Alerts per week

Alerts per week is how many signals a given rule would have fired, averaged over a trailing window — shown for the last 30 days and the last 90 days. It’s the volume cost of a filter: tighter rules (higher consensus, higher conviction) fire less often.

The 30-day and 90-day rates can diverge sharply because older data came from sparse leaderboard backfills while recent data comes from dense live polling. The 30-day figure is the better forward proxy, but it’s more volatile.