gameflow-skill-rating
GameFlow Skill Rating
GameFlow stores and computes player ratings for you. Each player has μ (mu), the skill estimate, and σ (sigma), the uncertainty. GameFlow derives ordinal = μ − 3σ for display and leaderboards. You create a model in the dashboard and attach it to a matchmaker. After each match you report who won, and GameFlow updates every player's rating.
Account setup, API keys, base URL and JSON conventions are covered in the gameflow skill. Queues and tickets are covered in gameflow-matchmaking, and the server SDK and payload in gameflow-servers. This skill doesn't repeat them.
Skill Rating is labelled Alpha in the docs. Field names below were checked against the API code (develop, 2026-10). If the GameFlow docs MCP is connected, check each request body against get_api_operation before writing it. If a field name differs from this skill, don't choose yourself or write code with either name. The docs can be ahead of the deployed API, and the wrong name is silently dropped. Show the user both names and ask which one their environment accepts.
Rules
- The API key never ships in a game client. Every call here is made by the dedicated game server or by the customer's backend. Clients that show a rank or leaderboard ask the customer's backend, which calls GameFlow.
- Only trusted code reports results. If the client reports, any player can award themselves wins.
matches:reportis not idempotent. Each successful call applies the rating update again, andexternalMatchIdis stored but not deduplicated. Read the retry policy before you add retries.- JSON is camelCase (
externalPlayerId,muAfter,skillModelId). Older docs show snake_case response fields, which are wrong. The one exception is the matchmaking payload the game server receives, which is snake_case (player_id,match_id).