Explained match intelligence
Footnote picks out the moments worth a graphic, words each for casual fans and analysts, links it to the evidence behind it and holds it until your stream reaches the moment.
Watch the showcase match as Nadia, the producer who approves each graphic before it airs, or as one of four fans.
Watch as
Start from
Speed
As Sam, from 23', at 2×. You can switch between viewers during the replay.
How a graphic is made
1Ranked
The processor finds the moments worth a graphic and words each with a template. A graphic's rank starts from a base for its kind of moment, raised by the match-state model and the moment's figures.
2Redrafted
Writers built with Microsoft Agent Framework on a Foundry model redraft the graphics where wording adds something, reading the match only through our MCP match tools.
3Checked
Writers put a placeholder where each figure goes. Code fills the figures in and checks every draft, refusing any that types one; a draft that fails or comes late leaves the wording already on screen.
4Delivered
An approved graphic shows for each viewer when their stream gets to the moment: in English, or in Spanish or French through Azure AI Translator.
Also on Microsoft Azure:
- Event Hubs: the event feed
- Container Apps: the services
- Azure Database for PostgreSQL: each replay's state and graphics
- Web PubSub: pushes to the page
- Application Insights: a trace of each agent draft