Sports data hub: transcript

This is a sports data hub we built for a prediction-media client: thousands of tennis pages that keep themselves current.

Players, tournaments and matches each get a page, with live scores, market prices and results as they happen.

It went live in October with about fifty-six hundred pages ready for search engines.

Behind it are one database, one service and two clocks. Here's how it works.

Every fifteen minutes, numbers move into a snapshot. Every three hours, the whole site is rebuilt as static pages and published worldwide.

Only the fixture feed may create a match or a page. Market data can only attach to what already exists, so a price can never invent a player.

Every player has one internal id, and every outside name goes through one strict alias ladder, so the same player never splits in two.

During matches, results settle from a thirty-second live poll, and a stale 'live' turns into 'result pending' instead of misleading anyone.

Player summaries are written by a fast model with a review pass, for under a cent each. A gate rule caught three wrong claims before any page published them.

Every job declares how stale it may get, every run records its outcome, and the health check queries the real data, not a heartbeat.

Every release is verified against the live site: a hundred and eighty-nine planted checks, zero broken.

The sports data hub: thousands of pages, always current.