Priorbook · Methodology

How the fair-value model works

Priorbook’s premise is that a single venue’s quoted price is one noisy signal among several. The composite model blends five independent layers into one fair-value estimate, so a “mispricing” means a venue has drifted from the consensus of price, polling, history, news and money flow — not just from one other exchange.

Status: this is the target design. The live terminal today (see /markets) shows raw venue prices only — the composite blend below isn’t computing live yet. This page describes what we’re building toward next.

The five layers

Live cross-venue price

30%

Kalshi, Polymarket & adjacents, sub-second

The raw quoted price on each venue, normalized for fee schedules and contract equivalence before blending. Raw spreads can lie, so this layer isn't taken at face value.

Polling & expert forecasts

22%

Public polling, forecaster panels

Public polling and expert-forecaster panels for the underlying question, weighted by recency and historical pollster accuracy.

Historical base rates

20%

Outcome distributions since 2014

How often similar-shaped questions have resolved a given way historically — a prior that keeps the model honest when the other layers are thin or noisy.

News & NLP sentiment

18%

Realtime ingest, entity-linked events

Realtime news ingest, entity-linked to the contract, scored for directional sentiment on the underlying event.

Wallet / smart-money flow

10%

Opt-in wallet graph, fund-grade sizing

Position flow from tracked, opt-in wallets and accounts — a smaller weight since it's a lagging, thinner-coverage signal today.

Worked example

Illustrative numbers, not live data — this shows the arithmetic the model will run once it’s wired to real polling, base-rate and news feeds.

Market: Fed cuts ≥25bp · December meeting · Kalshi quotes 0.34

LayerWeightLayer valueContribution
Live cross-venue price30%0.320.0960
Polling & expert forecasts22%0.450.0990
Historical base rates20%0.400.0800
News & NLP sentiment18%0.440.0792
Wallet / smart-money flow10%0.430.0430

Composite fair value

0.40

Edge vs. Kalshi (0.34)

+5.7 bps

Read-only by design. Priorbook never touches your funds and isn’t a venue — trade on whichever platform you choose. The model is an analytics layer on top of Kalshi, Polymarket and adjacent books.

Every layer auditable. Once live, any consensus number will decompose back to its five input layers, and weights will be configurable per market class.