Windows
For each post, the system measures price + volume change at four windows after the post timestamp: 15 min · 1 hour · 1 day · 3 days.
Baseline
15m and 1h windows compare against the average of the prior 30 minutes of intraday bars (intraday_pre_post). 1d and 3d windows compare against the prior 5-day daily-close average (prior_5d_avg).
Abnormal return (the headline number)
Each cell shows the move net of the market: the stock's change minus the S&P 500 (SPY) change over the identical window. A stock that rose 0.5% on a day the whole market rose 0.5% shows ~0% abnormal — so broad drift can't masquerade as a reaction. Hover any cell for the raw move and the market move. Where the index baseline can't be computed, the raw move is shown with a ~ marker.
Who profits
The point isn't that a stock moved — it's that the poster controls the move. Each case links the beneficiaries on file (donors, family holdings, contract-holders) and any documented trades in that ticker close to the post (president's family/associates where documented, plus members of Congress from the STOCK Act ledger). Stated as documented overlaps — never as proof of intent.
Data source
Intraday minute bars from Yahoo Finance's public chart endpoint (free, no API key). 1-minute bars are available for the last ~7 days; 5-minute bars for ~60 days. Older posts can only show 1d / 3d windows once daily prices have been ingested.
Honest labels (not "no data")
When a window can't be computed, the row shows why: intraday_unavailable (post older than 60 days), market_closed (weekend / holiday / off-hours), no_ticker (entity is mentioned but not publicly traded), pending (window has not elapsed yet).
Correlation only
Same-window price moves are reported; causal mechanism is not asserted. A defense-stock dip after a personnel announcement could be the announcement, the broader tape that day, an unrelated company event, or pure coincidence. The measurement documents the timing; the inference is the reader's.
Scope
Public posts from the sitting president (Truth Social and X). Mention list is operator-confirmed at file time — automated mention-extraction is intentionally narrow to avoid false ticker links. Source code: lib/post-window-impact.ts + lib/intraday-prices.ts.