Variance Signal Extraction

Variance Signal Extraction is the foundational mechanism of the protocol. It defines how Variance Markets derives structured signals from prediction markets before outcome resolution, without interacting with pricing, liquidity, or settlement.

Variance Markets continuously observes active markets on Polymarket and treats them as time-indexed data streams rather than decision outputs. The goal is not to reprice outcomes, but to identify and quantify belief divergence embedded in market behavior.

The extraction process operates exclusively on public, observable market data, including price evolution, traded volume, liquidity distribution, and reaction patterns to external information. No private trader data, identities, or intent inference is used. The protocol assumes that heterogeneous beliefs necessarily leave measurable traces in these observable variables, even when the aggregated price appears stable.

At each step, the system analyzes how the market responds to information over time. This includes identifying asymmetric reactions, delayed adjustments, partial reversions, and persistence of disagreement under similar conditions. These patterns indicate that multiple coherent belief regimes coexist within the same market, even though the base layer exposes only a single price.

Variance signals are extracted as pre-resolution metrics, meaning they are computed only while the market is live and unresolved. Once an outcome is resolved, signal extraction stops and the final outcome is used only as a settlement reference for downstream features.

The output of this feature is a set of raw variance signals that describe:

• the presence and intensity of belief divergence

• how long disagreement persists

• how belief updates differ across time and liquidity conditions

These signals are not probabilities, predictions, or alternative prices. They are structured measurements of how the market disagrees with itself prior to resolution. Raw signals produced by this feature are then passed to the normalization and scoring layers, where they become comparable across markets and usable by protocol products.