World events priced by prediction markets, on a calendar and linked to the assets they move. Each event shows how likely it is by every date; each arrow's weight is estimated from how the asset has moved with the market's probability. Pick an event and a date, and ask "what if?".
Nodes. Events are binary outcomes priced on Polymarket; their probability is the price of the YES share. Assets are futures, ETFs, stocks, yields and Bitcoin.
Edges. Each arrow means "a change in the source moves the target". For an event pointing at an asset, the weight β is the asset's log-return if the event's probability went from 0% to 100%. It is estimated by regressing daily asset returns on daily changes in the market's probability (same day and previous day, Newey-West errors), the method Leigh, Wolfers & Zitzewitz used in 2003 to price war with Iraq at about $10 a barrel. Each estimate is blended with a judgment prior, weighted by precision, so a noisy regression barely moves a confident prior.
What-if. Choosing "If YES" moves the event to 100% and passes the change along the arrows: β × change, summed over paths, up to four steps. It treats each arrow as a direct effect. Many estimates are co-movement, not cause (two Fed markets react to the same news), so read the results as "what the market has been pricing together", not as a forecast.
Cross-market checks. A Polymarket "Bitcoin above $K on date" market is a digital option, so Deribit's option prices give an independent probability (Breeden–Litzenberger on a smoothed smile, interpolated to Polymarket's settlement time). The December Fed decision is compared with the fed funds future. Gaps are after Polymarket's taker fees but before execution risk, settlement-source differences and the options' risk premium. They are candidates to check, not risk-free trades.