Method
TrumpWatch is built on three primitives: real-time transcripts, frequency counting, and private market signal scoring.
01 · Source
All mention data is pulled from akta.pro News and Signals in open-ended query mode. Every public appearance — rallies, pressers, social posts, pool sprays — is transcribed and indexed in near real time. There is no manual scraping, no curation. The query runs continuously.
02 · Counting
A "mention" is a normalized hit on the tracked term inside a verified transcript segment. Variants and obvious aliases collapse to a single canonical term. The frequency curve is a per-day count over the selected window.
03 · Signals
A spike is any day exceeding the rolling 30-day mean by a configurable z-score. For each spike, the corresponding private market signal — sourced from the same akta.pro signal feed — is captured in a 72-hour window and reported as a percent delta.
04 · Predictions
The weekly call is generated from a term's recent-cadence model versus the upcoming appearance schedule. Tiers (easy / medium / hard / long shot) reflect modeled probability bands. There is no betting and no scoring system. The reveal is live the moment a transcript hits.
src/lib/stub-data.ts to akta.pro endpoints when keys are available.