How Parlae grades every pick
Our +EV model went live on June 13, 2026. From day one, every pick the model surfaces is logged and graded against real settled results. This page leads with the rules — because that's the trust asset today; results accrue as bets settle.
No backfill. No fabrication. Ever.
- Picks are logged when surfaced — not after the fact.
Every +EV row is captured at the moment the model publishes it, with the exact price, line, book, fair odds, and timestamp. We never edit a pick after the result is known.
- Grading uses real settled results.
Outcomes come from official box scores and event results, never from estimates. Pushes are pushes; voids are voids.
- CLV is measured against the closing line.
For every graded pick we record the closing fair price and report Closing Line Value — the leading indicator that the edge was real, independent of one-off variance.
- Small samples are shown as small samples.
We won't dress up an N of 12 as a "track record." Until results accumulate, you'll see the sample size honestly, even when that number is zero.
- Sport-by-sport breakdowns — no cherry-picking.
ROI, units, win rate and CLV are reported by sport so a hot streak in one league can't hide a cold streak in another.
- Anti-backfill pledge.
If you ever spot a graded pick that wasn't first published live in the +EV feed, that's a bug — report it and we'll fix it publicly.
Engine CLV — live, accruing
tracking since —Closing-line value of Parlae's own picks. We lead with average CLV and sample size — not a binary "100%" — because dispersion is what tells a sharp the edge is real. Never backfilled.
Confidence intervals on mean CLV tighten meaningfully around N≈200. Until then: accruing.
Loading track record…
Loading full +EV ledger…
Results so far
tracking since —Loading…
Public leaderboard
opt-in · aggregates onlyOnly users who flip profile.is_public appear here. We show aggregate stats only — never individual bets, never PII.
Loading…