We will not hold AtCoder Regular Contests for a while. Probably the next ARC will be in March. Instead, there will be several sponsored tournaments.
The onsite rounds of these tournaments are only for Japanese (we are sorry for that), but we will try to translate majority of qualification rounds into English. In case the problems are translated, usually the contest is rated for <2800 (there can be exceptions, so please check individual announcements). The difficulty/quality of those rounds are at the same level as usual ARCs — so please participate in those contests! There will be one such contest this Saturday (20:00 JST, an hour earlier than usual), and I think it will be announced soon.
There will be two more AGCs this year. date/time of AGC029 date/time of AGC030
After these AGCs, we will announce the top 8 people (by GP30 scores), and they will be invited to World Tour Finals in February 21st!
UPD: We decided to move one AGC to the end of year because it overlapped with NEERC. In next two weeks there will be ABCs, an AGC on 15th, a contest rated for <2800 on 22nd, and an AGC again on 29th.
I noticed AtCoder has a theme of stagnation going. Something is announced and it's a big popular thing, but it eventually just fades. For example, someone here expressed that it's very hard to increase your rating past a small number of contests and it kinda fits. There has been no non-trivial (non-ABC) contest for a month. With these 2 AGCs, it makes less than one per month this year — that's way less than just the number of mirrored/sponsored div1 rounds on CF and seems like way too little to choose 8 people for WTF from (even one missed round matters a lot). You have custom colours of handles, but at this pace it might take years for some people to get them. You had international onsite competitions, but not anymore. Add to it that there's no real way to build a community there...
About the rating: that's the truth. Unless you are a beginner, you need a long-term effort to improve your skill. I don't want to change the ratings too much based on small number of contests (then the ratings will just represent your luck in recent contests).
About the community: I think the idea of CF is to gather all discussions related to competitive programming into a single place. I don't think it's a good idea to split it into CF and AtCoder.
About the number of contests: My intention is that the quality of ARCs are comparable with other regular contests and AGCs are something above them. Holding AGCs twice a month is unrealistic (and if I counted correctly, the number of AGCs this year will be exactly 12).
And you won't do that on AtCoder, given the low number of contests.
Edit: also, by the same logic, since you need long-term non-effort to decrease your skill, it should take a lot of contests to decrease in rating — but that's not what I observed. It's just easy to drop due to bad luck, but hard to get back up.
cgy4ever won his first contest and got rating 2928. simonlindholm won his 3rd contest and got rating 3201 (9th and 54th place in previous contests). How is that not getting very high rating due to being lucky early on?
ARC are definitely less difficult than CF div1 rounds. I can realistically expect to solve everything in an ARC contest and a lot more people can do that too. They're more like difficult div2 contests. Our tiny Slovak ACM training site holds contests with comparable difficulty (around 10 per year).
I'm not saying there needs to be a community around Atcoder. It's just that when you add everything together, there isn't much left. I would like to recommend the site to someone, but how do I do that? "Well, they did a lot of things before" isn't exactly compelling.
Speaking of AtCoder community, the biggest one is on Twitter, and most of the discussion is going on there. Few good coders from Japan write about contests in Codeforces (, where it is read by few Japanese as well).
It seems the way to partially resolve the issues mentioned by Xellos while maintaining stability would be to amend the APerf calculation. Currently, you use 0.9i, where i is the rank of your contest, as the weight of a contest. Have you considered using, e.g. 0.9t where t is the time in months since the contest? Imagine the following scenario:
In the current rating system, the time T does not influence the final rating, giving you rating 2076 (after subtracting f(2)). In my proposal, larger T results in larger final rating. The rationale is that, if two years have passed, chances are that you really improved significantly during that period, and system should rate you accordingly. If the above formula would be used, the rating will be 2395. If only two weeks have passed, then you likely just have a very large variance, and the rating system should be very cautious about you.
I miss ABC :(
I like AtCoder
I was wondering, what typically prevents you from organizing more contests? Is it the lack of problem setters? The time needed to coordinate the contests?
If it is about lack of problem setters, have you considered working with problem setters who don't speak Japanese? I think many people truly appreciate AtCoder as a platform (two major reasons being stability and full feedback on submissions) and would be willing to switch to holding contests there.
Both the lack of problem setters and the time needed to coordinate the contests are reasons. And maybe the third reason: sometimes it's hard to find a time slot that doesn't overlap with other contests.
Internally, we are already working on improving the situation. For example, now the system to receive problem proposals from many writers is ready. Please wait for a while until I finish reviewing the problems in the current queue — then I'll start receiving problems from more writers.
I agree I would love to have more AGCs, however one thing I would like to add to considerations is that AtCoder holds contests of highest quality only. I think it is pretty clear that average problem from AGC is much more interesting than for example an average problem from Codeforces Div1. I understand that it may be harder to provide problems of highest quality only. That of course doesn't have to mean that AtCoder needs to accept setters from Japan only, but not anybody who can prepare okayish round should be allowed to do so (so it's not like I am opposing your line of argument, just adding something to discussion).
I completely agree! The superior quality of problems is the biggest part of what makes AtCoder great for contestants. I imagine that maintaining this could mean having an invite-only system for contribution (e.g. you have to set a high-quality contest on an other platform before you are allowed to submit problems).
Now , I think AtCoder has best quality problems in contests more than other websites...but the problem that there's a big gap between every two contests...I know it's hard to have large number of contests with this high quality but I mean if there's a small gap between contests...It will be a great thing
Atcoder ARCs and AGCs problems have some of the best tasks that i love solving hope the tasks of these tournaments are of the same quality.
Auto comment: topic has been updated by rng_58 (previous revision, new revision, compare).