Full transparency
The track record, as logged
Every prediction is timestamped server-side before kickoff and graded after the final buzzer — wins and losses both stay in the record. This page renders straight from that log via the public API. loading…
Settled value bets
—
ROI
—
profit / staked
Profit
—
units (1u = 1% bankroll)
Hit rate
—
Graded predictions
—
all sides scored
Val. accuracy
—
holdout window
ROI by edge bucket
are bigger claimed edges real edges?
positive ROI
negative ROI
bucket = model edge over de-vigged market probability
Crowd engine self-score
the crowd layer graded by the same Brier standard as the modelGames scored
—
crowd 1x2 vs realized results
Crowd Brier
—
weighted crowd shares
Market Brier
—
de-vigged closing line
Calibration
when the model says 60%, does it win ~60%?| Predicted bucket | n | Avg predicted | Observed win rate | Gap |
|---|---|---|---|---|
| Loading… | ||||
Recent settled picks
most recent first| Matchup | Side | Odds | Model | Result | P/L |
|---|---|---|---|---|---|
| Loading… | |||||
Model quality: Brier — vs market baseline — (lower is better) · backend — · trained —. Free plan shows the 10 most recent picks; Premium unlocks full history with stakes.