Testing your integration

A test key plays the whole flow with no money in it. One endpoint exists only here: it decides a match, so you can see a result without two people playing one.

Deciding a match in test

The test environment can end a seated match on demand. You need a matchId first: seat two of your own test players with POST /partner/matches (see Paid play) — that call returns the matchId this endpoint decides.

Name every seat by the same playerId you minted its session with — a missing one is refused rather than filled in. The match finishes exactly as a played one does: same finish, same seal, same rows in the results feed, same webhook nudge. Your CI can then check that it paid the player you decided would win.

Test keys only. The handler cannot be reached without a token only a test key can produce. A live key gets the same 404 no such test match as a nonexistent id, so it learns nothing about which matches do — unless you're on a deployment pinned to one environment, where it's refused earlier with live_key_against_test. The match must also be between test players of yours.

What it refuses with. A match id that isn't yours, isn't a test match, or doesn't exist is 404. A match not in a decidable state — already finished, or never seated — is 409. A body missing a seat is 400. These three carry a message but no reason code, so branch on the status.

What a green suite covers. The half that is yours: the seat, the feed, the webhook, and your ledger paying the player you decided would win. It does not exercise our scoring engine or the anti-cheat replay behind it — the score comes from you, not moves — nor a payout of ours, since a test match moves no money on either side.