What this site covers
Everything Barry needs in one place. Each module is a stand-alone page, but they're also stitched together as weeks in the six-month plan. Pick where to start based on what you need today.
About the contest
Rules, dates, team composition, judging tiers, and what "Outstanding" actually means.
Anatomy of a winning paper
Summary sheet, assumptions, model, sensitivity, letter — section by section with judge commentary.
Modeling techniques
The dozen methods that keep showing up: AHP, TOPSIS, regression, ODEs, Monte Carlo, ARIMA, optimization.
Python toolkit
NumPy, SciPy, pandas, matplotlib, statsmodels, networkx, SimPy — snippets you can paste and adapt.
Past problems 2025–2011
Every problem with a full restatement, a modeling outline, key formulas, and pitfalls.
Original mock problems
Five problems written in HiMCM style with solution sketches, rubrics, and self-grading checklists.
Rubric self-check
Interactive 16-row scorecard that mirrors how COMAP judges actually read papers. Weighted total maps to Outstanding / Finalist / Meritorious bands.
Outstanding paper anatomy
Reverse-engineering the ~1% of HiMCM papers that win Outstanding: summary-sheet dissection, assumption quality (✗/✓), model coherence, real sensitivity, 27-item Outstanding-paper checklist.
Worked sample paper · 2025-A
Full reference paper for "Emergency Evacuation Sweeps" — summary sheet, 9 numbered equations, NetworkX + PuLP MILP code, Monte Carlo sensitivity tables, letter to the LEPC, AI-use appendix.
Worked sample paper · 2024-A
Full reference paper for "To Play or Not to Play" — summary sheet, 10 numbered equations, runnable Python, sensitivity tables, non-technical letter, AI-use appendix.
Worked sample paper · 2024-B
Full reference paper for "Environmental Impact of High-Powered Computing" — summary sheet, 10 numbered equations, per-region logistic + 6-state Markov mix Python, Sobol sensitivity, embodied + water extension, AI-use appendix.
Worked sample paper · 2023-A
Full reference paper for "Dandelions: Friend? Foe? Both? Neither?" — summary sheet, 9 numbered equations, Fisher–KPP PDE + EWM impact-factor Python, Sobol sensitivity tables, AI-use appendix.
Worked sample paper · 2023-B
Full reference paper for "Charging Ahead with E-buses" — summary sheet, 7 numbered equations, life-cycle CO₂ + per-cohort NPV + PuLP MILP roadmap code for three cities, Monte Carlo sensitivity, AI-use appendix.
Worked sample paper · 2022-A
Full reference paper for "The Need for Bees" — 9 equations, KMB-extended 5-compartment ODE in scipy.solve_ivp, Sobol sensitivity, hive-sizing for a 20-acre parcel under stress.
Worked sample paper · 2022-B
Full reference paper for "CO₂ and Global Warming" — 10 equations, six CO₂ models fit to Mauna Loa + GISTEMP, log T–C law with IPCC climate sensitivity range, ARIMA uncertainty band.
2021 problem set
Storing the Sun (utility-scale solar + storage sizing) and Tackling the Drought (Lake Mead reservoir operating policy).
2020 problem set
The Best Summer Job (MCDM warm-up: AHP + TOPSIS) and Funding Biodiversity Conservation (knapsack over Florida threatened plants).
2019 problem set
Charge! (US EV charging-network forecast + facility location) and Bottle Battles (campus single-use-bottle policy via compartmental flows).
2018 problem set
Roller Coaster (objective ranking on 300+ coasters with EWM/TOPSIS) and Cozy Smart House (thermal RC model + LP for HVAC scheduling).
2017 problem set
Drone Clusters as Sky Light Displays (swarm choreography + Hungarian assignment) and Ski Slope (DEM-based resort layout with M/M/c queueing).
2016 problem set
Swim, Bike, and Run (wave-start triathlon scheduling for 2,000 athletes under 5.5h closure) and Shop and Ship (continental warehouse set-cover with sales-tax overlay).
2015 problem set
Preventing Road Rage (Nagel–Schreckenberg CA + behavioral merge rule for late vs early merge) and City Crime and Safety (3-pillar composite index with AHP/EWM + DBSCAN hotspots).
2014 problem set
Unloading Commuter Trains (1D platform sim + Fruin stair capacity + k-center placement) and The Next Plague? (SEIR + severity decision table + LP resource allocation).
2013 problem set
Emergency Medical Response (MCLP ambulance placement, 8-min coverage, fleet degradation) and Bank Service Problem (M/G/c queueing with wait < 2 min and queue ≤ 2 targets).
2012 problem set
American Elk (stage-structured Lefkovitch matrix + stochastic PVA for Great Smoky Mountains) and How Much Gas Should I Buy This Week? (EIA price forecast + deadband decision rule).
2011 problem set
No More Space Shuttles (10-year ISS resupply MIP over Soyuz/Progress/ATV/HTV/Dragon/Cygnus) and Search and Find (nighttime penlight search with sweep-width model + parallel tracks).
Six-month plan
Week-by-week schedule from June 2026 through the contest weekend in November.
Resources & references
Books, courses, official PDFs, datasets, glossary of contest jargon and modeling vocabulary.
Suggested reading order
- Get oriented. Read About the contest end-to-end and skim two recent problems on the Past problems page so you have a sense of what 25 pages of analysis actually look like.
- Learn the paper. The structure of a HiMCM solution is fixed. Go through Anatomy of a winning paper next so every later technique has a place to land.
- Build the toolkit. Work through the modeling catalogue on Techniques and the matching code in the Python toolkit. Do small drills as you go.
- Practice on real problems. Pick a past problem and write a 5-page partial solution in a weekend. Compare against the judges' commentary linked from each problem page.
- Run a full mock. Once a month between July and October, do one mock problem under contest-like conditions (no outside help, 4 days, write the full paper).
- Final push. The six-month plan ramps to a dress-rehearsal in late October and the live contest Nov 4–17.
A few things judges actually reward
From COMAP's published judges' commentaries on Finalist and Outstanding papers:
- Simple models that you fully understand beat sophisticated models you bolt on from the internet. Pick math you can defend out loud.
- Assumptions tied to the model. List each assumption with its justification, then point to where in the model it's used.
- Sensitivity analysis. Show which parameters matter — by simulation, by partial derivatives, or by re-running with perturbed inputs.
- Strengths and limitations. Honest weaknesses signal modeling maturity. Hiding them never works.
- A real Executive Summary. Written last, after the paper is done, summarizing the actual results — not just the methodology.
- A non-technical letter that doesn't just paraphrase the executive summary. Different audience, different voice.
Each of these is unpacked in detail on the paper page.