90% of the coverage is reaction, not news
Sports news is a self-exciting process: one story produces several more. Fitting that separates what happened from how loudly it was reported.
Why a Hawkes process
A story count cannot distinguish a busy week from one event everybody covered. A self-exciting point process can: the background rate is how often genuinely new things happen, and the branching ratio is how much reaction each one generates. A ratio near one means the coverage is nearly self-sustaining.
What this is not
It is not a read on anybody's state of mind, and it never will be. It counts stories and scores the language of headlines against a published lexicon you can inspect word by word. No number here enters a match price until a backtest says it should, and that backtest needs a season of archive it does not yet have.
Exponential Hawkes fitted by maximum likelihood on publication times, a log-linear fit to story volume by age, and Shannon entropy over the team share of coverage.
One feed, football only, and an archive measured in weeks. Everything here describes the coverage, not the teams.