Verified track record
Every pick. Timestamped. Graded. Public.
Each AI consensus is written to an immutable Postgres record the moment it's generated — before kickoff — then graded automatically once the game is final. Nothing here is cherry-picked, back-fit, or fabricated.
Graded record
0-0
0 graded • 3 pending
Win rate
—
Awaiting graded results
Return (units)
0.00u
ROI — • flat 1u staking
Avg. confidence
51.7%
Avg. edge -12.2%
Profit curve
Cumulative units, flat 1u stakesHow a flat 1-unit bettor following every graded pick would have tracked over time.
Model calibration
Stated vs. actualWhen we say a pick is 70% likely, does it win ~70% of the time? Closer bars mean better-calibrated confidence.
Closing Line Value (CLV)
CLV compares the price we locked in when the pick was made against the market's closing line. Consistently beating the close is the single strongest statistical proof of a real edge — it shows the model finds value before the market corrects, independent of short-term win/loss variance.
No closing lines captured yet
CLV is recorded automatically: the pick-time price is stored when each AI consensus is generated, and the closing line is captured when the game is graded. Figures appear here once picks made before kickoff have been settled.
By confidence tier
By league
By market
How this record is kept honest
- Every AI consensus is written to the database the instant it's generated — with a server timestamp that predates kickoff.
- Results are graded automatically from final scores; we never hand-edit a graded outcome.
- ROI uses flat 1-unit staking at the recorded book price (standard -110 assumed where a book price wasn't captured).
- Win rate counts decided picks only; pushes are excluded from win rate and count as 0 units.
Full prediction ledger
Boston Celtics
moneyline
Manchester City
moneyline
Boston Celtics
moneyline