MarketPulse · live paper experiment

The shock-dip experiment

When a provocative post from the President lands on a day the Nasdaq-100 sells off, does the dip bounce? Three $10,000 paper books answer it live, one trading session at a time.

The experiment's snapshot isn't available right now. It is written after each US market close.

The rule

  1. Hold QQQ (the Nasdaq-100), fully invested.
  2. Every post is scored 0–10 for provocation by a local 14B model, using a prompt calibrated on known market movers.
  3. A post scoring 7+ on a day QQQ closes down more than 1% is a shock dip: rotate into leverage at that close for 5 sessions, then back to QQQ.
  4. In the interpreted book, an AI interpreter chooses the leverage (3x, 2x or 1x). It reasons under a style picked by SelinAI's reasoning-policy library, reads the posts, checks the scorer, and weighs the fear and recession signals and the ML forecasts by their measured trust. If it fails, the book falls back to the rule's 3x. The rules book always takes 3x, so the two together measure whether interpretation helps.

Why it might work

A study of Nasdaq-100 sell-offs since 1986 and of 34,000 presidential posts (2017–2026) found that dips on days with a provocative post bounced more than other dips over the next 5 sessions, in office and out of it. The post seems to label a headline-driven panic rather than cause it, and fear without a recession behind it has historically been followed by better returns.

Paper trading only: no real money and no broker connection. Historical replays are not a forecast. Not investment advice. Post excerpts are quoted from public Truth Social posts via CNN's public archive. Back to the live demos