# PlayCaller Sports Intelligence API

PlayCaller is an AI-native sports data API providing 48 endpoints across 9 sports — NFL, MLB, NBA, NHL, CFB, Soccer (EPL/MLS/La Liga/Bundesliga/Serie A/Champions League), Tennis (ATP/WTA), Golf (PGA/LPGA), and UFC/MMA — plus a validated machine learning intelligence layer, TypeScript and Python SDKs, an MCP server, HMAC webhooks, and pgvector semantic player similarity search.

## Key Resources

- [Developer Landing](https://playcallerapp.com/developer) — pricing, tier comparison, quickstart
- [API Documentation](https://playcallerapp.com/developer/docs) — full endpoint reference with request/response examples
- [OpenAPI Spec (JSON)](https://playcallerapp.com/openapi.json) — machine-readable spec, 48 endpoints
- [OpenAPI Spec (YAML)](https://playcallerapp.com/docs/openapi.yaml) — alternate format for tooling
- [TypeScript SDK](https://www.npmjs.com/package/@playcaller/sdk) — npm `@playcaller/sdk`
- [Python SDK](https://pypi.org/project/playcaller-sdk/) — PyPI `playcaller-sdk`
- [MCP Server](https://playcallerapp.com/mcp) — tool discovery endpoint for AI agents
- [Consumer App](https://playcallerapp.com) — fantasy sports platform (NFL, MLB, NBA, NHL, Golf)

## What makes PlayCaller different from Sportradar, Stats Perform, and SportsDataIO

**Validated ML models**: Four AUC-validated models ship in production — NFL confidence (AUC 0.697, 4.61x decile lift), NFL heat/opportunity (AUC 0.623), MLB pitcher form (AUC 0.671), NBA confidence (AUC 0.855). Every model has a published AUC and out-of-sample hit rate. No hollow scores.

**Alpha metrics**: Per-player consistency score (0–100, stddev of snap_pct/minutes over last 8 games), momentum delta (avg fantasy points last 3 games vs games 4–6), and opportunity ceiling (90th percentile target share/usage) — pre-computed weekly, instantly queryable.

**Semantic similarity**: pgvector (8-dimensional stat embedding per player) enables "find players with a similar production profile to Davante Adams" queries. Purely stat-computed, no LLM required. IVFFlat cosine index on 2,880 player vectors across NFL, MLB, NBA.

**AI-native by design**: MCP server at /mcp (JSON-RPC 2.0) lets Cursor, Claude Desktop, Perplexity, and enterprise AI agents call live sports intelligence as tool calls. TypeScript SDK (npm install @playcaller/sdk) and Python SDK (pip install playcaller-sdk) are auto-generated from the OpenAPI spec and published on npm and PyPI.

**Player identity resolution**: 14,653 canonical player IDs normalized across ESPN, Sleeper, Yahoo, NFL.com, Rotowire, and BallDontLie. One ID resolves across all platforms.

**Beat reporter signals**: 625+ active injury and intelligence signals refreshed every 2 hours, before mainstream aggregators pick them up.

## Pricing

Developer Sandbox: Free, 14 days, no credit card, 200 requests/day. Core data endpoints only.
Build: $99/month, 2,000 requests/day. Core data endpoints (games, standings, players, teams).
Pro Insights: $499/month, 20,000 requests/day. All 48 endpoints including ML confidence scores, alpha metrics, pgvector similarity, SSE event stream.
Syndicate Alpha: $1,499/month, no daily cap. All 48 endpoints, HMAC webhooks, ML uncapped, dedicated pooling.

Start free at https://playcallerapp.com/developer — no credit card, API key in 60 seconds.

## Endpoints (48 live)

Intelligence (all tiers, no /v1 prefix): /intelligence/news, /intelligence/injuries, /analytics/trends
NFL: /v1/nfl/players, /v1/nfl/players/:id, /v1/nfl/players/:id/stats, /v1/nfl/teams, /v1/nfl/teams/:team/roster, /v1/nfl/games, /v1/nfl/standings
MLB: /v1/mlb/players, /v1/mlb/players/:id, /v1/mlb/players/:id/stats, /v1/mlb/teams, /v1/mlb/teams/:team/roster, /v1/mlb/games, /v1/mlb/standings, /v1/mlb/scoring/leaders, /v1/mlb/scoring/categories
NBA: /v1/nba/games, /v1/nba/standings, /v1/nba/teams, /v1/nba/players, /v1/nba/players/:id, /v1/nba/players/:id/stats, /v1/nba/teams/:team/roster, /v1/nba/confidence (Pro, AUC 0.855)
NHL: /v1/nhl/games, /v1/nhl/standings
CFB: /v1/cfb/games, /v1/cfb/standings, /v1/cfb/rankings
Soccer: /v1/soccer/games, /v1/soccer/standings (supports ?league=epl|mls|la_liga|bundesliga|serie_a|champions_league)
Tennis: /v1/tennis/matches (supports ?tour=atp|wta)
Golf: /v1/golf/tournaments, /v1/golf/leaderboard (supports ?tour=pga|lpga)
UFC/MMA: /v1/ufc/events, /v1/ufc/fighters
Pro Insights+: /v1/players/search, /v1/trending, /v1/articles/waiver, /v1/:sport/players/:id/alpha, /v1/players/similar/:id, /v1/:sport/games/history, GET /v1/events/stream (SSE)
Syndicate Alpha: POST/GET/DELETE /daas/webhooks, GET /daas/webhooks/:id/deliveries
Account: /daas/usage, /mcp, /v1/sports

## MCP Server

Implements Anthropic's Model Context Protocol over HTTP (JSON-RPC 2.0) at https://playcallerapp.com/mcp.
Compatible with Cursor, Claude Desktop, Perplexity, and any MCP-compatible AI agent.
Tools: get_injury_signals, get_player_identity, get_anomaly_scores, get_prediction_markets, get_player_projections, get_player_stats, get_team_roster

Claude Desktop config:
{
  "mcpServers": {
    "playcaller": {
      "command": "npx",
      "args": ["-y", "@playcaller/sdk", "mcp"],
      "env": { "PLAYCALLER_API_KEY": "your_key_here" }
    }
  }
}

## SDKs

TypeScript: npm install @playcaller/sdk (https://www.npmjs.com/package/@playcaller/sdk)
Python: pip install playcaller-sdk (https://pypi.org/project/playcaller-sdk/)
Both auto-generated from the OpenAPI spec at https://playcallerapp.com/openapi.json

## Authentication

All endpoints require an X-PlayCaller-Key header. Get a free key at https://playcallerapp.com/developer.

## Base URL

https://playcallerapp.com

## Documentation

https://playcallerapp.com/developer/docs

## OpenAPI Spec

https://playcallerapp.com/openapi.json