NewsMarkets

Press vs. reality

The market gets a calibration page. This asks the same question of the reporting. For every story whose market has resolved, we take the outcome the market favoured when the story first appeared and ask whether it happened, grouped by how the coverage framed it and by how much coverage there was. Next to each hit rate is what the market itself expected, so you can see which side had it right.

By how the coverage framed it

Framing is read from each story's coverage as it is labelled, starting now, and a story only counts once its market has resolved. These rows fill in over the coming weeks.

Coverage Stories Favourite won Market expected Gap
Reported as settled — it has happened, or is reported as fact 0 — — —
Reported as expected — likely or imminent 0 — — —
Reported as contested — an open question; sources disagree 0 — — —
Reported as speculative — a possibility, rumour or proposal 0 — — —

By how much coverage there was

Publishers covering the story over its life. More coverage is the press saying it matters.

Publishers Stories Favourite won Market expected Gap
<5 46 76% 71% +5
5-9 5 80% 69% +11
10+ 1 100% 66% +35

Recently resolved

How this is counted

Only two-outcome markets count, so the favourite is unambiguous. The favourite is whichever outcome the story's best-matched market priced above 50% when the story first appeared, from the prices we record every few minutes; a story whose market was not being tracked at that moment is left out. "Market expected" is the average price it gave those favourites, which is what a perfectly calibrated market would have hit. Markets that resolved in the last 60 days are included; 52 stories so far.

The framing is read by the same model that writes each story's headline, from the sources' own titles and leads, as one of four words. It is a reading of tone, not of any single outlet's claim.