ForesightFlow: An Information Leakage Score Framework for Prediction Markets

Maksym Nechepurenko

Keywords: prediction markets, informed trading, information leakage, Murphy decomposition, proper scoring rules, market microstructure, blockchain forensics, Polymarket

TL;DR: The Information Leakage Score (ILS) quantifies pre-news price drift in prediction markets via the Murphy decomposition; pilot findings on Polymarket motivate a deadline-ILS extension covering documented insider-trading cases.

Abstract:

Decentralized prediction markets such as Polymarket aggregate dispersed beliefs into continuously updated price signals, but their on-chain transparency and pseudonymous participation also create unusually permissive conditions for trading on material non-public information. Recent empirical work has documented hundreds of millions of dollars in anomalous profits on Polymarket between 2024 and 2026; existing detection approaches are almost exclusively *post-hoc* and offer no actionable signal during the window when informed flow is moving prices.

We propose an information-theoretic framework for quantifying informed flow on prediction markets. We introduce the **Information Leakage Score** (ILS), a label generator that quantifies how much of a market's terminal information move was priced in before the corresponding public news event, and show that ILS admits a clean interpretation in terms of the **Murphy decomposition of the Brier score**: high ILS corresponds to *front-loaded resolution*. We specify the score's resolution-typology and operational scope conditions.

A pilot empirical study on $n{=}725$ event-resolved Polymarket markets produces structural negative findings: a resolution-anchored news-timestamp proxy does **not** separate event-resolved markets from a matched control population (Mann-Whitney $p{=}10^{-6}$, in the wrong direction), and **zero of 24** documented insider-trading cases satisfy the original ILS scope. The audit reveals that documented Polymarket insider cases are systematically *deadline-resolved* ("Will event $X$ occur by date $Y$?"), falling outside the original scope.

We accordingly extend the score to a **deadline-ILS** variant anchored at the underlying event timestamp, with a per-category constant-hazard baseline for the time-to-event distribution. The contribution is a methodologically transparent label generator with an explicit scoring-rule reading, scope conditions validated by negative findings, and an extension that closes the gap between methodology and the population in which informed trading has been empirically attested.

Submission Number: 73