AI Prediction Methodology

This page explains how predictions are generated, published, and tracked for transparency.

MVP Scope

• League: NBA.
• Markets: match winner (moneyline), spread, and total points over/under.

Data Inputs

• Bookmaker odds snapshots from configured providers.
• Match context signals (form, rest days, opponent strength, and injuries).
• Historical outcomes for calibration and drift checks.

Model Output

• Estimated probability by market selection.
• edge_percent as model probability minus implied market probability.
• Confidence tier (A/B/C) and unit stake suggestion.

Public Integrity

• Predictions are timestamped and stored before settlement.
• Settled picks remain in the public log with immutable historical records.
• Dashboard KPIs include hit rate, ROI, profit units, and drawdown.

Back to tracker