Applied probability, in public.
The Probability Lab is a small set of transparent statistical models for public-interest questions. We publish the methodology in full, the predictions before the event, and the accuracy data after. The goal is to make applied probability legible — not to compete with bookmakers.
Current artefacts
v1 · Live · 104 matches modelled
FIFA World Cup 2026 — Statistical Match Predictions
Match-by-match probabilities for all 104 fixtures of the 2026 FIFA World Cup, generated by two transparent statistical models — Dixon-Coles Poisson and an Elo baseline — and updated as results come in. Methodology published in full; accuracy tracked publicly from kick-off.
See the predictions→v1 · Live · Projected table + head-to-heads
Premier League 2026/27 — Season & Match Predictions
A projected final table with title, top-four and relegation probabilities, plus match-by-match predictions — from the same Dixon-Coles Poisson + Elo engine as the World Cup model, calibrated to Premier League scoring. The method and input ratings are published in full.
See the predictions→v1 · Forecast · 9 candidates tracked
Kenya 2027 — General Election Forecast
A fundamentals-plus-polls model for the August 2027 Kenyan presidential election, in the tradition of FiveThirtyEight and The Economist — adapted for the Kenyan 50%+25%-in-24-counties first-round rule and the post-2022 coalition landscape. Every coefficient, every assumption, and every source is published.
See the forecast→How this differs from a betting product
Probabilities and methodology — never odds, stakes, or tips.
Probabilities, not odds.
We report P(outcome) numbers that sum to one. We never convert these to bookmaker-style odds, payouts, or stakes. We do not link to any betting venue.
Models, not algorithms.
Every input, every equation, every parameter is published. No black boxes. If you don't understand why we predicted what we did, that's a bug — write to us.
Calibration, not bravado.
We track our predictions against actual results in public. A confident model that is wrong is worse than an uncertain one that says so. The accuracy page tells you which we are.
Related
Nearby on LeadAfrik.
The Probability Lab is one wing of a wider platform.
Models
Economic Models — all 15
Interactive AD-AS, IS-LM, Solow, Cournot, Phillips, Mundell-Fleming, Laffer, and more.
Data
Data Store
Sourced CSV and JSON datasets — the inputs to a lot of LeadAfrik's quantitative work.
Standards
Editorial standards
The integrity rules that govern every model and analysis we publish.
Analysis
Analysis & essays
Long-form notes on Kenya's economy and African markets — including the model launch announcements.