How our models and AI summaries work, and their limits.
Version 2026.09.1 · effective 2026-09-11
What our outputs are
Model probability, confidence, value, risk and grade are statistical estimates produced from historical results, live market prices and contextual data. They describe our model's view — not the truth about a future event.
What they are not
They are not guarantees, insider information, fixed odds, or professional advice. A high-confidence estimate can lose. A well-calibrated model is still wrong a predictable share of the time.
AI-generated text
Some explanations and summaries are generated automatically. They can be incomplete or wrong, and they are not reviewed line by line by a human before display.
Data limits
Estimates are only as good as the data behind them. Feeds can be delayed, partial or withdrawn, and injuries, lineups and weather change fast. When data is missing we say so instead of filling the gap.
Calibration and audit
We retain an internal audit trail of generated selections, model version and inputs so accuracy can be measured over time. Published calibration figures cover settled records only.
Stratos.Live is an independent sports analytics platform. We are not a sportsbook and we never accept, hold, settle or place wagers.