Updated November 2026
TL;DR
You’ll find 50 math research topics for high school students below, with a difficulty level and a short list of what you’ll need. They’re sorted into five groups: number theory and patterns; probability, statistics and data; applied math and modeling; geometry and topology; and discrete math and computer science. Most need only algebra, a free tool like Python or a spreadsheet, and a notebook. Worried you don’t know enough math yet, or that you’d be figuring it out alone? You can apply to the Horizon Academic Research Program.
If you like math enough to wonder what happens after the textbook gives you the answer, a research project is where you find out. You pick one question, work on it for a few weeks, and end up with something you can explain to a stranger. It is also a good way to see how your thinking holds up next to people who love math as much as you do, like a mentor or a judge at a science fair. This guide gives you 50 questions in five groups, and each one is small enough to start this week. Skim the groups, pick the one that matches what you already enjoy, and try the first step on paper before you open any code. If you’d like a ready-made program first, see our list of great math research programs for high school students.
What makes a good math research topic for high school?
A good topic is a question you can answer with the tools you have and a source you can read. Check three things before you commit. Can you state it as a question with a checkable answer? Can you finish a small version in a few weeks? Is there at least one book, article, or lecture note that explains the background?
If a topic fails the first test, narrow it. If it fails the second, pick one special case. Our guide to research programs for complete beginners shows what the first weeks tend to look like for people starting from zero.
Which math research topic fits what you already like?
| If you like | Start with topics |
| Puzzles and patterns | 3, 4, 8, 10, 45 |
| Coding | 6, 11, 23, 42, 48 |
| Sports, games, or money | 12, 14, 17, 26, 30 |
| Art and design | 2, 33, 34, 37 |
| Music | 29 |
| Biology or medicine | 21, 22, 35 |
| Physics or engineering | 24, 25, 28, 39 |
| Social questions | 16, 41, 49 |
Number theory and pattern research topics for high school students
Number theory studies whole numbers and the patterns hiding in them. Research here is mostly small cases, careful guesses, and proofs, so a notebook and some patience go a long way.
1. Do the spirals on a sunflower really follow the Fibonacci numbers?
Count the spirals running each way on photos of sunflowers, pinecones, or pineapples, and compare your counts with the Fibonacci numbers. Then check how the ratio of neighboring Fibonacci numbers settles near 1.618.
Level: Starter. You’ll need: photos from a library or your own camera, a spreadsheet.
2. Does the golden ratio really show up in famous art and buildings?
Pick a few works people say use the golden ratio and measure them from published images. Then test whether your answer changes when you move the edges you measure from, which is where many of these claims fall apart.
Level: Starter. You’ll need: images, a ruler or an image tool, a spreadsheet.
3. Can you explain why shading the odd numbers in Pascal’s Triangle makes a Sierpinski triangle?
Write out the first 32 rows, shade the odd entries, and watch the pattern appear. Then try to prove why it happens using how the entries are built from the row above.
Level: Starter. You’ll need: graph paper or a few lines of Python.
4. Can you find every right triangle with whole-number sides?
These are the Pythagorean triples, like 3-4-5 and 5-12-13. Start with Euclid’s formula, then check whether it catches every triple you can find by brute force.
Level: Starter. You’ll need: a pencil, then a short Python loop.
5. Can a check digit catch every single typing mistake?
Barcodes and book numbers end in a check digit that flags typos. Design your own scheme, then test how many one-digit mistakes and swapped neighbors it catches.
Level: Starter. You’ll need: a spreadsheet or Python.
6. Is it really hard to break a small RSA code?
RSA, described publicly in 1977, depends on the fact that multiplying two big primes is easy and un-multiplying them is slow. Build a toy version with small primes and time how long factoring takes as you make the primes bigger.
Level: Stretch. You’ll need: Python.
7. Do primes get rarer as numbers get bigger, and how fast?
Count the primes up to 10, 100, 1,000, and beyond, then compare your counts with n divided by the natural log of n. See how close the estimate gets.
Level: Stretch. You’ll need: Python.
8. Does the Collatz rule always end at 1?
Pick a number, halve it if it’s even, triple it and add one if it’s odd, and repeat. Nobody has proved that every start ends at 1, so your project is to explore how long the trips take and what patterns show up.
Level: Starter. You’ll need: a spreadsheet or Python.
9. Can you tell which primes are the sum of two squares?
Test the primes up to a few thousand and see which ones can be written as a square plus a square. Then check whether the ones that work, apart from 2, all leave a remainder of 1 when divided by 4.
Level: Stretch. You’ll need: Python.
10. Can every whole number be built from Fibonacci numbers in exactly one way?
Yes, if you never use two neighboring Fibonacci numbers, and that fact is called Zeckendorf’s theorem. Try it for the numbers 1 to 100, then try to explain why the choice is always unique.
Level: Stretch. You’ll need: pencil, then Python.
Probability, statistics and data research topics for high school students
This group is about drawing honest conclusions from numbers that don’t line up neatly. Most projects here start with a simulation or a dataset, which makes them easy to start and surprisingly hard to finish well.
11. Can you estimate pi by throwing random points at a square?
Drop random points in a square with a quarter circle inside it and count how many land inside. Then measure how the error shrinks as you add more points.
Level: Starter. You’ll need: Python or a spreadsheet.
12. Does the house always win at roulette?
An American wheel has 38 pockets, a European wheel has 37, and a straight-up bet pays 35 to 1 on both. Work out the average loss per bet on each wheel, which comes to about 5.26% and 2.70%, and see why the extra pocket matters.
Level: Starter. You’ll need: A pencil, then a quick simulation.
13. Do two people in a room of 23 share a birthday more than half the time?
Yes, about 50.7% of the time, which surprises almost everyone. Simulate thousands of rooms of different sizes and plot where the chance crosses 50%.
Level: Starter. You’ll need: Python or a spreadsheet.
14. Does a lottery ticket ever have a positive expected value?
Pick a lottery with published odds and prizes, and compute the average payout for a one-dollar ticket. Then ask what changes when a jackpot is split between winners.
Level: Starter. You’ll need: the lottery’s published odds, a spreadsheet.
15. Can pure random noise produce a “significant” result?
Run 20 tests on random data at the usual 0.05 cutoff and the chance of at least one false positive is about 64%. Simulate it to see for yourself, then explain why this trips up real studies.
Level: Stretch. You’ll need: Python or R.
16. Do more hours of sleep go with better quiz scores in your own survey?
Design a short survey, collect the answers, and fit a line to the data. Then write down what the line can say about the link and what it can’t say about cause. Ask your teacher about permission and privacy before you survey anyone.
Level: Starter. You’ll need: a short survey form, a spreadsheet.
17. Is home-court advantage changing over the seasons?
Pull game results for a league from its official stats page and track how often home teams win each season. Report how sure you are, since one number alone says little.
Level: Stretch. You’ll need: public game results, Python or a spreadsheet.
18. Does the Law of Large Numbers kick in as fast as you think?
Flip a virtual coin 10 times, then 100, then 10,000, and plot how the share of heads settles. Then look at how far off you can still be after many flips.
Level: Starter. You’ll need: Python or a spreadsheet.
19. Do the first digits of real data follow Benford’s Law?
Benford’s Law says about 30.1% of numbers in many real datasets start with a 1. Test it on city populations, river lengths, or prices, and find a dataset where it breaks.
Level: Stretch. You’ll need: a public dataset, Python or a spreadsheet.
20. Can you tell a real coin-flip sequence from a fake one?
People who invent flips tend to avoid long streaks, but a real 100-flip sequence has a run of five or more heads in a row about 81% of the time, and a run of five of the same side about 97% of the time. Collect fake sequences from friends, mix them with simulated ones, and see if a simple rule can sort them.
Level: Starter. You’ll need: friends, Python or a spreadsheet.
Applied math and modeling research topics for high school students
Modeling means writing equations for something real and checking whether they behave like the real thing. Calculus helps, but many of these start with a simple rule you repeat step by step.
21. Can a simple SIR model show how a disease peaks?
Split a population into susceptible, infected, and recovered groups and let the numbers move by simple rules. Change the contact rate, add vaccination, and see how the peak moves.
Level: Stretch. You’ll need: Python or a spreadsheet.
22. Does changing the starting numbers change a predator-prey cycle?
The Lotka-Volterra equations make populations of predators and prey rise and fall in cycles. Run them from different starting points and see how the size of the cycles changes.
Level: Stretch. You’ll need: Python.
23. Can a few simple rules create a traffic jam out of nowhere?
Put cars on a ring road, give each a speed rule and a small chance of braking, and run it. Jams can appear with no accident at all, and you can measure how the braking chance changes them.
Level: Stretch. You’ll need: Python.
24. Does a hot drink really cool the way Newton’s law says?
Record the temperature of a cooling cup every few minutes and fit it to Newton’s law of cooling. Then check how well the curve holds up and what might be throwing it off.
Level: Starter. You’ll need: a thermometer, a timer, a spreadsheet.
25. Do coffee filters really fall at a steady speed after a moment?
Drop stacks of coffee filters from the same height, film them, and measure how fast they fall. Then compare your results with the idea of terminal velocity.
Level: Stretch. You’ll need: coffee filters, a phone camera, a spreadsheet.
26. Do real stock prices move like a random walk?
Simulate a random walk and compare its ups and downs with daily prices from a public source. Look for where the model fits and where real prices clearly behave differently.
Level: Stretch. You’ll need: Python and public price history.
27. Does it matter whether two checkout lines merge into one?
Simulate a store with one shared line and a store with two separate lines, then compare the average waits. The result is a good example of how a small change in a rule changes a whole system.
Level: Stretch. You’ll need: Python.
28. Can a basic energy-balance model predict Earth’s temperature?
Set the sunlight Earth absorbs equal to the heat it sends back out. With a reflectivity of about 0.3, that simple model gives around 255 K, or roughly -18 degrees Celsius, which leaves room to ask what the atmosphere adds.
Level: Stretch. You’ll need: Python.
29. Does a perfect fifth really have a 3-to-2 frequency ratio?
Compare the pure 3-to-2 ratio with the one used on a piano, where each fifth is 2 raised to the power 7/12, or about 1.4983. Then work out how far off each interval drifts.
Level: Starter. You’ll need: a calculator or spreadsheet.
30. Does paying a little extra each month really save that much interest?
Build a loan table for a car or student loan and add a small extra payment. Compare the total interest and the payoff date with and without it.
Level: Starter. You’ll need: a spreadsheet.
Geometry and topology research topics for high school students
Geometry is about shape and space. Topology asks which properties survive stretching and bending, which is why a coffee mug and a doughnut count as the same object.
31. Do the angles of a triangle always add up to 180 degrees?
Yes on a flat page and no on a sphere. Draw big triangles on a globe or an orange and measure how far the angle sum goes over 180.
Level: Starter. You’ll need: a globe or orange, string, a protractor.
32. Can you measure how crinkly a coastline is?
Cover a map of a coastline with grids of different sizes and count the squares that touch it. The way the count grows gives you the fractal dimension, and the coast of Britain is a classic case at about 1.25.
Level: Stretch. You’ll need: a map image, a spreadsheet or Python.
33. Does the Koch snowflake really have an endless edge and a limited area?
Yes, and you can show it. The edge grows by a factor of 4/3 at every step and never stops, while the area settles at 8/5 of the starting triangle.
Level: Starter. You’ll need: pencil, then a spreadsheet.
34. Can you draw a street scene using only vanishing points?
Use the rules of perspective to draw a street, then compare your drawing with a photo. This is a hands-on way to meet projective geometry.
Level: Starter. You’ll need: paper, ruler, a photo for comparison.
35. Can you tell a knot from an unknot?
Knot theory asks how to prove two knots are really different. Try tricolorability, a simple coloring test that separates the trefoil knot from a plain loop, then read how knots show up in DNA.
Level: Starter. You’ll need: string, paper, colored pencils.
36. Does V – E + F always equal 2 for a solid with no holes?
Count corners, edges, and faces on cubes, pyramids, and other solids and test the pattern. Then see what happens when you try a shape with a hole, like a doughnut.
Level: Starter. You’ll need: paper models or drawings.
37. Can you cover a floor with just one shape and no gaps?
Test which regular shapes tile a plane without gaps. Then try irregular shapes of your own and find out which ones tile and which ones don’t.
Level: Starter. You’ll need: paper cutouts or GeoGebra.
38. Is a circle the shape that holds the most area for a fixed fence length?
Compare squares, rectangles, hexagons, and other shapes with the same perimeter. Then see how the area climbs as you add sides.
Level: Starter. You’ll need: a spreadsheet.
39. Can a robot arm reach every point you want?
Work out where the hand of a two-joint arm ends up from its joint angles, using only trigonometry. Plot every spot it can reach and find the places it can’t.
Level: Stretch. You’ll need: Python or GeoGebra.
40. Does a spinning 3D cube on screen really come from matrix multiplication?
Rotation, scaling, and projection are all done with matrices. Write code that spins a cube and draws it on screen, and check your rotation against hand calculations.
Level: Challenge. You’ll need: Python, some linear algebra from a free source.
Discrete math, logic and computer science research topics for high school students
Discrete math is the math of things you can count one by one: networks, codes, logic, and algorithms. It sits underneath most of computer science.
41. Can a network show who holds a friend group together?
Draw a club as a graph with people as dots and friendships as lines. Count each person’s connections and look for the people whose removal splits the group. Use invented names or get permission first.
Level: Starter. You’ll need: paper or Python.
42. Can you rank five web pages by hand like early Google did?
PageRank treats a page’s importance as something you can compute from who links to it. Build the link matrix for a tiny web of five pages and work out the ranking.
Level: Challenge. You’ll need: pencil, then Python.
43. Can a code fix its own mistakes?
A Hamming (7,4) code can find and fix any single flipped bit. Work one out by hand, then simulate noise on a message and see how often it recovers.
Level: Stretch. You’ll need: pencil, then Python.
44. Can you add two numbers with nothing but AND, OR and NOT?
Build a one-bit adder on paper from these three gates, then chain adders together for bigger numbers. Check your design against De Morgan’s laws.
Level: Starter. You’ll need: paper or a free logic simulator.
45. Can a Sudoku with very few clues still have only one solution?
Yes, and the smallest number that can work is 17, which computer searches proved in 2012. Write a solver and test which clue counts leave more than one answer.
Level: Stretch. You’ll need: Python.
46. Is the shortest route through five cities always easy to find?
With five cities there are only 12 distinct round trips to check, so brute force works. Then add cities one at a time and watch the number of routes explode.
Level: Starter. You’ll need: a spreadsheet or Python.
47. Can you explain P vs. NP in one page?
It is an open problem with a $1,000,000 prize from the Clay Mathematics Institute. You won’t solve it, but you can explain it with examples like the five-city route problem and sorting.
Level: Starter. You’ll need: library books, paper.
48. Can a toy blockchain notice when someone edits an old block?
Build a short chain in which each block holds a fingerprint of the one before it. Edit an early block and show how every later fingerprint breaks.
Level: Challenge. You’ll need: Python.
49. Can the same votes elect different winners under different rules?
Take one set of ranked ballots and count them with plurality, ranked choice, and Borda count. Then read about Arrow’s theorem to see why no rule is perfect.
Level: Starter. You’ll need: a spreadsheet.
50. Can a small program diagnose a bike problem from yes-or-no answers?
Write down rules like “if the chain skips, check the gears” and turn them into a tree of questions. Then test it on friends and fix the rules where it gets stuck.
Level: Starter. You’ll need: paper, then Python.
How do you turn a topic into a research question?
Shrink the topic until you can finish it in a few weeks. “How do ecosystems work?” is too big. “Does changing the starting numbers change a predator-prey cycle?” is a project. If you feel stuck, try the idea with the numbers 1 through 10 on paper first, because the pattern you spot there is often your question.
Then write down what would count as an answer before you start. A project that tests an idea and reports that it didn’t hold up is still real research, and our post on doing research in high school with no prior experience covers the worries beginners tend to have.
How hard is each group of topics?
| Group | Typical prerequisites | Main output | Good first tool |
| Number theory and patterns | Algebra, curiosity about proofs | A proof or a pattern with evidence | Pencil and Python |
| Probability, statistics and data | Algebra, basic probability | An analysis with uncertainty | Python or a spreadsheet |
| Applied math and modeling | Precalculus, some calculus helps | A model compared to data | Python or a spreadsheet |
| Geometry and topology | Geometry, some linear algebra | A construction or a measurement | GeoGebra |
| Discrete math and computer science | Algebra, some programming | An algorithm or a proof | Python |
If you’re heading toward a harder topic, a short program can help you get ready. Our guides to math camps, free university math summer programs, and summer math programs cover options for different schedules.
What free tools do you need for math research?
Python handles simulations, data work, and quick checks of a guess, and R is popular for statistics. A spreadsheet does more than people expect. Desmos and GeoGebra let you see a function or a construction before you try to prove anything about it, and SageMath adds number theory tools. For the write-up, a free LaTeX editor lets you typeset equations cleanly.
Tips to Make Your Math Research Project Work
1. Pick a topic where you can finish a small version
A narrow question you can answer beats a broad one you never close. Shrink it until a first draft fits in a few weeks.
2. Work small cases by hand before you code
Try the topic with small numbers on paper. The patterns you spot there give you something to test with code later.
3. Keep a log of what you tried
Write down every dead end with a sentence about why it failed. Those notes often become the strongest part of your write-up.
4. Read one real source early
A textbook chapter, a survey article, or a university lecture note gives you the vocabulary you need. It also tells you whether your question has already been answered.
5. Check your claims with a second method
If a simulation gives you a number, see whether a formula agrees. If a formula gives you a result, test it against a simulation or a small example.
6. Find someone who will argue with you
A teacher, a mentor, or a friend who asks “why is that true?” will catch gaps faster than rereading your own work. To see how selective programs judge this kind of project, read our guide on getting into a top high school research program.
Want to go deeper with a mentor?
If one of these topics keeps pulling you back, the Horizon Academic Research Program is a selective online program that pairs you one-on-one with a professor or PhD scholar. Fall, winter, and spring cohorts run 15 weeks and the summer cohort runs 10 weeks, ending in a research paper of about 20 pages in a quantitative or qualitative track. Financial aid is available, and the Winter Cohort deadline is November 22, 2026.
What you might still wonder before starting a math research project
No. Plenty of topics here need only algebra, and a few modeling topics are easier if you have seen calculus. You can learn the bit you need as you go.
Yes. Plenty of people start alone with a library book and a laptop. A mentor mostly helps you choose a better question and catch mistakes sooner.
No. A strong project can repeat a known result in your own way, test a model on new data, or explain a hard idea with original examples. If you want help narrowing to a question you can finish, the Horizon Academic Research Program pairs you one-on-one with a professor or PhD scholar and ends in a 20-page paper, with financial aid available.
No. Plenty of good projects end as a clear paper or a talk. Our post on whether a high school research paper needs to be published explains how to think about it.
Not always. Many number theory, geometry, and logic topics work on paper. Programming helps for simulations and data, and Python is a friendly place to start.
Yes, if the goal is to study it. You can explain the problem, work through related examples, and read about partial results, but a full solution is not a realistic target.
Yes. A shared topic can work well if each person owns a clear piece, like the simulation, the proof, or the write-up. Agree on who does what before you begin.
Yes, if the question is narrow. About 15 weeks is enough to go from a topic to a full paper, and our list of online math research programs shows what that looks like in practice.
Image source: Horizon Academic Research Program




