I was curious which problems you need to solve consistently to maintain a rating. After observing a few contests, I ended up with the following list:
Here X & Y means if you solve from start to X you will usually lose points and if you solve start to Y you will usually gain a few points. This list is just intended as a rough approximation that ignores individual contest difficulty, ranking, international master/grandmasters etc.
That said, am I close in my evaluation?
For grandmaster you mean Div1 D right?
It seems that solving Div1 A-C consistently and quickly leads to grandmaster. For example, in Codeforces Round 641 (Div. 1) ranks 30+ solved A-C. Similarly, for Codeforces Round 647 (Div. 1) - Thanks, Algo Muse! ranks 70+ solved A-C and no other problem. But it seems speed is very important for ratings to improve.
Solving D would likely guarantee an increase to grandmaster but doesn't seem to be necessary.
I actually think it's more C than D. If you solve most of the times A-C rellatively fast, that will make you a grandmaster. D div1 can be often very hard.
I've only solved up to 1C on my best rounds (like +102 when I was already Master), and I don't remember my rating going down after solving up to 1C. So I'm pretty sure that if I can solve 1C 'consistently', I should be able to reach GM on some rounds with difficult problems. I think if you can solve 1D 'consistently', you should be able to reach IGM even.
isn't it most time that div 1 and div 2 different by 2 problem. your data suggest that it is different by 3 problem (especially obvious in candidate master section). how did you get this data.
That's because if you solve Div1 A & B (usually Div2 C & D) quickly, that's enough to give you a better rank and a better rating than solving till Div2 D only.
You can see this by looking at blue/expert coders who usually solve Div2 till D but are still not candidate masters.
In my experience, it seems to be the case that Div1 C is usually harder than Div2 E. It might just be psychological though, pressure for a CM is way more in a Div1 contest than in a Div2.
aren't they usually the exact same problem
If there are two rounds (e.g. CF Round #641) then yes, it is exactly the same problem.
But what I'm trying to say is that whenever there are not two rounds and it is only Div2 (e.g. CF Round #651), problem E tends to be easier than a regular Div1 C.
I think specialist should be Div2 A&B and pupil Div2 A and sometimes B. When I was a specialist, i solved just A&B, now, I can solve problem C, and sometimes D, and I am a low expert (1650). But it is still quite precise.
Don't you think situation is different now as the number of participant has increased due to corona?I feel like even solving both A and B does not guarantee your place as a pupil now,pardon me if I am wrong,it's just a personal thought
I think, to stay pupil you just need to solve Div2 A & B fast enough.
Being an expert means that your rating is from $$$1600$$$ to $$$1900$$$, Solving A, B, and C fast enough can give you a low-expert rank ($$$1600$$$ to ~$$$1650$$$), and solving A and B fast enough can give you even a high specialist. It depends on how many people solved each problem, and how easy they were. But still, nice job!
In my first and most recent contest I solved till Div2 C but admittedly C took time and I also had a couple wrong submissions. This gave me +513 which means my actual rating is 1400 + (513-500) => 1413 => low specialist.
Another example is el_heffeh who commented below and has solved 3 problems in the past few Div2s without the rating changing much.
After seeing your contest history it looks like you maintained specialist by solving 2 problems in Div2. I don't know what to make of that, but I think Div2 C is still closer since it's more independent of speed.
It is hard to classify in that way because of penalty. It may be easier to classify in the number of points achieved during the contest.
P.S.: I think I will make a blogpost on that using API data.
Great, would love to see that.
I feel that sometimes the map makes sense, but there can be inconsistencies. Although on most contests I am able to get A, B, and C, there are times (even when the A and B problems aren't even rated that high) when I cannot even get a single problem. However, I do agree with this table in general.
Good mapping, some time ago it would be actual, but now due to rating inflation, it isn't correct enough. Some people can become cyan/blue/violet solving consistently B/C/D respectively(I have some examples).
No offense, but it's good to see newbie getting many upvotes, like it's becoming rare situation
His real rating is 1413
Good Insight !