# ParlayAI

> AI-powered sports betting analytics platform delivering data-backed picks across 7 live sports. Every pick is logged to a public track record with full transparency on win rate, ROI, and unit profit.

ParlayAI uses a dedicated machine-learning model per sport (50-234 features depending on the sport), Monte Carlo simulation (5,000-10,000 iterations), Bayesian overconfidence calibration, and an LLM validation layer called ATLAS before publishing any pick. Customers receive picks via Discord and the website dashboard. The system targets 52-56%+ win rate on spreads/totals (enough to beat -110 juice) and only posts picks with positive expected value.

## What's here

- [Home](https://parlayai.net/): Product overview, hero proof bar showing live win/loss/unit stats across all sports
- [Picks](https://parlayai.net/picks): Today's active picks (requires subscription for full detail; free tier sees MLB)
- [Track Record](https://parlayai.net/picks/history): Public historical performance (all-time and filtered, by sport)
- [Performance](https://parlayai.net/picks/performance): Sport-by-sport performance dashboard with unit P/L, win rate, ROI
- [Pricing](https://parlayai.net/tiers): Subscription tiers ($19.99/mo after 3-day trial; MLB always free)
- [Login](https://parlayai.net/login): Discord OAuth, Google OAuth, or email/password
- [Account](https://parlayai.net/account): Subscription management, user profile
- [Signup](https://parlayai.net/signup): Create account, verify email, start trial

## Sports coverage (live)

- MLB — 200 features, Statcast integration, park factors
- NFL — NGS (Next Gen Stats), QB metrics, 33 NGS features, player aggregation ensemble
- NBA — 234 elite features, rest/B2B/injuries/rolling stats
- NHL — goalies, Edge IQ, period scores, OT/SO handling, puckline-copy ML strategy
- NCAAB — KenPom, tempo metrics, Four Factors, regression totals
- UFC — fighter stats, style matchups, method prediction (63-feature winner model, 74% ML accuracy)
- CFB — SP+, Elo, EPA, transfer portal impact, returning production

## Sports in shadow-mode testing

Soccer (7 leagues), Tennis (30 tournaments), Golf (PGA + 4 majors), College Baseball, CS2, League of Legends. These unlock automatically as they graduate accuracy thresholds.

## How picks are generated

1. Data collection from sport-specific APIs (OddsAPI, ESPN, Statcast, KenPom, DataGolf, bo3.gg, nflverse/nflfastR)
2. Feature extraction (50-234 features per sport)
3. ML prediction (GradientBoosting / LightGBM / XGBoost per market)
4. Monte Carlo simulation (5K-10K sims)
5. Bayesian calibration for overconfidence correction
6. ATLAS LLM ensemble validation
7. Publishing to Discord + database
8. Auto-grading after games end
9. Continuous learning — models retrain when threshold met

## Track record approach

Every pick logged at post-time: teams, odds, confidence, edge, bookmaker, game time, sport. After games end, results auto-graded from result APIs. Public track record at /track-record shows real win/loss/push counts and units profit. All-time vs filtered (current model era) both displayed for transparency.

## Markets covered

Moneyline, spread (point spread / puckline / runline), totals (over/under), player props (NBA / NHL / NFL / NCAAB / Soccer), and sport-specific markets (NRFI, F5, team totals, first basket, etc.).

## Contact

- Discord: https://discord.gg/parlayai
- X / Twitter: @ParlayAIHQ
- Support email: support@parlayai.net

## License / terms

Picks are for entertainment and informational purposes. Users must be 21+ and gambling-legal in their jurisdiction. Past performance does not guarantee future results. See /terms on site.