Sharp-anchored edges.
De-vigged every build. Graded in public.

A composite 0 to 100 confidence number for each card. It blends the model's projection vs sharp fair, the edge over the book, historical hit rate for that prop and side, and kill signals from the playbook. See Methodology for the exact recipe.

Tiers derived from Alpha Score + kill signals. PRESS and STRONG are the plays we're willing to size; LEAN is directional but capped; WATCH is a signal to track, not press. The Slate reaches PRESS and STRONG first. Game markets need an external sharp market to confirm the play; player props need the model's edge to clear a hard confidence bar before they press. On a thin board the Slate ships its best available calls at LEAN sizing instead of going dark.

ESPN scoreboard for schedule + DK moneyline/spread/total odds. Kalshi and Polymarket for sharp market anchors on ML/totals. When The Odds API key is wired, PrizePicks + book consensus for player-prop lines. The sharp anchor pulled per card is stamped on the card detail page.

Automatically, from the ESPN box score the morning after. Grader writes win / loss / push / pending to data/grades/. Cumulative record is public on Scorecard. No cherry-picking - every published card is graded, pending losses posted alongside the wins.

The Board is the browsing surface: every playable card, filterable and sortable. The Slateis the product: the small deterministic set of plays we'd actually press tonight, Kelly-sized, capped at 8 units of exposure.

Read the Playbook once. Five steps: Slate → set bankroll → hunt Board → tail cards → come back tomorrow. After the first read you already know the loop.

Every card the pipeline produces is publishable and gets graded. We surface WATCH tier cards even though we wouldn't bet them, because filtering out losses would rig the track record. The Slate curates from that pool; the Board shows the full pool.

Scores refresh site-wide every 60 seconds - on the board tiles, the Kickoff Timeline, card detail, and each tail row on /me. Tailed cards get a ▲ LEADING (green) or ▼ TRAILING (red) chip that answers “would this bet win if the game ended now?” The top of /me shows an aggregate rollup - IN PLAY · N · $X at stake · Leading $A · Trailing $B - so you can watch multiple bets at once without opening card detail.

Yes. Every filter on /board, /edges, and /movers lives in the URL - ?sport=MLB&tier=STRONG restores exactly that view when a friend opens the link. Hit ⌘K then ★ to save your own view names for one-tap recall from the command palette. Managed on /saved-views.

No. Bet Data Labs is an analytics tool. Bets are your decision. Kelly recommendations are a mathematical framework, not a promise. Play only what you can afford to lose.