NOI Singapore: The Five-Hour Contest Format Guide

NOI Singapore runs on a format that changes how a student has to prepare: a single five-hour session with four to five tasks. That is not a longer version of a school paper. It is a different sport.

Five hours, four to five tasks

The National Olympiad in Informatics has been organised by the School of Computing at the National University of Singapore since the first contest in 1998, and it selects Singapore's team for the International Olympiad in Informatics.

The contest is one five-hour session. Submitted programs are judged on correctness and on resource use — a solution must produce the right output and do so within the specified time and memory limits.

NOI Singapore is run as a preliminary contest followed by a final contest, so a student's first encounter is the preliminary round rather than the full selection event.

Why resource limits change the preparation

In a contest judged only on correctness, a student who finds any working solution has finished the task. Here they have not.

A correct algorithm that is too slow scores little or nothing on the large test cases, which means a student must reason about complexity before implementing rather than after. Deciding that an input size rules out a quadratic approach is the actual skill, and it is one students frequently skip.

This is also why C++ is the practical language at this level. Python is an excellent language for learning the algorithms, but a correct Python solution can exceed a time limit that the identical algorithm clears comfortably in C++. At the point where resource use is scored, the language stops being neutral.

Pacing five hours

Most lost marks in a long contest come from pacing rather than from algorithmic weakness, and students who have only sat short papers get this wrong on their first attempt.

The reliable approach is to read every task first, spending ten to fifteen minutes on nothing but reading. Tasks in a set like this are not ordered strictly by difficulty, and a student who works through in order sometimes spends two hours on the task they find hardest while an easier one sits unread.

Then bank the straightforward tasks completely before starting anything hard. A fully solved easy task is worth more than two partially solved hard ones in almost every scoring scheme.

Finally, stop coding before time is up. A submission that is never tested is a submission that frequently fails, and the last twenty minutes are better spent checking edge cases than writing new logic.

Partial scoring and subtasks

Olympiad-format tasks typically award marks for solving restricted versions of the problem — smaller inputs, special cases, simplified constraints.

That makes the brute-force solution a genuine strategy rather than a fallback. On a hard task, writing the obvious slow approach and banking the small-input subtask is usually better than holding out for the full solution and submitting nothing.

Students coming from mathematics olympiads find this counter-intuitive, because an incomplete proof earns little. Here an incomplete solution can earn most of the marks on offer.

What the tasks test

The algorithmic ground is the standard olympiad toolkit, and the depth rather than the breadth is what distinguishes the final contest.

Sorting and searching including binary search on an answer. Greedy reasoning with an argument for why the greedy choice is safe. Dynamic programming, which is the single technique that most separates students at this level. Graph algorithms — traversal, shortest paths, and tree reasoning. And data structures chosen for their complexity rather than their familiarity, because resource use is scored.

Our NOI Singapore preparation course works tasks under full five-hour conditions, because a student who has only solved problems one at a time has not prepared for the format they will meet.

Eligibility, stated plainly

A student must attend a local junior college or secondary school. Nationality is not a restriction.

In practice that means an international student enrolled at a Singapore school may compete, while a student schooled outside Singapore may not — the opposite of the assumption most families make.

Students outside Singapore who want the same training should look at contests open to international entrants. The Australian Informatics Olympiad and the Canadian Computing Competition both accept overseas schools, and USACO runs online contests open to anyone.

Building towards it

A student who is not yet ready for a five-hour algorithmic contest should not start with one.

Computational thinking contests like Bebras build the reasoning without requiring code, and shorter programming contests build implementation fluency before endurance becomes the constraint. The order matters: a student who can reason but cannot implement wastes a five-hour contest, and so does one who can implement but cannot pace themselves.

The gap between a student's first long contest and their second is usually the largest improvement they will record, and it comes almost entirely from having experienced the format once.

How a school prepares students for it

Schools that do well at NOI Singapore tend to run a weekly problem session rather than a term of lectures, because the skill being built is practice rather than coverage.

The most valuable single exercise is a full-length mock contest. Four or five tasks, five hours, no help, and a review afterwards that looks at where the time went rather than only at which solutions were right. A student who has done two of those arrives at the NOI Singapore preliminary contest knowing how the day feels, and that knowledge is worth more than an extra algorithm.

The second most valuable exercise is reading other people's solutions after the fact. A student who solved a task slowly and then reads a clean fast solution learns the technique in context, which is how most competitive programmers actually acquire their toolkit.

Where it leads

Beyond international selection, the preparation transfers directly to university computing and to technical interviewing, which is built on the same material.

Formats, dates, task counts and eligibility are set by the organisers each cycle. Confirm the current details with the NUS School of Computing before planning around them.