I dont know the answer to that, other folks please chime in. I am not sure there is a feasible career like this open for everyone.

A few things I can think of — CP trainers (Need to be really good and popular and be in a place where there is a demand), Quant firms ( I have heard it involves hardcore problem solving, and gives crazy money. But you need to be really good to crack their interviews ) , SDE at big companies like Google (I guess <5% of your work will revolve around problem solving, but it pays good.

Wait I forgot about AI. I think AI research companies have good amount of algorithmic and problem solving related work. I hope they will pay good.

A few question to better understand you.

• Have you partially solved/considered solving CSES problemset ?

• Have you trained DSA and Algorithm outside CF ? Like CLRS book , or a College courses of DSA/Algo ?

• Have you solved "standard" algorithmic problems ? For example how much of these standard problems on Leetcode you think you already know of https://leetcode.com/problemset/?listId=wpwgkgt&page=1&difficulty=HARD

• What limits you in your contest performances among "Limiting factor categorization" below ?

• What do you think are your weak areas ? (Dp/ counting problems / graph and trees / adhoc question )

• What limits you in your contest performances most of the time among "Limiting factor categorization" below ?

• What do you think are your weak areas ? (Dp/ counting problems / graph and trees / adhoc question / etc )

• I see that you have solved 50 problems of 1600 difficulty and 38 problems of 1700 difficulty. How many of these 88 problems you solved completely on your own / solve with some hints from editorial / solved by fully seeing the editorial ?

Since I haven't studied FFT, I cant help you there. I have 3-4 upsolves pending on FFT. So I will be studying FFT soon. Once I understand, and implement it, and solve a few problems. I will probably get back to you.

My thinking is that its pretty advanced topic, and below CM, anyone would better focus of DP, Maths, Graphs & Trees, and Data structures.

you basically want all possible combination of 2*b = a+c , so in other words if you are able to compute all possible values of a+c in the array, you will be able to check for each a+c , how many 2*b exist, and to calculate all possible a+c , you will need FFT, you will keep all the values in the array as powers , and then multiply the polynomial with itself, thus getting all possible combinations of a+c as the powers of the new polynomial, this way you will be able to calculate the number of b which satisfy the given constraint. It's a really nice question. I also wasn't able to solve it during contest unfortunately :(

One of the problem I have no idea how to solve (unthinkable). But it seems interesting and very classical at the same time.

Feels thats for now you are sort of stuck at pupil, but its only a problem if you aren't able to solve problems on your own.

• Try to understand what limits you in your contest performances most of the time among "Limiting factor categorization" below ?

• If its like misunderstandings / silly mistakes / slow or bad implementation / slow coming solutions — then you would improve by practicing more. No need to worry.

• If its more like unthinkable solutions and wrong solutions. Then you might need to change the way you practice.

• I see that you have solved ~80 problems of 1600+ difficulty. How many of these 88 problems you solved completely on your own / solve with some hints from editorial / solved by fully seeing the editorial ? If you are seeing the editorial too much / too early, than it might be a problem

You are solving way too much of 800 rated problems (215 out of 362 are 800 rated).

(1) You must solve more challenging problems. (1100, 1200, 1300)

(2) You must try hard to solve them on your own, and use editorial to solve at max 10% of the problems

(3) Make sure to upsolve the contests you give.

From your id , looks like you have just started. But I see you have solve 67 problems, our of which 50 are 800 difficulty level.

Same 3 points as this comment — https://mirror.codeforces.com/blog/entry/139837?#comment-1249428

By the time you would have solved 50+ (1200-1300) rated problems "on your own", you should be near pupil.

To get to specialist, I would consider two things — Standard and classical problem solving capability and Codeforces practice.

(1a) Solve Top solved Leetcode hard problems . These are mostly standard and classical/well-known problems, but they are very good for growth and understanding, if you are a beginner.

(1b) Also solve CSES Problemset. Again, classical and standard problems that are very educational. For specialist target to solve 60%+ of all except Additional and Advanced sections.

(2a) Make sure to upsolve all contests.

(2b) Progressively solve problems of difficulty rating 100-200 more than your current rating. (Prioritise classical problem solving stated above first)

I would guess, that once you have solved 100+ Leetcode hard on your own, and 60% of CSES problems except Additional and Advanced sections and 100 problems in codeforces of difficulty (1400-1500) "on your own", there is a very high chance you will be specialist.

Regarding topics. I would say dont do topicwise if you are a beginner. Solve all standard problems, and if you feel that you are specifically weak at something, then you can consider solving more on that topic.

But first consider improving your all round problem solving to a decent level

20-30 problems

A bit off topic, but one thing I an just guessing is that you might adjust your mindset about C++, if you are too negative.

I understand the frustration for some simple string relating, array slicing, mapping operations you have to write huge code in C++ and you curse the language and feel why they made it so complicated. But try to understand that there is a reason for that. They had to make things efficient, and thats why its like that. Try to "Respect and enjoy" the language.

I used to have this mindset about Java, where I would curse the verbosity.

You may not use C++ as your main language, but you should 99% of time be able to rewrite your code in C++ whenever you have to.

By the way, if you havent already, you can see conqueror_of_tourist, he is a Python LGM.

Just a small fact about language versatility, I recently had to learn Go for my project work, so I solve practice problems in Go as well now. I think you might be overestimating the toughness of using both python and c++, hand in hand.

Your profile looks great. You contest performance consistency is extremely remarkable, so is your practice. I am in awe. Great !! .

Just out of curiosity, would like to know what is the major limiting factor for you performance in contests.

Among the 176 problems you solved, 130 were in 800-900 difficulty range.

Its better to solve more problems that are above your level to see better growth.

At beginner level, I would advise also practicing standard and classical problems, they are very educational and improves you a lot.

• Solve Top solved Leetcode hard problems . These are mostly standard and classical/well-known problems, but they are very good for growth and understanding, if you are a beginner.

• Also solve CSES Problemset. Again, classical and standard problems that are very educational.

• Make sure to upsolve all contests you give on codeforces.

• You can consider giving Atcoder ABC contests as well, they are very educational, and beginner friendly. Problem quality is really good and you can discuss problems on codeforces after contest. There is usually a blog post for each contest.

• Target to solve more 1200-1300 difficutly problems.

"I came up with a solution, but I couldn't code it correctly in time, which decreased my rating" — That is okay. It sometimes happens. It will improve with practice. At beginner level I would say you can just stop caring about ratings at all. Just focus on learning, "being able to solve problems".

Once you are green/cyan. It wont matter how many ups downs you had as a grey. What matters is that you reached green/cyan.

If you had to sacrifice ratings of one contest for the sake of learning in long term, you should do that. (For example — instead of running to implement a relatively-shitty solutions, think back 3-4 min, and think of a more elgant approach).

Sure. Its too late for me right now. But I will solve this problem tomorrow morning and then find issues in your solutions.

I solved it. My submission — https://mirror.codeforces.com/contest/2027/submission/307335386

Having a bit of trouble understanding your solution. Can you define what precisely dp[i][j].s and rem stores ? And give a brief 2-3 liner rundown of your solution.

I understood that dp[i][j].f stores the min cost considering array a[0...i-1], and considering that you start with k=j .

inifnite Finally found issue in your approach. The subproblem definition unfortunately is the problem. It doesn't follows the optimal substructure property.

Lets denote the subproblem by $$$ P_{ij} $$$ , We say that optimal solution of $$$ P_{ij} $$$ is $$$ DP_{ij} $$$ stored as $$$ (f,s) $$$, where $$$f$$$ is the min cost and $$$s$$$ is max remaining sum. As per your above definition

Optimal solution $$$ P_{ij} $$$ cannot always be calculated from previous optimal solutions of $$$ P_{xy} $$$ (where $$$x \lt i$$$ or $$$y \lt j$$$ ).

For example while arriving at the solutions of $$$ P_{i,j} $$$ from $$$ P_{i-i,j} $$$ , a suboptimal solutions of $$$ P_{i-i,j} $$$ might be better, which doesn't necessary minimizes $$$f$$$ (the min cost), but provides more $$$s$$$ (max remaining sum) . So that later instead of doing an operation costing $$$m-j$$$, you could just use the remaining sum

1
6 8
7 6 2 2 8 6
17 16 15 14 12 7 3 1

set k=4, b[k] = 14, eat [7,6]
cost += 8-4

set k=5, b[k] = 12, eat [2,2,8]
cost += 8-5

set k=6, b[k] = 7, eat [6]
cost += 8-7

total cost = 9

PS : This was really hard to get to. I had to write a stress testing script and generate random test cases.

Your solve count is good, and practice consistency is good. No abvious suggestions for now. Contest performance is also very consistent.

I would like to understand you better to be able to suggest anything .

(1) What are the major limiting factor for you performance in contests.

(2) How much of the problems (1400+) have you solved completely on your own / solve with some hints from editorial / solved by fully seeing the editorial ?

(3) Any weak areas ? (Dp/graph/counting/maths/etc)

(4) Solved CSES problems set ? If yes what percentage (excluding additional and advanced sections)

(5) What is your strategy when you cant solve a problem ? Whether you give up and check editorial to quickly, or try a lot, or give a fix amount of time ?

By the way, "Slow Fourier Transform" is lit. I wish codeforces allowed me to be in more than one organization.

SDE Job, since June 2023.

Your solve count is good. Although your performance is sometimes very inconsistent, so maybe you are relying more on Memory and Analogy, rather then creativity. (See https://mirror.codeforces.com/blog/entry/112856). Its possible that you are reading the editorials too quick, although that might not be the case. A few questions

(1) Based on the above blog I linked, what one/two among the three — (Memory, Analogy and Creativity) majorly gets you to solve the hard problems.

(2) Any weak areas ? (Dp/graph/counting/maths/etc)

(3) What is your strategy when you cant solve a problem ? Whether you give up and check editorial to quickly, or try a lot, or give a fix amount of time ?

(4) What are the major limiting factor for you performance in contests.

About missed obeservations — I dont really understand the question. If I miss an observation, I would not be able to solve then problem, and after trying really hard, I would "have" to read the editorial to see that observation. Then I would get that observation, and would implement the solution. Maybe I would also think about the intuition behind the observation, and try to see what thought process could have made me arrive at that observation. But thats all, what else can we do.

Based on you problem solve count, ideally you should be at rated at least 1200-1300. Also the unsolved count if a bit high, suggesting that you are leaving "hard-for-you" problems for later (Those problems that you weren't able to solve, contains the corners of problems and ideas that are hard for you. Those are the most educational problems for you).

Are you giving up and checking editorial too quickly ? Do you upsolve all contests you give ? Any weak areas ? (Dp/graph/counting/maths/etc)

Based on the this blog ( https://mirror.codeforces.com/blog/entry/112856 ), what one/two among the three — (Memory, Analogy and Creativity) majorly gets you to solve the hard problems. It you are mostly relying on Memory and Analogy, then it might be a problem.

What are the major limiting factor for you performance in contests ?

" 8/10 problems I solve, I have to look at edu's" — This is really bad. You should be able to solve problems on your own, at lest 70% of the problems. The key is the skill to solve the problems, not "knowing the solutions".

It wont do much good to solve most problems with editorials, After few months, you will forget the solutions, and will be back to square one.

The suggestion is to internalize and understand that just knowing the solutions wont do you any good, and coming up with the solution is the real deal.

Try really hard to solve problem, and when you cant, only look at the editorial partially. Like see the first few lines, and as soon as you come across any idea you haven't thought of before, close the editorial and start trying again with that idea.

Before all that, you need to understand the reason why you want to do CP. And if you really want to and have to do CP or not ?

What if you are allowed to trade all your time you spent on CP, on another skill — would you take the offer ?

If you have no good reason and not enough motivation, and have better things to do, then you shouldn't do CP

"I think its because I keep trying to fully prove my ideas instead of sending a guess". This is really good. This tells that you are a real problem solver. I have many slow submissions due to this.

In fact, for a while, I did this. I would solve and only submit is last 5 min, to improve my problem solving. So I had no option to guess, but to "really solve" the problem.

So do you feel you are slow in only coming up with the correct solutions you believe, or you are also slow in implementation ?

What about other factors like "Bad Solution", "Misunderstanding Problems" . Does that happens with you or not ?

For example see this, and tell if you have any other major limiting factor for you performance in contests, apart from "Slow coming solutions" or "Slow implementation"

Also another question — How many of these (1400-1800) rated problems have you solved on your own / solve with some hints from editorial / solved by fully seeing the editorial ?

Looks like too many of your problems solved are 800-900.

I would recommend solving 1300,1400 and 1500 rated problems now.

And make sure to try really hard at the problems, to maximize learning. When you cant solve problems, only look at the editorial partially. Like see the first few lines, and as soon as you come across any idea you haven't thought of before, close the editorial and start trying again with that idea.

2nd thing is that if you havent solve standard and classical problems, do that too. Those problems are very educational and will be good for your current level.

• CSES Problemset (Except Additional and Advanced section)

• Leetcode Hards ( Pick 40-50 top solved problems)

At beginner level, I would suggest you to get some good foundation in DSA and Algorithms, to have good fundamental and conceptual knowledge. It will help you a lot. I personally watched this Free MIT course — https://www.youtube.com/playlist?list=PLUl4u3cNGP6317WaSNfmCvGym2ucw3oGp , and read a few relevant chapters or CLRS book, and it helped me a lot.

So you can do some classical learning via books and course while you practice on CF. That would be great.

Unsolved count is too high in my opinion suggesting that you are leaving "hard-for-you" problems for later (Those problems that you weren't able to solve, contains the corners of problems and ideas that are hard for you. Those are the most educational problems for you).

Since you have peaked 1550, you can now focus on solving 1500, 1600, 1700, 1800 rated problems. Count for problems solved for these difficulty is pretty low.

What is your major Limiting factor in contest performances ?