PM Atlas PM Atlas home

Concept · information-theory

Noise Decomposition

Separating observed price movements into components attributable to genuine new information versus random microstructure noise · used to assess how much of a market's short-term variance reflects real signal. Closely tied to market-microstructure theory (Kyle, Glosten-Milgrom, Roll variance ratios) but specialized for PMs by the binary-payoff and short-horizon structure.

Key insights

In their words

Noisy traders fund the probability space rather than serve as exit liquidity.· functionSPACE
Together explain 87.3% of variance on Kalshi.· Nam Anh Le
Persistent underconfidence in political markets where prices compress toward 50%.· Nam Anh Le

Where it matters

Noise decomposition is the diagnostic that lets a builder say honestly what their market is good at. Aggregate Brier scores treat all variance as forecast quality; decomposition separates "the market correctly updated on news" from "the market thrashed because a 5¢ tick read as a 1pp probability change" (semantic tick size). For PM platforms, this matters in three ways: (1) it lets you exclude microstructure noise from accuracy claims, (2) it identifies categories where noise dominates and informed flow is absent (signaling minimum-viable-liquidity failures), (3) it surfaces the platform-specific microstructure effects (Kalshi's large trades amplify underconfidence; Polymarket's don't) that determine whether a contract behaves like a real probability or a token price. For Dekant, the distribution-market surface should naturally decompose noise across the distribution, with traders rewarded for variance compression (entropy reduction) rather than for being on the right side of a binary threshold.

Connections

Platforms linked to this concept

Related concepts

Sources

Open in the interactive atlas →