Enrollment for the ISI-CMI is Live! Apply Now β†’
EduGlobal Logo
Edu Global Institute For Olympiads and Research
Competitive Mathematics

Maths Physics Coding: One Exam Trend for 2027

Three subjects, three governing bodies, no coordination β€” and the same shift away from technique.

R Rohan Kumar Singh Founder & Lead Mentor, IIT Madras alumnus 5 min read
A student planning a competition preparation timetable

This guide covers maths physics coding. We have spent the past few weeks reading primary sources across three subjects β€” MAA announcements, Oxford's own admissions pages, UKMT notices, the IPhO statutes, and a new study from ETH Zurich.

We went in expecting to write three separate updates. What we found instead was one pattern, appearing independently in places that do not coordinate with each other.

Across mathematics, physics and computing, assessment is moving away from technique and toward understanding. Here is the evidence, and what we think it means for a student preparing now.

Four changes in mathematics, pointing one way

For 2026–27, AMC scores no longer affect USA(J)MO qualification once a student has qualified for the AIME. Previously the AMC was combined with the AIME score, weighted ten times β€” twenty times in 2025–26.

The 2027 AIME moves to Pearson centres only, restricts entry to age 13 and over, drops the alternate AIME II date, and splits its three hours into two sealed halves of eight and seven questions. The MAA cited competition integrity.

Oxford retired the MAT, used from 2007 to 2025, and now requires the TMUA of all Mathematics and Computer Science applicants.

UKMT is trialling 000–999 answers for the last three Senior Mathematical Challenge questions, replacing multiple choice on the hardest problems.

Each change independently reduces what test-taking technique can buy you. Elimination, backward-working from options, roaming the paper for easier marks, and compensating for a weak round with a strong earlier one β€” all four have been narrowed.

Physics has been saying it for years

The IPhO syllabus does not need to change, because it already states the principle outright: problems test "creativity and understanding of physics rather than mathematical virtuosity or speed of working." Mathematical manipulation is to carry only a small share of the marks.

The structure enforces it. Theory is 30 marks; the experimental paper is 20. Forty percent of the result is decided in a laboratory β€” measuring honestly, analysing uncertainty, and designing procedures the paper deliberately leaves open.

And problems may introduce concepts outside the syllabus entirely, with an explanation in the text, which makes memorised coverage structurally insufficient.

Physics has been assessing understanding over technique for years. Most preparation simply has not caught up.

Computing reached the same place from the other direction

No rule changed here. The world did.

The ETH Zurich study of 100 students found that computer science knowledge had the greatest effect on performance at AI-assisted programming tasks, holding even after controlling for general cognitive ability. Writing skill mattered too. And students who used large language models heavily in daily life performed worse β€” a correlation the authors explicitly decline to explain causally, and so do we.

The lesson is not that AI is bad. It is that AI assistance raises the value of understanding rather than replacing it. You cannot specify precisely what you want, recognise a wrong answer, or locate a fault without knowing how programs work.

Competitive programming holds the line for the same reason: a correct solution that is too slow scores zero, and no amount of prompting changes the complexity of an algorithm.

What connects them

Three subjects, three different governing bodies, no coordination β€” and the same conclusion.

In every case, the assessment is being moved closer to the question "does this student actually understand?" and further from "can this student work the format?"

We think there is a straightforward reason. Technique is now cheap. Past papers are free and abundant, coaching is widely available, and anything that can be systematised can be taught quickly to many students. When a proxy becomes easy to optimise, it stops measuring the thing it was standing in for β€” so the examiners move the measurement.

Understanding resists that compression. You cannot shortcut a student into recognising which technique a disguised problem reduces to, or into estimating an uncertainty honestly, or into knowing why one algorithm is logarithmic and another quadratic.

What this should change in preparation

Four things, applying across all three subjects.

Practise in the real format, immediately. AIME mocks split eight-then-seven. SMC final questions with the options covered. USACO problems against real time limits on large inputs. Preparing in an outdated format trains a strategy that will not be available.

Weight preparation by where the marks are, not by what is comfortable. The clearest case is physics: 40% of the examination sits in the laboratory and receives a small fraction of most students' time. A strong theorist improving at theory feels productive while losing on the part that decides the result.

Classify errors instead of counting them. A score tells you nothing actionable. Whether a mark was lost to a concept gap, a careless slip, a slow method, a misread or poor question selection determines entirely what next week should contain. Four of those five are habits, not knowledge.

Build understanding before technique, not after. This is the thread through everything above. The examiners have decided that understanding is what they want to measure, and across three subjects they have redesigned assessments to get at it.

A note on reading this yourself

Every claim here comes from a primary source, and we have linked them. We would encourage checking them, for a specific reason: a great deal of what circulates about these changes is out of date, and in a year when this much moved at once, out-of-date guidance is actively harmful.

Where sources disagreed, we have said so rather than picking the tidier version. Two published analyses of the 2026 AMC 10A reach different conclusions about whether the topic mix shifted β€” one finds it broadly familiar, the other finds the hardest questions spread more evenly across areas. We do not think the question is settled, and we are not going to pretend otherwise to make a cleaner story.

What is settled is the direction. Four mathematics changes, a physics syllabus that has said this for years, and a computing study pointing the same way. For a student starting now, the preparation that suits this landscape is the kind that was always better and is now simply necessary: slower at the start, built on understanding, and honest about where the marks are actually going.

Everything above is how we approach maths physics coding with our own students, and it is the method behind the results on this site.

Everything above is how we approach maths physics coding with our own students, and it is the method behind the results on this site.

Sources

  1. American Invitational Mathematics Examination - format and 2027 changes, https://en.wikipedia.org/wiki/American_Invitational_Mathematics_Examination
  2. University of Oxford Mathematical Institute - Maths Admissions Test notice, https://www.maths.ox.ac.uk/study-here/undergraduate-study/maths-admissions-test
  3. UKMT - Senior Mathematical Challenge, https://ukmt.org.uk/senior-challenges/senior-mathematical-challenge
  4. IPhO - Statutes of the International Physics Olympiads and Syllabus, https://www.ipho-new.org/statutes-syllabus/
  5. TechXplore - What skills do people need to successfully program with AI? (ETH Zurich study), https://techxplore.com/news/2026-04-skills-people-successfully-ai.html

Questions people ask

What is the single biggest change for 2026-27?

In mathematics, that AMC 10 and 12 scores no longer affect USA(J)MO qualification once a student has qualified for the AIME. Previously the AMC was combined with the AIME score, weighted ten times and twenty times in 2025-26. It turns the AMC into a gate rather than a contribution, and changes how preparation hours should be allocated.

Is this really a pattern, or separate coincidences?

They are genuinely independent - the MAA, Oxford, UKMT, the IPhO and a Swiss university research group do not coordinate. What makes it a pattern is the direction. Every change narrows what test-taking technique can buy and widens what understanding is worth. We think the reason is that technique has become cheap to teach, so it stopped measuring what it stood in for.

What should we change first?

Practise in the current format rather than an outdated one, and weight preparation by where the marks are. The clearest example is physics, where 40% of the IPhO result sits in the experimental paper and receives a small fraction of most students' preparation.

Does this mean past papers are less useful?

No, but how you use them matters more. A score out of a paper tells you nothing actionable. Classifying each lost mark as a concept gap, careless slip, slow method, misread or poor question selection determines what the next week should contain - and four of those five are habits rather than knowledge.

Is AI making computing competitions irrelevant?

The evidence points the other way. The ETH Zurich study found computer science knowledge was the strongest predictor of success at AI-assisted programming, holding after controlling for cognitive ability. Competitive programming builds that understanding, and resists outsourcing because a correct but too-slow solution scores nothing regardless of how it was written.

How do we know this guidance is current?

Every claim links to a primary source - MAA announcements, Oxford's own admissions pages, UKMT's notices, the IPhO statutes and the published study. We would encourage checking them. Where sources disagreed, such as on whether the 2026 AMC 10A topic mix shifted, we have said so rather than choosing the tidier version.

Where should a student start given all this?

With a diagnostic rather than a syllabus, because the right starting point depends on the error profile rather than the age or the contest. The landscape above rewards preparation that is slower at the start and built on understanding - which was always the better approach and is now simply the necessary one.

Get a study plan for this

Tell us the class and what they are working towards, and we will send a plan built around it, plus the next free trial class. No cost, and we will not pass your details on.

One reply from a real person, usually the same day. Unsubscribe from any email.

Read next

Free resources, straight to your inbox

Problem sets, strategy guides and olympiad registration deadlines β€” sent when they matter, never more than twice a month.