Mentions & Social · Quantitative#281

Multi-Word Bundle Joint Probability

Markets that ask whether a speaker says all of several words, or any of them, are joint-probability questions, and the words are strongly correlated — a speech that mentions tariffs almost certainly mentions China. Traders price these by multiplying independent guesses, which badly underprices the 'all of them' case and overprices 'any'. You estimate the correlation structure from the transcript corpus and trade the bundles against the naive independent price.

What you need to run it

Co-occurrence statistics between target words across the speaker's past transcripts
Joint-probability model that respects correlation rather than assuming independence
Topic conditioning, since the correlation depends on the subject of the speech

Where this applies

Markets on Polymarket where multi-word bundle joint probability is the natural play:

Will the president say both 'China' and 'tariffs' in the address?
Will any of these three words be said during the debate?
Will the speech mention all four listed topics?

Capabilities this demands

Model / quantData ingestionDomain knowledge

At a glance

CategoryQuantitative

MarketMentions & Social

Requirements3

CapabilitiesModel / quant, Data ingestion, Domain knowledge

VenuePolymarket (CLOB, Polygon)

Build it

Related mentions & social strategies

This is documentation, not advice. Poly Research & Robotics publishes how these strategies work because the method should be checkable — not as a recommendation to trade them. See the full strategy database (297 strategies) or the data resources directory.