Oddsy

Bettors who use multiple sportsbooks, DraftKings, FanDuel, BetMGM, Caesars, and a dozen others, have no single source of truth for their own history. Bets live scattered across apps, results get forgotten, and there's no easy way to tell if a betting strategy is actually profitable over time or just winning by luck this week.

Manually logging every bet is exactly the kind of friction that kills a tracking habit before it starts. Oddsy sends a screenshot of a bet slip straight to a GPT-4o-mini vision model for structured extraction, no manual typing, with a Tesseract.js OCR pass as an automatic fallback if the vision call fails. A custom team-alias resolver covering the NFL, NBA, MLB, NHL, WNBA, MLS, and international soccer catches what's left, turning it into a structured, trackable bet in seconds.

Once a bet is logged, Oddsy grades it automatically against live game state, no manual "did I win" bookkeeping. It also tracks closing-line value: whether the odds moved in the bettor's favor or against them after the bet was placed, one of the sharpest signals in sports betting for whether someone is actually skilled or just running hot.

On top of the tracked data sits an OpenAI-backed natural-language query layer, ask "how am I doing on NBA player props this month" and get a real answer pulled from your own betting history, not a generic model guess. It sits alongside a library of 60+ sport- and prop-specific models for bettors who want to go deeper.

Oddsy ships as native iOS and Android apps via Capacitor with a PWA fallback, backed by Supabase for auth, data, and edge functions. Subscriptions run through Stripe on web and native StoreKit2 / Google Play Billing on mobile. It's a real, published consumer app, not a prototype.