Research mode. Forecasts are available for analysis. Staking remains unavailable until the required validation gates pass. Why? ▸Hide ▾
This model generation has not yet passed validation against the closing market. Probabilities and the market comparison are shown for analysis; no stake is recommended for any fixture.
Live evaluation (C6): checking
Upcoming verified fixtures
Each fixture shows the model's forecast beside the bookmaker's fair price. Open one for the full evidence.
Soft coverage — higher uncertainty while provider evidence is limited.
Football forecasts, evidence and market context — clearly explained
Every forecast shows the evidence behind it, what is missing, and the bookmaker's price beside it. Missing evidence is listed, never filled in.
Data integrity. Configured providers are reconciled, and every gap is listed with where it came from.
Model governance. Artifact and validation status appear only when the backend confirms them.
Market comparison. Model probabilities sit beside the bookmaker's de-vigged price, with no stake while validation is pending.
Model status
Model generation
Validation
Availability
Leagues covered
Live evaluation (C6)
Explore a manual matchupHypothetical: not a scheduled fixture, so no price or stake applies.
How SabiScore works
Five evidence-gated steps from raw data to an explained verdict.
Collect evidence
Technical detail ▸Hide detail ▾
Configured sources are queried for fixtures, lineups, injuries, odds, and standings when available.
Validate & reconcile
Technical detail ▸Hide detail ▾
Fixture identity resolved across providers. Conflicts surfaced explicitly — never silently merged.
Run the active model
Technical detail ▸Hide detail ▾
The backend reports the served artifact, feature schema, and calibration state; missing certification stays unavailable.
Compare to market
Technical detail ▸Hide detail ▾
Fair market probabilities de-vigged from bookmaker odds. Edge = model probability minus fair market probability.
Surface the result
Technical detail ▸Hide detail ▾
Six evidence-gated verdict levels with a plain-English rationale and explicit data-gap report.
Understanding verdicts
Every match card leads with one decision state. The verdict tier beneath it describes the evidence.
- Play
- A stake is permitted: a certified forecast, complete evidence, and a price with positive expected return. Not available while this generation is in research mode.
- Pass
- Evaluated, and the price offers no value. No stake.
- Withheld
- Not evaluable: a blocking gap, such as an uncertified model or a missing price. No stake.
Each verdict is an evidence gate, not a confidence dial. Stronger labels require more independent sources.
Partial Data
Critical evidence is missing, so the fixture is not evaluated.
No Bet
Evidence is complete, and the price offers no value.
Hold
Model and market are close. Evidence is thin; watch for updates.
Speculative
A tentative signal for research only. No stake is permitted.
Actionable
Can carry a stake only when every evidence and certification gate is open.
High Conviction
Strong independent evidence. An unvalidated generation stays research-only.
No verdict is a certain outcome. Stronger evidence reduces uncertainty — it does not eliminate it.
Supported competitions
Listed coverage is subject to the active artifact and league policy reported by the backend.
Premier League
ENG
La Liga
ESP
Bundesliga
GER
Serie A
ITA
Ligue 1
FRA
Eredivisie
NED
Champions League
Europe