get_predictions

5 creditsStats + aggregationPOST /v1/get_predictions

The model's win probability and A/B/C grade for each NHL game on the current slate, with the model version and a backtest note.

Use it when you want the model's pre-game view of tonight's games: home and away win probability and a confidence grade per game, to feed a dashboard, a bot or your own comparison. Current slate only; no date argument. A game's probabilities are the model's own, with no odds, prices, goalie or injury notes. An empty slate costs nothing, and neither does a call that fails because the feed is down or stale.

Parameters

No parameters.

Notes

Over MCP, ask for it in plain English and the client calls get_predictions itself. Point any MCP client at https://mcp.puckapi.com/mcp?key=YOUR_API_KEY.
Each call costs 5 credits, taken before it runs and refunded if the server errors.
Not the call you need? Find the right call by what you are trying to answer.

Request

curl -X POST https://mcp.puckapi.com/v1/get_predictions \
  -H "Content-Type: application/json" \
  -H "x-api-key: YOUR_API_KEY"

Response

json

{
  "generatedAt": "2026-10-15T19:17:39.008480+00:00",
  "modelVersion": "v2.0",
  "accuracyNote": "60.7% over 7,428 games (2021-22 to 2025-26) is a walk-forward backtest, not a live record. Live 2026-27 results are at puckcast.ai/model-health.",
  "games": [
    {
      "gameId": "2026020108",
      "gameDate": "2026-10-15",
      "startTimeUtc": "2026-10-15T23:00:00+00:00",
      "homeTeam": "TOR",
      "awayTeam": "BUF",
      "homeWinProb": 0.4311,
      "awayWinProb": 0.5689,
      "grade": "B"
    },
    {
      "gameId": "2026020111",
      "gameDate": "2026-10-15",
      "startTimeUtc": "2026-10-15T23:00:00+00:00",
      "homeTeam": "TBL",
      "awayTeam": "SEA",
      "homeWinProb": 0.664,
      "awayWinProb": 0.336,
      "grade": "B"
    }
  ],
  "count": 2
}

Response fields

generatedAtstring: When the slate was generated (ISO 8601, UTC)
modelVersionstring: Model version from the accuracy feed (e.g. "v2.0")
accuracyNotestring: The backtest accuracy and its basis: a walk-forward backtest, not a live record
games[].gameIdstring: NHL game ID
games[].gameDatestring: Date (YYYY-MM-DD)
games[].startTimeUtcstring: Scheduled start (ISO 8601, UTC)
games[].homeTeamstring: Home team abbreviation
games[].awayTeamstring: Away team abbreviation
games[].homeWinProbnumber: The model's probability that the home team wins (0 to 1)
games[].awayWinProbnumber: The model's probability that the away team wins (0 to 1); the two sum to 1
games[].gradestring: The model's confidence grade: A, B or C
countinteger: Games on the slate