solved all the problems officially, and I would say that more than 95% of all the unofficial purple contestants solved at most 4 out of the 5 problems.
In Codeforces Round #477 (rated, Div. 1, based on VK Cup 2018 Round 3), let's assume that the contestants who solved 4 problems would have a chance of solving all problems in the corresponding div. 2 round. Less than 50 (including red and yellow rated people) out of more than 450 div1 contestants solved at least 4 problems.
In Educational Codeforces Round 42 (Rated for Div. 2) only one red rated person solved all the problems and the vast majority of purple rated unofficial contestants could not solve 2 of the contests' problems. Also 3 legendary grand masters contestants could not solve all the problems.
With that being said I suggest that most div. 2 only contests should allow rated participation of users with rating up to 2200, or less depending on the difficulty of the specific round. But please do allow participation of more users in Div. 2 rounds when the difficulty is sufficient, this is healthy for the plataform!
I am open to discussing ideias in this regard.
Thanks for reading :)
I agree, in my opinion the Div 1. lower limit should stay the same and people with rating ove 1900 should be forced to participate in Div 1. Rounds. But when there are Div 2. Only rounds purple people should be rated or at least some part of purple. Maybe even split purple in two subcolours?
wouldn't splitting purple to two sub-colors mess the whole rating up meaning if <= 2050 is rated for div 2 when one of them wins the round it would give him a push to >= 2200 which he would most most probably not get participating in a div 1 round.
Me engravida!
I agree with your point that div 2 problems, sometimes are hard and it would be reasonable to bring Candidate Masters into div 2. But, an important factor here is that whenever a div 1 contest is held, around 800 people register and out of which only around 500 — 600 people participate in contest, and out of which around 60 — 70 % are Candidate masters. So, if we remove them from div 1, then there would be hardly around 200 people who would participate in div 1. So, it would be very unreasonable on part of the authors to make 3 — 4 new problems only for 200 participants. So, I guess that's why it would not be possible to bring Candidate masters into div 2.
When there is a Div. 1 contest along with a Div. 2, the current participation distribution could reamin the same. But the key reason to allow rated participation for purple contestants in div. 2 only rounds is because div. 1 contests are way more infrequent.
It makes no sense to make a purple guy wait for a div. 1 contest, which are rare due to the difficulty to come up with a Div1D/E problem, when some purple users (including me) dont even read those harder problems at all.
This is basically whats described in fbosnjak_ comment, not what your blog says.
I believe that only Educational Rounds could be the exception, rated to candidate masters. They are a bit longer than div.2 usual contests and usually have a Div.1 D level problem.
Maybe I could be more precise in my words by saying that the Div. 2 upper limit sould be increased only in div. 2 only contests. Other possibility is always making the contest a div. 1 + 2 contest instead of making two separated contests. Or keep the Division as it is but allow more people to participate, calling Div. 1.5, but this is a matter of "notation" for me. The important thing is the pratical implication that my suggestion has.
Regarding your opinion that only Educational Rounds should be rated for candidate masters I disagree, as shown in the first and second examples of the rounds mentioned in my blog.
I assume their difficulty should be sufficient to not transform the contest in a "type race"
Impossible. Even now, some kind of type race is presented on div2 contests: participants who solve 3-4 problems very fast are on the higher places.
But this is a current problem that would not be significantly augmented by increasing the div2 upper rating limit. Of course speed is very important, but the fact that there is a harder unsolved problem means that there is a chance for you to perform well even though not being fast or precise.
Wouldn't that make reaching Div1 harder because a lot of the positive rating change will go to participants from Div1 already?
Well it should be more important to distingish better the purple rated users with rounds that are hard enough for them. Again this is a matter of notation, if making Div. 1 harder to reach is a concern becuase it will discouraje people, decrease the Div. 1 lower limit and create a DIv. 0 or something.
I don't know if this makes sense — Div 2 contests already have about 5-8 times as many participants as Div 1 contests. This will skew it even further.
Please read this comment above
well i don't feel like a purple coder needs to solve problems like div2A and div2B before finally moving to a problem of his standard and if there is not a div1 contest i feel like its not that bad to participate in an unrated contest
Well I dont see why having to solve problems that you consider very easy is that big of a deal. If they are indeed easy you should get them fast. Otherwhise you get the opportunitty to learn something or improve your speed on easier problems.
Even if they are easy they take atleast 20 minutes to code generally, infact this time can be added to think more on d or e problems because after certain point of time you don't want to increase your speed but increase your thinking ability for difficult problems.
What about creating a new division? Div3, Div2, Div1.
Div3 problem set: A, B, C, D, E.
Div2 problem set: B, C, D, E, F.
Div1 problem set: C, D, E, F, G.
This is a valid suggestion, but I would like to stress that my post is more concerned towards the Div. 2 only rounds and making them rated for purple guys. In your suggestion, the Div. 2 only round could be transformed to div. 2 + 3 maybe.
This would create more reasons for Div1 contests to be even more infrequent than it is now because preparing Div3 only or Div3 and Div2 contests is easier.
Is it a guess or a fact?)
I know about a different dynamic happening among the users. People stop participating once they reach div1. Probably these people would have stayed if they knew that there is one more level.
Who knows what will happen when the new division appears. Maybe the amount of active users will increase twofold.
It's of course a wild guess and may be wrong. But I'm not talking about participating. I'm talking about preparing. I mean problem setters would find it easier to make Div3 + Div2 contests than Div3 + Div2 + Div1 contests.
We don't know whether complexity/easyness is the real reason for perceived rarity of div1 contests :)
People leave the website. If you decrease the pool of participants, probably, you are decreasing the pool of potential authors.
I agree. It may, as you mentioned, be some other factors that I didn't consider.
Preparing Div. 2 or Div. 2 + 3 are indeed easier, but why would that configuration make div. 1 rounds more infrequent? Increasing the participation in Div. 2 rounds shouldn't have major effects (if any at all) on the frequency of harder contests that require harder problems to exist (and nothing else).
By the way, we don't even know what are the variables that affect the frequency of div1 contests. Maybe, it's about the money — if we increase this amount by 2x, we may get 5x amount of contests. Or maybe it's about prestige — authors of div1 contests may want to feel special compared to those who make div2 contests and a simple badge or icon or blog post of different colour may lead to something.
But we don't know that. We don't even know if administration even consider that as a problem. Maybe there is some reason why they do that on purpose by for example filtering contest proposals.
I guess not many people could have predicted, that only 4 days after egor.okhterov posted this comment, it will actually pretty much come true...: "As an experiment, we plan to host rounds of simplified complexity (Div. 3)"
Oh, and if we have div 3, say bye bye to div 2 contests...
If there only have div2,purple master can rated. otherwise,purple master only rated at div1.
And purple master with about 2000 rating wins the contest and with the current rating formula becomes master. In Div.1 contest it's harder as you must place top 30/40 to get an increase of 200 (the needed increas to get master if you have 2000 rating). Do you think winning Div.2 is the same as getting top 30/40 in Div.1? In my opinion the latter is much harder.
I believe a solution, that is applicable even In the current division distribution, is to saturate the rating to the upper limit of the division. That is, no one can get a rating increase in a div. 2 only contest so big his/her rating surpasses a user that was already in div. 1 before that contest. Of course this brings other issues as things are not this simple, but I believe something in this regard is worth trying.
I also do believe that increasing the participation of purple users (without loss os quality of problems for those users) is more important than the concern you mentioned. More participation is more important than a few users with inflated rating.
Yes,I agree.So if your rating 2000+,you can join the div1 contest.
“I estimate pretty much 60-70% of the time all the problems are easily solvable by any div1 fellow”
Your estimation is provably very wrong. Have you seen the examples in my post? I didnt look for them, I just took the most recent rounds. But you can check any div. 2 only contest in the past year and look for the unofficial performance of purple rated people there. After doing this research I am sure you will change your mind.
But that's the point of a contest. I believe the last problem shld be solvable by at most 10 ppl, if not it will be a contest of speed. If you increase the limit to 2100, I believe there will be a lot more people with full problem, and that will make getting to 2100 a contest of speed, which I don't think it is what people would prefer.
Also, your data for div2 contest(only) is not a great indicator, most div1 contestant don't take unrated seriously. For instance, sometimes when I take part, and I know how to solve the problem but the implementation is too messy, I will just skip it. Or I will rage quit after a few pretest fail, or due to lagness. Or maybe sometime I only have an hour to do it, I will just do it anyway since it's unrated. And I know red coders that does the same, so I think the data is not quite accurate IMO.
To add on further:
Also 3 legendary grand masters contestants could not solve all the problems.
2 of the LGMs didn't submit anything pass 35 minutes, and the last one was hacked. So your point here looks quite invalid.
About the not taking unrated seriously — I also think so and don't take it seriously myself. Also I was purple not long ago. This is just to give some more credibility to your statement.
Ok, my estimate may be off a little. After some looking I see about 30-40%.
There are still a couple of issues though.
In your own example, round 476, you said "I would say that more than 95% of all the unofficial purple contestants solved at most 4 out of the 5 problems." So it's still a type race?
And also, rating inflation is a massive deal. Every time I participate in a combined-division contest, even if I do terribly, I (and other high-level players) get boosted by the div2 users. And if your suggestion happens, then it will be even worse, because there won't be any reds or yellows to keep the purples in check, so then some purple fellow whose skill level is 2000 can easily boost him/herself to 2200+ without any real improvements. If there is no accurate rating, then what is the point of having a rating in the first place?
I do like the idea of capping the rating gain for div2 at the boundary for div1, as in you can only get up to 1900 rating if you are in div2. That's a nice suggestion that's sure to kill the "I won a div2 contest and got to 2190" syndrome.
"In Educational Codeforces Round 42 the vast majority of purple rated unofficial contestants could not solve 2 of the contests' problems."
Some Div 1 contestants skip easier problems in educational round and only focus on harder problem (usually from the fifth problem).
I'm not purple, but I know many purple guys drop back to blue/go back and forth, and many purple guys become "stuck" at purple.
Pushing up rating boundary only delays the issue.
Why not allow purple users to participate in either div2 and div1 contest since they are one of the larger groups of codeforces but get so few contest because of less div1 rounds.
UPD: Seems like my suggestion become a reality.
I think that it is an interesting topic, so I investigated about the percentage of Div.1 participants and active users, and the mental reason for people who want to be Div.2 by some information of the following graph.
[The number of active users]
[The percentage of Div.1 participants for each contest platform]
My Opinion
I changed the blog to be more clear about my suggestion. I do agree with you that "The distribution of participants in div. 2 and div. 1 (when there is a corresponding Div. 1 contest) could and should remain the same."
What I want to address is that most div. 2 only contests should be rated for purple users also. Much like Div. 1.5 rounds.
Auto comment: topic has been updated by VLamarca (previous revision, new revision, compare).
I am wondering if unofficial participation is more fun for purples. Let's say you have a standard 5 problem div 2 contest. If you are purple A, B and C should be fairly straightforward. If you compete unofficially you can ignore A, B and C and challenge yourself to solve D and E in the two hours. If you were rated doing this would be a bad idea since you would likely lose rating points.
Bad idea. It's obvious that this will create rating inflation. It will be something like easily jumping from 1900 directly to 2200, since purple users will be competing with the same people they were before. If you want to solve the problems, you can do it unofficially.
I think this CHANGE will encourage coder join Div.1 to change their color and adapt Div.1 contest.
I hope Codeforces will work more great!
The main objective of my comment is to make most Div. 2 only rounds rated for rating users upto International Grandmaster
Some data to back my suggestion:
In Codeforces Round #480 (Div. 2 only) no user which consists of users from [Master to International Grandmaster] could solve 6th problem.
It is just because the other problems were rather difficult and participants had no time to solve it
THANKS FOR SUGGESTION BROOOOOOOO
INFLATION YO GO HIGH HIGH HIGH!