AI Picks Track Record
Every AI pick is graded the moment its game goes final and written to an immutable ledger — the recommended side, the confidence, the odds it was made at, and the result. Nothing is edited after the fact. Every number below is computed live from that ledger, wins and losses alike.
Results assume a flat 1-unit stake on every “strong” or “lean” pick at the recommended side's money line. Picks the AI graded “no-pick” (no real edge) are never counted. For informational purposes only — not betting advice. 21+.
Confidence calibration
Picks grouped by the confidence the AI assigned. A well-calibrated model's dots hug the dashed diagonal: picks it rates 70% should win about 70% of the time. Larger dots = more settled picks in that band.
The calibration chart appears once picks start settling.
By league
No settled picks yet.
Recent results
Graded picks appear here as games go final.
Methodology
- Each pick is snapshotted at the odds it was made at and graded against the final score. The row is append-only and never edited after settlement.
- Profit uses a flat 1-unit stake. A winning +150 pick returns +1.50u; a winning -200 pick returns +0.50u; a loss is -1u; a push is 0.
- Only disciplined bets count. Picks the model grades “no-pick” (no real edge vs. the market) are excluded entirely, so ROI reflects actionable picks only.
- Win rate excludes pushes. ROI is net units divided by the number of bets with money staked.