AI sports analytics. 11 independent models. 308,000+ predictions and growing logged. Every selection published — wins and losses.
We publish every prediction — right and wrong — so you can verify our track record. Tracking predictions since February 2026 against real odds from live bookmakers.
| Sport | Market | Predictions tracked | Strike rate | ROI | Status |
|---|---|---|---|---|---|
| Data updating… | |||||
Data current as of March 2026. Rows show the best-performing model per sport.
All figures are based on simulated paper trading — not real-money wagering. Each prediction is recorded with a flat notional stake of AUD $100 at the best available Australian bookmaker odds at prediction time. ROI is calculated as total simulated profit or loss divided by total notional amount staked. These figures do not account for bookmaker account limitations, odds movement between prediction and placement, market liquidity, or other real-world execution factors. Actual results could differ materially from the reported figures. Each row shows the best-performing model for that sport — other models may have produced different results. Paper trading since February 2026.
We build statistical models that analyse sports data. Here is exactly what they do, what they cannot do, and why we tell you both.
Our 11 independent models ingest historical match data, team performance statistics, and real-time bookmaker odds across 15 sports. Each model uses a different statistical methodology to estimate outcome probabilities. When a model's probability diverges significantly from the market's implied probability, we flag it and publish the analysis.
Every model runs independently -- no shared predictions, no committee decisions. We track each model's accuracy by sport and market type over time. The model with the strongest verified track record in a given sport earns the right to generate subscriber alerts. Models that underperform are demoted automatically. The data decides, not us.
Models are retrained and recalibrated regularly as new data accumulates. Calibration parameters are updated weekly. New statistical approaches are tested in shadow mode -- logging predictions without generating alerts -- until they demonstrate sustained accuracy against real market odds.
Each model uses a different statistical approach — running independently, with no shared predictions, using real-time odds from multiple bookmakers.
Every market is scanned for statistical divergence. An alert is generated only when a model's calculated probability exceeds the bookmaker's implied probability by at least 5%. Data, not opinion.
Each sport and market combination is tracked across all 11 models. Only the most accurate model generates your alerts — and it must earn that position with verified results.
Model alerts are published the moment a divergence is identified — with the probability, the odds, and the data behind it. Premium subscribers also see the consensus score: how many of the 11 models agree.
Follow the public channel to see real model analysis in real time before subscribing. No credit card. No commitment.
We're building something different — rigorous, transparent AI sports analytics with a verified track record. Register your interest to be among the first to access gpunt when we launch Foundation Subscriptions.