Méthode

White noise measures attention, not effect. Components: stories tagged to the player or club in the league's own news feed (7 and 30 days); Wikipedia daily pageviews of the player's or club's article (7 and 30 days, against a 90-day median as the baseline); social posts naming the player or club in the last 7 days across the connected networks (Bluesky public search (AT Protocol AppView)), with likes, reposts, replies and comments as engagement; roster moves in the last 7 days; an active injury listing. Each is log-scaled, weighted (players: news 30%, pageviews 30%, social posts 20%, social engagement 10%, moves 7%, listing 3%; clubs: news 30%, pageviews 25%, social posts 20%, social engagement 10%, the roster's own stories 15%) and turned into a percentile among players or among clubs. Rising is this week's daily pageviews over the 90-day median; social rising is this week's posts over the median of the last 30 daily readings, once a week of readings exists. A common name's social count is scaled by the share of sampled posts that read as hockey talk. Sources : NHL.com stories (league content feed); Wikimedia pageviews API (English Wikipedia, user traffic); Bluesky public search (AT Protocol AppView); CapOrCup roster moves and injury listings (licensed feeds). Reconstruit chaque matin; les relevés de chaque jour sont conservés, pour que dans une saison la question « le bruit précède-t-il un changement dans le jeu ? » puisse être posée avec des données plutôt qu'une impression.