GoalGorithm MCP Server

Soccer match predictions using xG data and Poisson distribution, exposed as MCP tools for Claude Desktop/Code.

Install

Or run directly:

Claude Desktop Config

Add to your Claude Desktop config (claude_desktop_config.json):

Example Usage

Once configured, just ask Claude naturally:

You: "Predict Arsenal vs Chelsea this weekend"

Claude will call the predict_match tool and respond with something like:

Claude: Here's the prediction for Arsenal vs Chelsea (Premier League):

Outcome	Probability
Arsenal Win	52.4%
Draw	22.7%
Chelsea Win	24.9%

• Expected Goals: Arsenal 1.85 — Chelsea 1.23

• Over 2.5 Goals: 58.3% | Under 2.5: 41.7%

• Both Teams to Score: Yes 52.1% | No 47.9%

• Most Likely Scores: 1-0 (12.8%), 1-1 (11.2%), 2-1 (10.5%)

Arsenal are clear favorites at home with stronger attacking xG.

Other things you can ask:

• "Show me the La Liga xG table" — calls get_league_table

• "Which leagues are available?" — calls list_leagues

• "Who's more likely to win, Bayern or Dortmund?" — calls predict_match

Tools

predict_match

Predict soccer match outcome using xG-based Poisson model.

Returns: win/draw/loss %, over/under 2.5, BTTS, top 3 scores, expected goals, score matrix.

list_leagues

List all supported soccer leagues with IDs and slugs.

get_league_table

Get all teams in a league with their xG statistics, sorted by attacking strength.

Supported Leagues

How It Works

1. Fetches team xG/xGA stats from Understat.com

2. Computes attack/defense strength relative to league average

3. Applies Poisson distribution to calculate goal probabilities

4. Builds 6x6 score matrix for all possible scorelines (0-5 goals each)

5. Derives match outcomes: W/D/L, Over/Under 2.5, BTTS

Data Source

All data from Understat.com public JSON API. Results cached locally for 12 hours.

License

GPL v2 or later