/ PROOF

We don't force picks where the data says we lack edge.

Every selectivity gate below cites its sample. Every CLV figure comes from the live picks table. If a sport×market combination doesn't clear the bar, the model stays silent — that's how we keep the slate honest.

Currently paused buckets

No buckets paused. Every sport/market combination is publishing today.

How to read this

1. Hit rate is the percentage of graded picks that won. Looks great in marketing copy but lies in small samples.
2. ROI is profit divided by amount risked. Tells you whether the past N picks made money, but variance dominates short horizons.
3. CLV (closing-line value)is the only metric that survives variance. If the model bought a price the market later tightened against, that's an edge — regardless of whether that specific pick won.
4. The gates aboveare how the model enforces discipline. When the data on a bucket says we don't have edge, we don't publish picks there. No forced bets to keep the slate looking busy.

The raw data

• → /api/model/selectivity — gate matrix JSON
• → /api/model/clv — 7d/30d/90d CLV by sport+market
• → /record — live append-only ledger of every pick

Pre-gate baseline (30d, $25 flat staking): 530 picks · 67.3% hit · -2.4% ROI. Gates reviewed monthly. Entertainment only. 21+. 1-800-GAMBLER.