Notebook FC
How Notebook FC works
From the final whistle to the page: how a match becomes a record, and the record becomes analysis.
Gameday engine
Records every match as it is played and after it ends: the result, the goals and cards in order, the lineups, the substitutions and each player’s figures.
Knowledge Engine
Turns the record into knowledge, and checks each piece against the record before it is published.
- Discovery finds what stands out in a season, then verifies each finding on its own
- Articles research a development and write it up, with every figure checked
- Briefs short analyses of a match or a run of form
- Exhibits charts with a short story
- Experiences interactive simulations, such as how a playoff race could finish
Data API
Serves the record and everything published from it to the site, in fixed shapes the site checks before it shows anything.
Notebook FC
Places each piece where it helps a reader most, and learns from what readers read which pieces to show first.
The record: the Gameday engine
Everything on Notebook FC starts with the match record. The Gameday engine records each match in the NWSL, the Women’s Super League and the UEFA Women’s Champions League while it is played and after the final whistle: the score, the goals and cards in the order they happened, the starting lineups, every substitution and each player’s figures.
Every version of a match is kept. When a result or an event is corrected, the correction is added as a new version rather than written over the old one, so the record can always say what it held and when. A match that has not been recorded yet is shown as not recorded, never filled in with a guess.
Analysis: the Knowledge Engine
The Knowledge Engine turns the record into analysis. Each of its services publishes its own work, and each piece is checked against the record before it reaches the site.
Discovery reads the season record and finds what stands out: a club’s longest run, the widest gap between goals and expected goals, a player’s share of her club’s goals. A separate check recalculates every finding before it is published.
Articles report a development, such as a result, a transfer or a player’s return. They are written with language models from our record and published reporting, then checked: every figure must match the record and every claim must be supported by a source the article links. An article that fails the check is not published.
Briefs are short analyses of a match or a run of form, checked the same way.
Exhibits are charts with a short story, each dated and readable as a table.
Experiences are interactive explorations. A simulation plays out the rest of a season, so you can see how a race for the table or the playoffs could finish. Simulated figures are labelled as modelled.
Delivery: the Data API
The Data API serves the record and everything the Knowledge Engine publishes to the site. It answers in fixed, versioned shapes, and the site checks every answer against them before it shows anything. When something cannot be read, the site says so rather than showing something else in its place.
The site: Notebook FC
Notebook FC decides where each piece appears: the newest leads, and a page shows what is most relevant to it, such as a club’s numbers on its own page. With readers’ consent, it counts in aggregate which pieces are read, and uses those counts to decide which pieces to show first.
How we use machine learning
Language models write Articles and the stories that go with Exhibits. They never publish on their own: a separate check compares every figure with the record and every claim with its sources.
Simulation plays out the remaining matches of a season to show how a race could finish. It is labelled as a model, not a prediction.
Expected goals (xG) and expected assists (xA) are model estimates of the quality of a chance, shown beside the goals and assists they explain.
Placement learns from what readers read, in aggregate, which pieces to show first.
How we check and correct our records is set out in Methodology. What the site covers is set out in About.
