Signal-Driven#145

Counterparty Smart-Money Clustering & Fade/Follow

Group on-chain Polymarket traders into clusters by how much real edge they've shown and how well they time entries, then follow the consistently profitable group and bet against the consistently losing group whenever either places a fresh, oversized position. The edge is attributing skill at the wallet level using the whole trade graph, rather than blindly copying one famous whale.

What you need to run it

Full historical fills indexed by wallet with realized-PnL labels
Clustering/skill-scoring model with out-of-sample validation
Real-time new-position detection per tracked cohort
Position sizing decorrelated across overlapping cohort signals

Where this applies

Markets on Polymarket where counterparty smart-money clustering & fade/follow is the natural play:

Will the Republican candidate win the 2028 presidential election?
Will a US-China trade deal be signed before the end of 2027?
Will the UK hold a snap general election before the end of 2027?

Capabilities this demands

On-chain / walletModel / quantData ingestionCustom code / API

At a glance

CategorySignal-Driven

Requirements4

CapabilitiesOn-chain / wallet, Model / quant, Data ingestion, Custom code / API

VenuePolymarket (CLOB, Polygon)

Build it

Related signal-driven 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 (147 strategies) or the data resources directory.