Skip to main content
Prediction markets have become a remarkably efficient way to price beliefs about future events. But the signal they produce is trapped in hundreds of isolated, binary contracts – fragmented, noisy, and constantly expiring. Every other asset class solved this problem the same way: with indices. Equities got the Dow in 1896, volatility got the VIX in 1993. Event risk has had no equivalent. Belief Index is that layer.

Prediction Markets as a Distinct Risk Domain

Prediction markets trade contracts that pay $1 if a specified event occurs and $0 if it does not. The market price of a contract is the crowd’s implied probability of the event. Taken together, these markets price a risk domain with genuinely distinctive characteristics: Event probabilities move on political shifts, regulatory decisions, and macroeconomic surprises – factors largely orthogonal to equity beta or credit spreads. That makes them a genuinely novel signal for anyone measuring, researching, or managing real-world risk. But the discontinuous, binary nature of individual contracts makes any single market a poor measurement instrument on its own. That is exactly the problem indexing solves.

Why the Signal Needs an Index

From Binary Readings to a Continuous Series

Individual prediction markets resolve to exactly $1 or $0. Read in isolation, a single market is a discontinuous series that one surprise can terminate at zero – informative about one event, useless as a measure of a theme. Aggregating multiple markets into a rules-based basket smooths this discontinuity – the same principle that makes an equity index a better barometer than any single stock.
Single market: A contract on whether the Fed cuts rates in March trades at $0.85. The Fed surprises with no cut. The contract goes from $0.85 to $0.00 overnight – the series simply terminates at zero.Index of 7 equally weighted markets: The same Fed surprise takes one constituent from $0.85 to $0.00, but the other six markets in the basket are unaffected (or may even move higher on the news). The index moves from roughly 65 to 53 – a sharp, informative repricing of the theme, not a terminal event.If only one of the seven constituents resolves unfavorably and the rest are unchanged, the maximum move from that single event is bounded at approximately 14.3% of the basket (1/7) – not 100%.
The more uncorrelated markets in the basket, the smoother and more interpretable the aggregate series. This is what turns settlement noise into a benchmark.
Diversification reduces the magnitude of any single event’s impact on the level, but does not eliminate event risk from the series. If multiple correlated constituents resolve unfavorably at once, the level can move substantially. See Risk Factors.

Category-Level Questions Deserve Category-Level Answers

Each Belief Index series tracks a coherent theme – interest rate expectations, election outcomes, macroeconomic data releases, or crypto regulatory events. The index answers questions at the category level rather than the event level: This is how measurement works everywhere else in finance: analysts quote the S&P 500, not a survey of 500 tickers; they quote the VIX, not an options chain. Thematic aggregation is what makes a market citable.

One Series Instead of Dozens of Order Books

Following a theme through raw prediction market data means:
  • Monitoring dozens of order books with varying liquidity profiles
  • Tracking resolution dates and settlement rules for each individual market
  • Splicing series together by hand as markets resolve and new ones list
  • Deciding – ad hoc, and differently from everyone else – how to weight each market’s reading
A Belief Index series reduces that to one published number per theme, computed by fixed rules, continuous across resolutions, and comparable across time. Two researchers citing the same series at the same window are, by construction, citing the same number.

Continuity Across Resolutions

Prediction markets expire; themes do not. When constituents resolve, a self-built tracker breaks – the series ends, and any continuation involves arbitrary splicing choices. Belief Index handles this systematically: settlement values fold into the level as constituents resolve, and Perpetual Series replace resolved markets on a scheduled, rules-based review cadence with chain-linked continuity. The published series measures the theme, not the turnover.

Where the Signal Is Irreplaceable

Not every theme needs a prediction market benchmark equally, and it is worth being precise about where the value concentrates. Some risk domains already have deep, listed instruments that imply probabilities. The market-implied path of U.S. interest rates, for example, can be read from fed funds futures and overnight index swaps – instruments with enormous depth and decades of history. For themes like these, a Belief Index reads as a cross-check from a different crowd: a useful second opinion, aggregated by a different mechanism from a different participant base, and validatable against the listed curve. Other risk domains have no instrument at all. There is no futures curve for a presidential primary, no swap that prices a regulatory ruling, no listed market in government-shutdown risk or geopolitical escalation. For these themes – elections, policy and regulatory outcomes, court decisions, conflict trajectories – prediction markets are the only venue where the probability trades, and a Belief Index series is the only continuous, citable measurement of it. The exposures that increasingly drive portfolio outcomes – elections, policy shifts, geopolitical ruptures – sit almost entirely in the bottom rows. That gap between what moves portfolios and what has a quotable instrument is the measurement problem Belief Index exists to close.

A History That Cannot Be Rebuilt

Prediction market order books are ephemeral. Quotes are not archived the way equity prices are; resolved contracts disappear from venue interfaces; and any attempt to reconstruct a themed series after the fact forces splicing and selection choices that contaminate the result with hindsight. Belief Index levels are computed and recorded as of each valuation window, under a methodology whose revisions are dated and prospective-only. That produces something no amount of later effort can recreate: a point-in-time record of how markets actually priced a theme, at each moment, with no lookahead. For anyone backtesting a signal, studying how event expectations move, or citing a probability in published work, the value is not just today’s level – it is the accumulated series behind it, captured live, window by window, since inception. Every published window extends a dataset that cannot be backfilled. This is how equity research got its foundational datasets: not from clever reconstruction, but from someone keeping the tape.

A Number You Can Put Your Name Next To

Raw prediction market prices are public. What an institution cannot easily produce from them is a number it can rely on – one that survives an investment-committee question, a model-validation review, or a footnote in published research. A hand-picked quote from one contract’s order book reflects an analyst’s private judgment: which market, which venue, which moment, how weighted against related contracts. A Belief Index level replaces that private judgment with a public rule: published selection criteria, a published weighting formula, a replicable valuation methodology, and governance around revisions and corrections. Two people citing the same series at the same window cite the same number, and either of them can show exactly where it came from. That is the practical difference between information and a benchmark. The information is free. The benchmark is what makes the information usable inside an institution.

Who Uses Index Data?

Different users rely on different parts of what an index publisher produces. The live level is only one of them.

Risk & Strategy Teams

Rely on: the live series and its citability. A monitor for event risk orthogonal to the market data they already consume – election scenarios, policy risk, and regime-change indicators as continuous series they can put in a risk dashboard or committee deck and defend, because the methodology behind the number is published and replicable.

Systematic Researchers

Rely on: the point-in-time history. A backtest is only as good as its data discipline. Because levels are recorded as of each valuation window under prospective-only methodology revisions, the series is free of the lookahead and splicing bias that invalidates self-built prediction market datasets.

Researchers & Academics

Rely on: methodological stability. Continuous, citable time series of market-implied event probabilities – replicable from public data, and free of the ad-hoc construction choices that make self-built datasets hard to defend in peer review.

Journalists & Analysts

Rely on: attributability. A single number for how markets price a macro or political theme – “easing expectations rose X% this month” – rather than a hand-picked quote from one contract’s order book. Attribution: Source: Belief Systems (beliefsystems.xyz).

Product Issuers

Rely on: governance. Rules-based, transparently computed underlyings for event-linked products – methodology, levels, constituent files, and calculation-agent services under license via Index Services.

Allocators

Rely on: independence. As prediction-market strategies emerge, evaluating any of them requires a passive, rules-based reference constructed independently of the managers being measured. A benchmark is the yardstick fiduciary evaluation runs on.

What the Indices Do Not Do

Honesty about limitations is as important as articulating what the indices measure:
  • They do not eliminate binary risk from the domain. The underlying markets still settle at exactly $0 or $1. Diversification smooths the series; it does not make event risk continuous.
  • They are not execution prices. Levels are theoretical values computed from order book midprices. They do not include spread, slippage, or market impact, and thin books make that gap material. See Risk Factors.
  • They do not predict outcomes. An index level is a measurement of the market’s current implied probabilities – a reading of consensus, not a forecast by Belief Systems. The consensus can be wrong.
  • They depend on their data source. The indices are computed from a single category of venue; venue disruptions or oracle failures affect the published series.
  • They are not investment products. The published indices are informational. No public offering of any security is being made, and index data is not investment advice. A private, invitation-only Alpha Program exists separately, under the conditions in Disclosures – and history suggests that is the natural order: in equities, the index came decades before the index fund.

Benchmark Governance

Methodology versioning, corrections policy, and publication integrity.

How It Works

The production pipeline from market data to published level.

NAV Methodology

The transparent valuation methodology.

Risk Factors

Methodology limitations and data caveats.