A sports-analytics MCP

A betting model you can interrogate.

good-sport is an analyst for sports betting markets, reachable inside Claude, ChatGPT, or Grok. Ask where the edge is, where the smart money is moving, and the part nobody else will show you: how often the model is actually right.

predicted actual
It said 63%. They won 57%. Every graded bet, grouped by confidence and checked against what actually happened. Below the line means overconfident, and it shows you the gap instead of hiding it.

How it works

You ask in plain language. It answers with its reasoning.

  1. You Where's the edge on tonight's slate?

    ↳ prop_slate(min_edge: 0.10)

    good-sport A handful of legs clear the bar. On batter props with a 10%+ edge, the model has hit 9% where the market priced 47% — pitcher strikeouts run overconfident, so they're scored on their own (next question). Two books are tight against the strongest one, so treat it as a lean, not a lock.

  2. You Is the model overconfident on pitcher strikeouts?

    ↳ model_performance(prop: "pitcher_strikeouts")

    good-sport Yes, about 12 points overconfident across 2700 graded bets, and it's the one prop type running a loss. I'd down-weight it until the re-audit lands.

  3. You Where's the smart money moving?

    ↳ smart_money(sport: "all")

    good-sport One side has steamed four points of implied probability since open across five books. That's a market signal, not my edge. I'm flagging it, not endorsing it.

Why it's different

Built to keep itself honest.

9% hit 47% market

Calibration

It grades itself against what happened. On its batter-prop leans clearing a 10%+ edge, it has hit 9% where the market priced 47%. Pitcher strikeouts run the other way and are scored separately. On thin edges it's noise, and it tells you which is which.

MLB 79% NHL 46%

Closing-line value

Did it beat the closing line? That's the textbook sign an edge is real. On MLB, 79% of the time. On NHL, 46%: at market, and it says so.

Tennis −1.8% MLB −32%

It reports the misses

Over the last 30 days, tennis is down 1.8% across 76 graded bets, while MLB game lines came in about break-even. A good analyst shows the flat stretches and the misses, not just the wins.

No new app

Works where you already think.

It's a connector for the assistants you use. Add the server once; ask it anything, anywhere those models go.

good-sport-mcp

See where the edge is, and where it isn't.

Full tooling

$5/mo

Predictions, props, parlays, calibration & CLV receipts — the whole research surface.

Cancel anytime. No spam, no picks in your inbox.