How SportsLab works
SportsLab is the sports-forecasting front-end on top of the POE1 prediction engine. The same engine powers finance, energy, health and other domains — SportsLab adds the sports vocabulary, market structure, and settlement loop.
The pipeline
Five stages, evidence-first.
- 1. You ask a sports question.
Pick a mode (team vs team, tournament winner, field ranking, etc.), provide context, and submit. SportsLab routes the question to POE1's sports-forecasting domain pack.
- 2. Evidence retrieval.
The engine pulls relevant historical cases, form clusters, head-to-head outcomes, and approved domain laws from indexed sports sources. Nothing is invented — missing evidence is flagged.
- 3. Multi-hypothesis prediction.
Multiple outcomes are scored in parallel, then weighted by evidence strength, contradiction, recency, and historical stability of the engine's calibration in that segment.
- 4. Honest, calibrated output.
You get a probability with confidence band, contributing evidence, and known contradictions — or an explicit insufficient-evidence response. The engine never fakes certainty.
- 5. Settlement and learning.
When the real-world outcome resolves, it flows back into the engine. Calibration scores update; future forecasts in similar segments tighten.
Operating principles
Every output ships with a confidence band. We'd rather say 'we don't know' than guess.
The engine cites the evidence it used. If retrieval came up thin, the response says so.
SportsLab is a thinking tool, not a sportsbook. It does not accept stakes, hold funds, or route money.
Real outcomes update the model. Stale calibration is treated as a data-quality bug, not a feature.
- Decision-support only
- Not betting advice
- No guaranteed outcomes
- No wager placement
- Uncertainty-first outputs
- Insufficient-evidence warnings when data is thin
SportsLab does not accept wagers, hold funds, or place bets on your behalf. All outputs are probabilistic estimates. Past calibration does not guarantee future outcomes. 18+ only.