InsidePoly

LIVE · 2026

insider-trading surveillance for Polymarket — every trade watched, every wallet scored

fig. 1 — the 0–100 insider-likelihood scale, scoring livewallet 0x4f2…a91 · score 0

NORMAL 0–29

WATCHLIST 30–59

SUSPICIOUS 60–79

FLAGGED INSIDER 80–100

01THE PROBLEMinformation asymmetry, fully public

Prediction markets are uniquely vulnerable: the data is all on-chain and public — yet no tooling existed to flag wallets trading with suspiciously prescient timing.

InsidePoly closes that gap for Polymarket, the world's largest prediction market on Polygon. It continuously watches every trade and scores each trader on a 0–100 insider-trading likelihood scale — wallets scoring 80+ get flagged as suspected insiders on a public leaderboard.

02THE ARCHITECTURETypeScript monorepo · common / backend / frontend

One pipeline, five stages — from a raw OrderFilled log to a scored wallet on a public leaderboard, with no application-layer round-trips in the hot path.

INGEST

every OrderFilled event, as it lands

The Graph · Polymarket Subgraph

↓

ENRICH

resolve token + condition IDs

Alchemy RPC

↓

STORE

the canonical record + job queues

PostgreSQL · Drizzle · Redis

↓

SCORE

five signals, one pass, in the database

PL/pgSQL · score_wallets()

↓

SERVE

REST + live push to the leaderboard

Express · socket.io · Next.js

STACK — TypeScript · Next.js · Express · PostgreSQL · Redis · Drizzle · socket.io · The Graph · Alchemy

03THE HARD PARTscore_wallets() · PL/pgSQL · single pass

The scoring engine is one PL/pgSQL function computing five behavioral signals in a single pass — no round-trips through the application layer. Each signal contributes up to 25 points:

bet concentration

top-market volume share

market count

breadth of activity

position size

capital at risk

entry timing

how late in a market's life the wallet first traded

wallet age

gap between first USDC.e receipt and first trade

04IN THE WILDplates 01–03 · the live product

[next study — Flux→][view the source→]