I see, thank you for clearing my misunderstanding!

SuperJ6 doesn't seem to have read my question yet so can someone (hopefully chinese) answer my question please? I'm curious to know and I've waited 3+ days hoping for a reply...

OI referers to Informatics Olympiads in general

https://contest.yandex.com/ioi/contests/

I think oj.uz is a tad better than yandex. Just opinion though, and there are a couple of output onlys that only yandex has iirc

I think by OI problems he meant problems with subtasks (that have points). It's easier to solve a hard OI problem, because you can try to solve it part by part. It's also giving you a stress solution most of the time. In an ICPC style task you are aiming for the full solution from the very start, which is obviously harder.

There are too many ICPC-style competition and each of them has too many problems. I feel it’s hard to filter what is good or not. And they are less consistent (ex: CERC 2017 and 2018 have radically different quality)

Also something really good about OI is that they generally foster long-term thinking, so if you have ICPC problem and OI problems of same quality, it’s better to encourage them doing the latter because they help them to solve harder things in their life. ICPC problems are generally not posed for this purpose (although Snark’s New Year Contest changed the culture a bit)

For techniques, I think “exposed to techniques” are good and we should learn more about other’s approaches. But for OI they are not expected to “use” the technique but “invent” such kind of them. Anyway, yes, to be exposed at various stuff, solve from various sources.

Don't get me wrong I think you should do all types of contest as that exposes you to the most things, but I agree with ko_osaga that I find it much harder to filter out good problems from bad in ICPC, because they have to produce more problems and more end up not great quality (of course doing the more prestigious regions and wf kind of fixes that), but mainly it seemed like the op had only been doing codeforces and oi is a much more drastic change from what he'd been doing, so I think he should try more of those.

Also I very much disagree oi will limit what you learn. The only big thing oi doesn't really have that others do is flow. I wish they had more (any?) flow, but for example I've seen more hard string problems, not just constructive stuff using strings, in oi than cf.

Code

int a,b,c,d,i,k,n,m,e,dx[10]={1,0,-1,0,1,1,-1,-1},dy[10]={0,1,0,-1,1,-1,1,-1};

Sure, why not, let's start our code with declaring 10 one-letter global variables.

  scanf("%d",&e);
  for(int tt=1;tt<=e;tt++)
  {
    scanf("%d",&a);
    k=a;
    for(int t=1;t<=a;t++)
    scanf("%d",&o[t]);
    for(int t=1;t<=a;j[t]=0,t++)
      scanf("%d",&l[t]);

And use them later in a way that doesn't reflect any names from the statement.

It is that bad and I really think it will make a difference

How many of them were without looking at the editorial or taking any hint?

What task list are you talking about though?

You're just trying to come up with excuses for yourself instead of admitting you suck

Wow, even mentioning my name leads to downvoting.

You can change one of those right now though

its nice but it stuck even more than 6 years :(

Wow, you was stuck long... How did you escape?

If CF is your main source of practice, I think you need to solve this many problems to become stable red. But this number varies from person to person.

If I know it, I will never write such like this post. I think a good platform, it can give you the difficulty of problems, filter which problem is good, see the solutions. That's why I usually use codeforces.

My approach is just participating every div2,div1 contest and solve till difficulty 2400~2600

'What works best' differs from person to person. Use at least atcoder besides codeforces. I think I am good in typical / standard problems, but not so good in observation based adhoc problems, hence I am trying atcoder now. Atcoder problems' difficulty can be found here.

General strategy for everyone is solving problem in difficulty in which you are not fully comfortable, but which is not so hard for you. If you can solve 7/10 random problems of a certain difficulty within reasonable time, you should move to higher difficulty.

Well my question isn't really about which problems you solve, but how do you spend time on each problem, like

• How often you are stuck
• How long do you think of a solution while not seeing editorial
• In general how long it takes to solve one problem

(I'm sorry, these are very hard questions to answer)