Often these days it feels like the problems are too hard for their positions but it's almost never the other way. I'm not saying these problems are fundamentally bad — it's good to confront them to improve our problem solving skills — but it's not very healthy in a contest's perspective.
Let me bring up some examples:
Example 1. Yesterday's contest, Codeforces Round 945 (Div. 2), had no official all solves. Most of Grandmasters, and even LGMs suffered from the difficulty of problem F. Less than 1% of the official participants were able to solve any of the harder half of problems of the contest.
Example 2. Codeforces Round 944 (Div. 4), had the hardest problem 1971H - ±1 that requires knowledge of 2-SAT. Only a very few official participants were able to solve it even though the topic was pretty well-known, just because the subject's level itself is too high. I think this is one of the harder topics even for the last problem of Div. 3, but it appeared on a Div. 4 contest.
Example 3. 1932G - Moving Platforms. I once created a very similar problem to this (guess what, the problem title is literally "teleporting platform") with constraints that do not require extgcd, and even so the problem was evaluated as hard as a typical 2F problem. This problem does require extgcd, but this was a Div. 3 problem with no official solves.
My suggestion is NOT that we should remove harder problems from contests. Instead, I want to say: Why not hold a higher division contest with these problems? Many of Div. $$$n$$$ contests just feel perfect to be held as Div. $$$n-1$$$ contests, especially for most of Div. 4 and Div. 3 contests. In recent contests, the hardest problems from Div. 4 are not easy even for Experts, and if you can all solve a Div. 3 contest you're most likely near Master level.
Here are a few points from me:
In a contest's perspective, problems with little to no official solves are 'wasted' because they cannot affect the official standings at all. If the writers prepared 6 problems, but only like 1% of the participants can really do anything about the later 3 problems, wouldn't these problems be more enjoyable for higher level unofficial participants if they were also rated, as well as making the competition more interesting?
For all we know, even if the hardest problems have a few solves, most of them are from alts, meaning that they're still too hard for actual legit official participants. For lower divisions, these solves often come from people whose rating is low just because they didn't participate in many CF contests.
Let's talk about a little history about Div. 2-only contests. Until a long time ago (around 6 years?), Div. 2-only contests were rated only up to Experts. Then, someone brought up that the harder problems in these contests were quite challenging even for CMs, and that's why we started to include CMs in Div. 2-only contests. But what we have in Div. 2-only contests nowadays are problems that are too hard even for Masters. I don't think this is what we expected when we decided to extend the rated range.
We shoudn't be afraid of having a too 'easy' contest once in a while. Even if an extreme case happens, 95% of the participants still won't even have a chance to tackle the last problem, so it's still pretty good for most people.
Because these contests are not rated for many high-rated people, they cannot fully enjoy the contests even when the problems are pretty interesting to them. I think the drought of Div. 1 contests, and comparably low frequency of Div. 2 contests to Div. 2-4 contests altogether, are asking for necessity for a more lenient difficulty threshold for the divisions.
Yes, I completely agree. As a pupil a Div 2 contest is hit or miss, I usually solve A, B and just completely give up, cause the C will be around 1600 rated and I can't really solve up to that level. So my rank will be entirely dependent on how fast I solve two relatively easy problem. And If I get a wrong submission on any of the two, my rank will plummet below people whom I could have overtaken If say the C was around 1400.
(I just stopped giving Div 2 altogether, because of this.)
Also I find myself a lot more involved in Div 3/4 cause it gives more problems that feels like it could be solved by me.
Hope this changes.
(yay I am a specialist now, no thanks to div 2)
IMO you should not stop giving div2’s altogether. It’s better to have a lower true rating than a higher fake rating. Plus, the only way to get better at solving div2C’s in contest is to attempt doing so.
Regardless of the high or low, I think rating after solving 4 or more tasks is more true than solving 2 or less. The latter tends to be more easily fluctuated by luck than the former.
Is it true for 4 800-rated tasks? In some instances, it's safe to say that div4 ABCD < div2 AB. And I'm not even talking about the fact, that the actual rating on most of div3 and div4 problems is less than what was given to them by cf rating system.
*800 *1000 *1300 *1600 *1900 *2100 is more true for div2 than *800 *1300 *1700 *2100 *2500 *3000. Div2 is rated for ~1899, and no 1899 can solve *3000(except for future GM).
Difficult div2 often feels like div1.5, which may be good for M or CM, but is disastrous for E. Creating difficult well-made task is very difficult, and I think it is impossible for the rated participants of div2 to solve *2500 or harder task. Why not save the most difficult task for div1? Wouldn't it be satisfactory for everyone if put in a moderately easy problem(yes, around *1000~*1500) instead?
Emm, I would argue that 800 1000 1300 1600 1900 2100 is never true for div2
I genuinely am sorry for your inability to solve harder problems, but i can ensure you most of >=1800-rated participants can pretty consistently solve 2100-rated problems and div2 is not meant to be AKable by 500+ people. I agree that 3000* is extreme for div2, however 2500+ is just fine for CM-M (don't forget div2 problems are also inflated ;)
The rating distribution over problems you proposed would suit a normal (not even hard) div3, but in no scenario a div2
Contests are not meant to please everyone with low difficulty, otherwise there is no point in contest. If you want moderately easy problems you will be able to solve, you can freely open archive and solve as many as you wish there.
Back few years ago, 800-1200-1600-2000-2400 would be a normal div2 problems rating distribution, the fact that you got used to recent contests being easier doesn't make it normal and common
Contest is stands for rated participants, I think. If not, then what I'm about to say is just.., yes. Kind of nonsense.
I'm not advocating that problemsetter must just lower the difficulty. I'm saying that it's necessary to moderate the difficulty for the people who will be solve them.
(Not Div.2 only.) Div.2, with Div.1 considers only those with a rating of 1899 or less to be rated participants. This means that very few people can solve *2500 or higher during the competition.
It was my mistake to mislead by using an extreme example and not explaining it. *800 *1100 *1400 *1700 *2100 *2500 like contest would be better then examples I gave before. However, I think need to consider what kind of rated partipants will be solving this task during the competition.
If a *2100 task is too standard to be in Div.2 or Div.1, why not put it in Educational CodeForces instead of Div.4? Why does Div.2 with Div.1, have to obey the formula D2F=D1D? I believe contest divisions could be improvable and give everyone more satisfaction.
I consider problemsetters shouldn't "moderate difficulty" for anyone. Point of contest is to be challenging and for you to get out of your comfort zone, otherwise you won't progress and won't learn anything.
Umm... do you really think existense of *2500 or harder task matter the comfort zone of 1600~1899? I think not. You should missed the point... The point is *2100 is not suitable for ~1399, *2600 is not suitable for ~1899. The comfort zone doesnt matter.
Adding extremely hard task that 99.99% of rated participants cannot solve is meaningless, because only chosen 1% or less try that task in contest time and finally that buried in archive. This is no good. This is all I want to say.
Or you just want to say *3100 D2F is OK. But I don't think so.
But point of contest is not to satisfy comfort zone of some subset of participants... If we're talking about that, there exist problems on A-D positions for 1600-1899 rated people. I can't say they're intended to solve more.
Also I had some 2600 being solved (alright, not in contest) while being in 1600-1899 range. Which means it gives chance of one in thousand to solve it.
Why do you even care then if you're not going to try it? There are enough problems in contest for you to solve even without the last one.
Where exactly did I say that? In my above message, I rather said directy "I agree that 3000* is extreme for div2, however 2500+ is just fine", please read better.
That's what exactly what I said. Even you, now are CM, cannot solved *2600 during the contest. That is what I afraid, very good task goes to archive, and waiting someone.
Contest, again, stands for rated participants, and solvability during the contest is important. Even me, given plenty of time and my target is in my speciality zone, can solve one of the hardest one in Petrozavodsk Camp. So I would ask you not to misrepresent my intentions: it would be a waste to have a task that no one can solve, and I don't want to solve easy problems just for the sake of it, either.
standings — example.
Do you claim all people who solved E1 are fake-CMs/Es and secretly have 2500 rating?
Your point is invalid.
Erm... That is Div.2 only, and participants are ~2099. You have to read again my comments.
And yes, also in this case, E2 is not suitable for this range of participants(note that I did not said that task is no use or quality is low or no good! Just good task missed for its seat.). Actually only 9 rated participants solved E2.
It would be better that range of rated participants are ~2399 for *2700 and *3100 last two tasks(or unlimited), but I save my words because I have no authority to adjust range of rated participants.
Plus) I can't get rid of the impression that you choose your words wierd. In your link, who solved E2 and F are not fake at all. They just got a low rating compared to their level. They will probably go up. That's it.
You're just making excuses to justify your weakness. Div2 is not meant to be easy for experts. Once you get to Master and still feel that div2 D/E is too hard for you, feel free to complain. On your current level it's not intended for you to be able to solve more. If you want a contest where you'll be able to solve many problems, go on, do div3...
I will kindly remind you, that in div1, for considerable majority of it's "rated participants" it's normal to solve just 2 problems. Noone complains and asks to make div1 1300*-1700*-1900*-2100*-2500*. Remember, in div2 you're not different...
As for problems noone rated can solve, please show me any such recent rounds apart from round 945. The fact that it's a single exception doesn't prove your point.
Did I say any of them is fake? I meant higher-rated (in reality) alts doing div2s which can justify solving hard problems (like E2 and F), but please learn to read. In my comment I attached E1 as example, which is 2500 rated (the one you were insisting in your comment that is impossible for div2 people to solve). Do you claim that
are higher-rated alts? (look, some of them definitely are, but not all 80)
If you're not saying they're all alts, your point is wrong since those are experts and cms who successfully solved a 2500-rated problem in contest.
broski's spitting facts damn
the definition of rating X is that on an average you can solve problems with rating X. It thus does not make sense that someone rated 1.8k can solve problems with a rating of 2.1k consistently
You're missinterpretting it. The fact that it "is intended to have person of rating X solve X" doesn't mean that person of rating < X can't solve X.
I can solve 2400-rated tasks in contests pretty frequently, which, sadly, doesn't make me 2400-rated. As well, I was being able to solve ~2100-2200 rated problems, while, okay, being not 1800, but 1900 rated, if you care that much.
https://codeforces.net/blog/entry/62865
Also, I'm sure the reason your rating isn't something like 2200 now is because you lost 288 rating by being skipped and you kept only the deltas for the contests afterwards. Your performance for 4 rounds after that were like 2155-2419 according to Carrot.
that's what i said, the fact that it is intended to be so, doesn't exactly mean it is so...
and i wouldn't've been 2400 rated without -288 either.
...Like 2 years ago
The key word here is consistently.
What's 'consistently'? If you mean solving them like 20% of the time is consistent then I have no more words. In the recent 5 of the rounds you participated in, there were 4 2400-rated problems and you solved only 1 of them, and 1 out of 2 2300-rated problems. This isn't consistent at all and that is why your performance averages at a little above 2200 and not 2400. Same goes for 1800-rated participants and if they can consistently solve 2100-rated problems then they should get to 2100+ in a few rounds for sure.
I agree that there are some inflation for the top problems of each division, but not as much as you stated. The problem ratings, at least for low-mid problems, are working as nearly as intended.
Consistently, no, but on a good day, yes. I'm roughly 1900, and I've solved 2300 in contest before. Sometimes you'll do really well in a contest and end up with a lot of time, or said 2100-rated problem will happen to be about a topic that you're strong at. Chances are this situation will happen for a significant number of people in a div2 contest.
I do agree that problems that are 2700+ should not show up on div2, though.
I agree with this view. My personal experience: I'm 1700 and I've solved 2500 (two, in fact) in contest before. But both of them were very lucky solves. I still think though that problems which only, say < 10 true (as in, not alts, and not newly registered accounts, actual accounts, with actual average skill around the level of div 2) rated participants are going to end up solving aren't really fit for a div 2 contest. For example, in TimDee and prvocislo's Div 2, the first five problems were fantastic, and really nice. But the last problem, while interesting, didn't really do anything to distinguish participants, because only like 2 rated participants solved it, which means it could have been entirely removed from the problem set without affecting the results (this is different for div 1 where the difference in performance rating between first and second place is a lot, and there is more volatility as well, where say, some IGM can be ahead of tourist in a very specific situation).
you cheated in some contest anyway... So I downvoted(even not reading this)
We'll talk when you reach master
why?
lol youre just making it speedforces
I think probably the problem setters want these rounds to be “educative”. For example, I participated all the 3 rounds you mentioned in your post, and I learned a lot from them like 2-sat and using exgcd to solve Diophantine equations.
Perhaps different people have different standards to measure if one problem or round is “interesting”. I agree with that rating plays an important role in CP, but IMO seeing much harder things is also a must since it stimulates us to keep learning and practicing.
And also, beside these hardest problems in the rounds, the difficulty of other problems seems normal to the participants from respective divisions. So they can enjoy the contests while learning new stuff.
Yes, but the thing is that, you can still participate and try the same problems in a higher division. Every Div. 4 and Div. 3 participants can still participate if these 2-SAT and exgcd problems were in Div. 3 and Div. 2, respectively. It's just that more higher-rated people can also enjoy it.
Although I've mainly mentioned the hardest problems, it's also becoming more and more common that the mid-difficulty problems get less solves as the time goes on. This is why I also stated that it's not happy to see that 99% of the participants can only solve at most half of the problems.
as a not-so-high-rated div1 user, i cant seem to agree. Basically, as in any high-level competition, standard problems are not recommended: they aren't gonna distinguish anything other than whether you know the algorithm and its standard applications well enough. This isn't algorithmic thinking.
I agree with your opinion on div4 and div3, but cannot agree on div2. as some famous lgm said, "there shouldn't be standard problems in div1s." div2s problems are usually standard to some extent, so maybe they won't be a good fit for a div1 contest. Of course, exceptions can be made, but i wanna hear the community's opinion on this too:)
I suppose these problems for unofficial participants, to make rounds more interesting for them
Deleted
This has been talked about before and the general consensus seems to be that while these last problems of Div. 2 contests may be hard, they are also too standard to be suitable for a Div. 1 contest.
Some may be standard for IGMs and LGMs, but I don't think it holds true for most Masters and low reds. Many, not all, Div 2 rounds are definitely suitable for a "Div 1.5" and the coordinator should be able to decide this.
What's the case in a div 1 + div 2 contest? Are the C-F in div 2 (usually A-D in div 1) less standard than the C-F in a normal div 2 contest?
In general, yes.
Seems like you have a good (maybe quantitative) way of knowing the "standardness" of problems. Please do share.
For me personally, the latest div3 tasks were very useful when I was a specialist/expert. And I think that only by solving them, I was able to quickly improve from 1500 to 1900 (you can look at the graph and make sure).
Therefore, I think that it is necessary to look for such tasks where it will be possible to pump your skills quickly with a minimum of effort. And one of these tasks for beginners is div3/div4, just get out of your comfort zone.
the current format is more educational than competitive which i like for lower divisions and when u get to div 1 where competition starts then its fair for everybody i think divisions which target lower rated participants should focus more on the educational side than the competition side
I think it is a good idea for "standard" or "template" problems rated violet/yellow to appear in the last section in a Div.4/3. Textbook problems which you can solve after learning the basics of a topic is something I think is missing on Codeforces. You can't place them in a competitive Div.2. They can serve as a bonus for some unofficial participants who are bored.
Nevertheless, the gap between 2-SAT and XOUR/Circle problems are a bit to wide.
Sorry for repeating this for the third time, but I think I have not received any solid objections yet. Why can't we make Div. 2 partially rated for Div. 1 contestants? For example, it might mean that, if you are not in Div. 2, your rating change is lower: specifically, if you are rated $$$2100+d$$$, your rating change is multiplied by $$$e^{-d/400}$$$.
About some contests I've coordinated:
Of course, this can be generalized to Div. 3 and Div. 4.
Does this mean that if tourist participated in the round and got rank 1, he would only get something like +1, and even if he did really badly (like getting last place), he would at most be rewarded with -6? Then what's the point in it being partially rated for LGMs :cat_think:
Just too complicated
Movable rated boundaries make much more sense than this partial rated thing. But the issue with movable rated boundaries is its very subjective.
I think its a great idea (movable rating boundaries). Considering how many testers the average round has, I don't think it would be too subjective. Also, some other platforms already do this (for example DM::OJ).
Misplacing problems by difficulty happens. You can't hold a higher division contest with them because they already appeared in contests, because mistakes were made in estimating their difficulty. Contest divisions are fine as is, problem assignment to divisions is the topic.
I suppose what you mean is "can we assume problems are harder than they appear in testing?"
It was well known in testing that the 2SAT last problem of div4 is immensely difficult to implement (from scratch) and 2SAT as a topic is quite high level
That doesnt make it fair to put it in div3 (or even in div4, but well it was)
I'm a Newbie and I still enjoy these contests whether it be hard or moderate. I enjoy when I solve it and enjoy knowing how it can be solved when I couldn't solve it. The one thing which I know is, I will be able to solve most kind of problems if I keep learning and solving good problems. Atleast I know I'm growing thanks to the codeforces!
Many E/F problems in Div2 are classical,so in many people's eyes,they are easy.But for some participants,they haven't seen these tricks before,so they felt that difficult.
I would say these problem has taught me many topics like 2-sat, DP with bitmasking. I would say these problems are more educational than competitive which motivate the lower rating participants to learn new topics . And, since the topic has come in recent contest, there are many elegant solutions which comes on youtube which helps us in learning these topics easily...