Personal data product

NCAA 2026 Bracket Model

Krystian built an interactive bracket model from public basketball data, evaluated MLX models, a production ensemble, and 10,000 Monte Carlo simulations.

Personal Labs Shipped 2026
  1. Source data
  2. Matchup features
  3. Model ensemble
  4. Simulation + UI
End-to-end project path from public basketball data through evaluated predictions and a decision-support interface.

Interactive product

Explore the full model room.

Test bracket paths, title odds, upset scenarios, model notes, and all 68 teams in the complete data product.

Open full interactive model
NCAA 2026 model room showing Michigan over Florida in the projected title game, model metrics, and the bracket dashboard
A captured view of the shipped interface. Controls are available in the full interactive model.

Exact role

Independent builder

Current status

Shipped

2026

Problem

What needed to change

I wanted a practical office-pool decision tool that went beyond gut picks by combining public basketball data, modeling, simulation, and a readable product interface.

Users

Who the system serves

Bracket-pool participants and people reviewing the model, simulation, and interface.

Team context

Where I fit

Independent build with end-to-end ownership from source collection and cleanup through modeling, simulation, and the interactive interface.

Important technical decision

Compare model approaches, use an ensemble for production picks, and expose confidence and bracket paths instead of a bare prediction list.

Tradeoff or limitation

Validation accuracy and simulation output support comparison; they do not guarantee tournament outcomes.

Implementation

What I personally owned

  • Owned the full workflow from data sourcing and cleanup through modeling, simulation, evaluation, and UI packaging.
  • Built the static interactive interface around the model outputs rather than leaving them in notebooks.

Constraints

Hard parts that shaped the work

  • Collected and normalized data from NCAA NET, AP Top 25, Sports Reference, Bracket Matrix, official bracket, schedules, logs, boxscores, and historical tournament rows.
  • Compared MLX logistic baselines, priors, a tree benchmark, and a production ensemble.
  • Resolved the full field and turned probabilistic model outputs into readable bracket decisions.

Shipped result

What exists now

  • Office-pool decision tool and playful portfolio UI.
  • 10,000 Monte Carlo simulations.
  • Bracket paths, title odds, upset boards, model notes, and team search.

Result

Verified result and current state

  • The shipped lab makes the model inputs, evaluation result, simulations, bracket paths, and confidence outputs visible in one interface.
10,000 Monte Carlo simulations.68 of 68 field teams resolved.Production ensemble reached 68.5% validation accuracy.

Technical scope

Tools used

PythonPandasBeautifulSoupMLXModel evaluationMonte Carlo simulationStatic HTMLJavaScriptData visualization