math modeling

math modeling competitions are the most underrated competitions for builders. unlike contest math (AMC/AIME/USAMO), math modeling gives you a real-world problem and asks you to build a mathematical model to solve it. no memorization, no tricks — just applied thinking. the skills transfer directly to research, product building, and engineering work.

if you like building things and you like math, this is your competition.

why math modeling is special

contest math asks: "solve this problem that has a known elegant solution." math modeling asks: "here's a messy real-world situation. build a model. make predictions. justify your approach."

the second one is closer to what engineers and researchers actually do. the skills you develop — framing problems, choosing assumptions, validating models, writing technical reports — are the same skills you use in research papers, product analytics, and engineering internships.

also: math modeling competitions are team-based (usually 3-4 people), which means you're developing collaboration skills. and the results are strong on a resume — a top-10% finish in MCM/ICM as a high schooler is a real accomplishment.

the competitions

HiMCM (High School Mathematical Contest in Modeling)

format: 36 hours, teams of up to 4, November
organizer: COMAP
what we did: fire/evacuation scenario with Monte Carlo simulation and spectral bisection
team: three teammates and me
my take: this is the entry point. massively underrated. 36 hours is enough time to build a real model without the sleep-deprivation of a hackathon. the problems are genuinely interesting — real-world scenarios with no single right answer.

MCM/ICM (Mathematical Contest in Modeling / Interdisciplinary Contest in Modeling)

format: 4 days (Jan/Feb), teams of up to 3
organizer: COMAP
result: Meritorious (top 10%)
what we did: spectral bisection, cellular automata, graph traversal
my take: this is the college version, but high schoolers can compete. placing Meritorious (top 10%) as high schoolers competing against college teams was a genuine accomplishment. 4 days is enough time to go deep — our paper was substantial.
2026 dates: January 29 - February 2

M3 Challenge (MathWorks Math Modeling Challenge)

format: 14 hours, one weekend (Feb/Mar)
organizer: SIAM (Society for Industrial and Applied Mathematics)
result: 143/770 (top 19.8%), qualified for second round
my take: free to enter, $100,000+ in total prizes. the 14-hour constraint is brutal but teaches you to scope aggressively — the same skill you need for hackathons. high school juniors and seniors eligible.

MTFC (Modeling the Future Challenge)

result: semifinalist (2025-26)
what we did: equitable bus routing
my take: the problems have a social-impact angle, which makes the modeling more interesting. "optimize bus routing for equity" is the kind of problem that doesn't have a clean mathematical answer — you have to make value judgments and defend them.

IMMC (International Mathematical Modeling Challenge)

format: 6-day sprint (March)
international pool of competitors
similar flavor to HiMCM/MCM but with more time and an international dimension

techniques worth learning

these come up repeatedly across modeling competitions:

Monte Carlo simulation — randomly sample scenarios to estimate probabilities and outcomes. we used this for the fire evacuation model. versatile, intuitive, and works when analytical solutions are intractable.
spectral methods — eigenvalue-based approaches for graph problems. spectral bisection was key in both our HiMCM and MCM papers.
cellular automata — grid-based simulations where each cell follows local rules. emergent behavior from simple rules. great for modeling physical processes (fire spread, traffic, epidemics).
graph traversal and network analysis — most real-world problems can be modeled as graphs. shortest paths, flow optimization, community detection.
optimization — linear programming, integer programming, gradient descent. the core of most modeling problems is "maximize/minimize something subject to constraints."
sensitivity analysis — how does your answer change when you change your assumptions? judges care about this a lot. it shows you understand the limitations of your model.
FEM simulation — finite element methods for physics-based modeling. more advanced, but powerful for structural/thermal/fluid problems.

the modeling mindset extends beyond competitions. problems I've worked on:

elevator scheduling optimization
drone routing for delivery
disease testing strategies
Olympics scoring system analysis
irrigation optimization
sheep population dynamics
trading algorithm design
firefighter sweeping route optimization

each of these is a self-contained modeling exercise that could become a research paper, a section of a competition submission, or a portfolio piece for job applications.

practical advice

practice with past problems. COMAP archives past MCM/ICM and HiMCM problems. SIAM archives past M3 Challenge problems. work through them under time constraints.
learn LaTeX. your submission is a technical paper. LaTeX (via Overleaf) is the standard for mathematical typesetting. judges notice well-formatted papers.
master one simulation tool. Python with NumPy/SciPy/matplotlib is the default. MATLAB works too (and MathWorks sponsors M3 Challenge). the tool doesn't matter — fluency does.
write clearly. the paper matters as much as the model. judges read dozens of papers in a day. clear writing with good figures stands out. explain your assumptions, justify your choices, acknowledge limitations.
build your team deliberately. you want: someone strong at mathematical formulation, someone strong at coding/simulation, and someone strong at writing/visualization. overlap is fine but coverage matters.
start with HiMCM. it's the most accessible, the time pressure is manageable, and it teaches you the modeling-paper workflow. then move to MCM/ICM.

math modeling is where mathematical thinking meets building things. the competitions are the structured version, but the modeling mindset applies everywhere — any time you're trying to understand a system, make predictions, or optimize something, you're modeling.

see competitions-hackathons for the full competition landscape and publishing-research for turning modeling work into papers.