The IB Math IA is scored out of 20, across five criteria: presentation, communication, math communication, personal engagement, and reflection. The single biggest predictor of a 6 or 7 grade is whether your research question is narrow enough to generate genuine mathematical insight — not whether your math is hard.
That's why "modeling the spread of a virus" scores worse than "comparing two SIR-model variants on 2020 COVID data from my home city." The first is a textbook chapter. The second is your own work on your own data.
Topics that score well (and why)
- Optimizing the shape of a specific real object — not "cylinders in general," but the actual bottle on your desk, with measured constraints and a defended trade-off.
- Comparing two actual models against data you collected — regression analysis on your running splits, your basketball shot arcs, your Spotify listening pattern.
- A named paradox worked out rigorously — Simpson's paradox in a small dataset you own, the Monty Hall problem extended to n doors with simulation + proof.
- Markov chains on something you can count — weather at your town, song transitions in an album you like, board-game mechanics.
- Error analysis of a physical measurement — how Fermi estimation compares to direct measurement for something you can actually measure.
Topics that consistently score 4 or below
- "Modeling population growth with exponential functions" — the whole topic is a textbook exercise; there's nothing for you to add.
- "The mathematics of music" — too broad, and almost everyone writes about Fourier series without really using them.
- "Fibonacci in nature" — every examiner has read this 100 times. They grade it harshly by default.
- "The Birthday Paradox" — a great topic if you extend it to something specific; a dead topic if you just restate the standard calculation.
- Anything that's really a physics or economics problem with some math dressing — the IA is graded on mathematical engagement, not domain knowledge.
The HL vs SL scoping difference
SL IAs score well when the math is clean and the engagement is visible. Don't try to use HL-only topics like Taylor series or differential equations at SL — the graders flag mismatched scope and you lose communication points.
HL IAs need at least one non-trivial technique. Regression alone isn't enough — pair it with a formal derivation, a constrained optimization, or a proof. The HL rubric explicitly rewards depth.
How to turn a weak topic into a strong one
A worked example: "Modeling COVID spread" is weak. Here's the transformation to strong:
- Narrow the geography: "COVID spread in Lahore, March–July 2020."
- Narrow the model: "Comparing the SIR and SEIR models."
- Narrow the data: "Using WHO daily case data and Google Mobility data."
- Add a personal angle: "...against the lockdown dates my school observed."
- Add a trade-off to defend: "...and explaining why SEIR's delay term was necessary for that window."
Now you have a 5-sentence research question that nobody else can write, requires real math, and gives you something to reflect on at the end. That's a 6 or 7.
