NFL · NBA · MLB · NHL
Picks from models that
show their record.
Every pick on Shark Snip comes from a model you can backtest, audit, and compare against the closing line. No black boxes, no tout slips. Free to start. Not a sportsbook.
$0
Free, forever. Build models, backtest, look at last week's picks.
$10 / wk
Live picks from every house model. Cancel any Sunday.
$30 / mo
Everything: 25 AI Sharp calls, DFS, props, API, VIP Discord.
Or pay $1 for a single Claude-powered Sharp call. No card to start. No surprise charges. Full pricing →
The Builder · free workbench
Build your own model in the browser.
Drag blocks, pick a target, train on 5+ years of play-by-play, tracking, and market data — all client-side with TF.js. Backtest with walk-forward CV, then watch your picks land on the same board as the house models. The free building tier never sends your training data to Shark Snip.
Real Data, Open Methods
Every dot here is a real play, a real shot, a real backtest.
No stock images. No "AI-generated charts." These render directly from local CSVs and the same pipeline that feeds the Builder, Picks, and Consensus.
NFL EPA + Success Atlas
35,050 plays · 2024 NFL play-by-play · pass/run efficiency, weekly EPA drift, top team EPA per play.
NBA Shot Yield Map
Every 2-pt and 3-pt event · 2024 NBA play-by-play · shot distance vs points, team efficiency, quarter shape.
Schedule Strength Ridge
Implied team totals × week from the schedule-strength dataset. The look-ahead view your slate models use.
Backtest Error Field
Projection vs actual for every published model. The track record nobody can fake.
Picks · across four sports
Daily model picks for NFL, NBA, MLB, and NHL — spreads, totals, props, and DFS. Each one links to the model behind it and its closing-line-value record.
Builder · the workbench
Train TF.js models in your browser on years of play-by-play, tracking, and market data. Backtest with walk-forward CV. The free building tier never sends your training data to Shark Snip.
Consensus · the wisdom of crowds
The weighted vote of every opted-in model. Brier-aware weighting, 5%-per-creator cap, fork dedup. Free tier sees the top pick; Premium sees the full feed. Methodology is public.