SportIntel MCP - AI-Powered Sports Analytics

An MCP (Model Context Protocol) server that provides AI-powered analytics for Daily Fantasy Sports (DFS). It delivers explainable player projections, lineup optimization, and real-time betting odds.

Overview

SportIntel MCP is a tool for developers and DFS players. It uses AI to generate player projections with SHAP-based explanations, builds optimized lineups, and fetches live odds from multiple sportsbooks. It runs in two modes: a single-execution batch mode for getting results, or a persistent server mode for integration with MCP clients like Claude Desktop.

Key Features

• Player Projections (get_player_projections): Fetches DFS projections for 200+ players per slate. Includes DraftKings/FanDuel salaries, projected fantasy points with floor/ceiling ranges, value scores, ownership estimates, confidence scores, and SHAP explanations for AI reasoning.
• Lineup Optimizer (optimize_lineup): Generates optimal DFS lineups. Supports strategies for cash games or tournaments, handles positional constraints, and allows for player exclusions/inclusions and stack identification.
• Live Odds (get_live_odds): Retrieves real-time betting odds from 10+ sportsbooks. Covers spreads, totals, moneylines, and player props, with best-odds finding and line movement tracking.
• Explain Recommendation (explain_recommendation): Provides SHAP-powered explainability for any player recommendation, showing feature importance and human-readable reasoning.

How to Use

The actor primarily operates in batch mode. You configure a run with an input specifying the tool and its arguments.

Input Structure

The basic input JSON requires a mode and a tool with its arguments.

Batch Mode Example:

{
  "mode": "batch",
  "tool": "get_player_projections",
  "arguments": {
    "sport": "NBA",
    "slate": "main",
    "minSalary": 8000
  }
}

Server Mode Example:
To run as a persistent MCP server for client connections:

{
  "mode": "server"
}

Available Tools & Arguments

• get_player_projections

  • sport (required): "NBA", "NFL", "MLB", "NHL"
  • slate (optional): "main", "early", "afternoon", "evening"
  • date (optional): ISO 8601 date (default: today)
  • minSalary / maxSalary (optional): Salary filters.
  • positions (optional): Array of positions to include.

• optimize_lineup

  • sport (required): "NBA", "NFL", "MLB", "NHL"
  • projections (required): Array of player projection objects.
  • strategy (optional): "cash", "tournament", "balanced".
  • lineupCount (optional): Number of lineups (1-150).
  • requiredPlayers / excludedPlayers (optional): Player ID arrays.
  • maxPlayersPerTeam (optional): Default is 4.

• get_live_odds

  • sport (required): "NBA", "NFL", "MLB", "NHL"
  • markets (optional): e.g., ["spreads", "totals", "moneylines", "player_props"]
  • team (optional): Filter by team name.
  • bookmakers (optional): Array of specific bookmaker names.

• explain_recommendation

  • playerId or playerName (optional): Identifier for the player to explain.
  • method (optional): "shap", "lime", "feature_importance".

Input/Output

Example Output

A call to get_player_projections returns structured data like this:

{
  "sport": "NBA",
  "slate": "main",
  "date": "2025-11-23",
  "projections": [
    {
      "playerName": "Giannis Antetokounmpo",
      "team": "MIL",
      "position": "PF",
      "salary": 5500,
      "projectedPoints": 63.4,
      "floor": 50.7,
      "ceiling": 76.1,
      "value": 11.53,
      "ownership": 45.2,
      "confidence": 0.95,
      "explanation": {
        "topFactors": [
          {
            "factor": "recent_performance",
            "impact": 63.4375,
            "description": "Averaging 63.4 FP over last 8 games"
          }
        ]
      }
    }
    // ... more players
  ]
}

Data Storage

Results are saved within the Apify platform:
* Dataset: Contains the primary data (player projections, generated lineups, odds data).
* Key-Value Store: Holds metadata, cache, and run status information.