Fair Line Sports
A live subscription SaaS for +EV sports betting — every prop, game line, and DFS pick devigged against sharp consensus and ranked by edge, refreshed on a tipoff-aware cycle.
The problem
Sportsbook odds carry hidden margin. A bettor comparing lines by eye cannot see the true price of a bet — or which book is currently mispriced against the sharp market.
The approach
A compute pipeline devigs every line against a sharpness-weighted consensus of reference books (weights fit from historical accuracy), blends in calibrated projection models, and prices each edge as EV. Postgres does the heavy lifting: row-level security IS the paywall, realtime push lands each compute cycle in the UI, and a tipoff-aware cache tightens the refresh from 15 minutes at idle to 2 minutes near game time.
Highlights
- Four selectable fair-price modes — sharp consensus, single-book, exchange-aware, projection blend
- Line-shift devig: re-prices a target line from reference books quoting different numbers
- 16 books and DFS apps, 10+ leagues from the NBA to ATP tennis and PGA golf, 25 prop-market families
Results
- Live at fairlinesports.com — Stripe subscriptions, OAuth, 7-day trial
- Backtested in the open: 3,145 bets over 6 months, +3.48% ROI, published on the landing page
- ~27,000 lines of TypeScript and SQL across the app, edge functions, and pipeline
- API cost engineering enforced in Postgres — clients cannot burn a credit against a fresh cache
A complete subscription business — quant model, product, billing, and brand — designed, built, and shipped end to end.