Hello , i just wanna ask that what and how all coders expert or above think when they see a problem , yesterday one told me that it is different when you reach a certain level . so all i wanna know is how to bring that shift as i am practicing 1600 right now but not able to think in that way , even after reading the editorial i am not able to digest that way of approaching the problem , so if any one of you knows something please tell me it'll surely help me.
Thankyou.
one of my chinese expert friend told me "LEARN AS MUCH ALGORITHM AS YOU CAN, MAKE NOTES OF ALL ALGORITHMS MEMORIZE IT TO YOUR HEART CONTENT AND KEEP GIVING AS MANY CONTEST AS YOU CAN"
i mentioned chinese because different countries have different programs and this may not represent method which people from other countries followed to become an expert.
imo Chinese are implementing, Russian are observating
From this blog, I would say that American problems type are implementing, Chinese tends to be about technique (complex data structres ig), and Russian is obervatiom.
I think that CF tends to be Russian (not because it is made by a Mike, a Russian, but because problem setters have to come up with weird problems so that it won't be the same as the older ones, so it's kinda ad-hoc) so I don't think that one should learn as many algorithms as possible. OI on the other hand, sometimes are heavy in data structures/being very classic (take an OI in Vietnam as an example, one of its problem is the tower of Hanoi problem but with 4 poles), so I don't think that being better in CF makes you better at OI.
IOI which is arguably the ultimate OI has shown very noticeable correlation between cf and IOI across multiple years even with people like rainboy throwing off the data https://codeforces.com/blog/entry/145502. I don't think its a coincidence that the last four ioi winners were all lgm
What you mean about Russian?
https://codeforces.com/blog/entry/126310
But I am Russian...
wtf you mean learn as much algorithm as you can? Um_nik once said that all you need to reach CM is 2 pointers and binary search (and also some basic graph knowledge if you like). As I can see, most of the problems I've done don't need complex data structures/ algorithm (I'm not a expert so idk tho)
Maybe these days you need some more like basic understanding of graphs and trees, know DP, often d2D is segtree (just to black box PURQ and know the basic applications, but you do not need to know walk or lazy prop). I think underrated is learning "solution-shape algorithms" instead of "data-structure algorithms". Like how looking at the extreme values can occasionally instantly solve 1600-1900 tier problems
did bro just self plug
also this guys just yapping i can barely do knapsack DP right and im comfortably 1600+
Oops i was talking about cm. You probably don’t need segtree for expert but I’d make the argument that dp (and maybe knowing some of its more niche variants) will help you get to expert.
also “comfortably” is a term I would use with caution here…
lol I like how cf doxxes whether I am on phone or not based off the quotes
oh mb im blind I looked at the blog title instead of the guys comment specifically
also nah to prove this i have registered for the next contest and i am confident I will not lose blue after it.
Yeah bro, completely agree with you! I’ve actually wrapped up over 200 problems on LeetCode, but I recently switched over to Codeforces. Do you have any tips on how to adapt efficiently and get the hang of the style here?
pretty sure um_nik wont say this in 2026, lol just to become pupil now u have to solve till D in contests
If you can't do it with editorial, then you're trying something too hard. In my opinion, you should be able to solve 30-50% of questions of the rating you are doing without editorial (at least this is what I aim for).
You should definitely be able to understand editorials for the rating you are practising though, or if they're badly worded try get an AI to explain them to you as you can ask questions and gain a better understanding.
TL;DR try a lower rating for now, you're aiming too high, only go up to 1600 when you're comfortable with 1500s and 1400s along with their basic algos — Binary Search, DP, BFS/DFS, idk what else rlly at that rating
https://codeforces.com/blog/entry/98806 might help
Learn to be satisfied with your best performance and wake up trying to improve however you define your best as, even if it's very incremental progress (Coming from a non-expert though)
Dude, you are already doing contests with expert performance, you can go to expert if you do more contest
But I guess this is either a smurf account or a person using AI, or both cuz no way someone just registered and got Expert performace instantly
Try to master graphs and trees based questions rated around 1500-1700 ,these not only teach trees and graphs manipulations but also builds intution for dp like finding subtree sizes etc etc which is later used in some dp problems as well (fenwick tree total inversion count etc etc).Recently graph trees have appeared in almost every contest which really became the diffrentiator..Hope it helps