Prediction markets price thousands of questions, but they do not yet have a shared way to describe what the market as a whole is saying. A reader who wants to understand the rate path, macro stress, geopolitical risk, or the distribution of attention must still move between venues, reconcile different contract rules, and decide which prices deserve weight. The Mensor is building a standard for turning those contracts into legible instruments.
This is a working paper, not a declaration that the standard has been fully achieved. It separates three things that are too often collapsed: the standard an index should meet, the system currently implemented, and the evidence required before an indicator can be treated as a professional benchmark.
The current figures are experimental research indicators. They are not approved reference benchmarks, investable products, or settlement instruments.
Why prediction markets need indices
An index is a convention: a public rule for what counts, how much each component counts, and how the result is expressed. Its value is not the presence of a formula. Its value is that different readers can inspect the same object, understand its exposure, and reproduce its value.
Prediction markets make that unusually difficult. Similar-looking contracts can resolve under different authorities, dates, thresholds, or definitions. A venue sees only its own order book. Traded volume measures activity but not necessarily the quality or executability of a price. Thin markets can look precise while resting on very little capital. A useful benchmark must therefore do more than average displayed probabilities.
The Mensor proposes five requirements.
- Cross-venue by design. An index should be able to read every venue whose contracts and data terms permit. Actual venue concentration must be reported on every observation; cross-venue design is not the same as cross-venue coverage.
- Resolution-aware. Contracts belong to one canonical event only when their resolution criteria are compatible. A disagreement over rules is not a price disagreement and must not be averaged away.
- Depth-aware. For price indices, weight should reflect executable two-sided liquidity near the midpoint rather than trailing volume alone. Displayed depth can be cancelled or manipulated, so the band, cadence, persistence test, and concentration must also be disclosed.
- Recomputable. A published observation should carry its constituent prices, depths, directions, weights, parameters, methodology version, input checksum, and calculation checksum.
- Governed. Selection rules, methodology changes, exceptions, corrections, and publication failures should be versioned and decided through a visible process.
Selection and weighting are different decisions
Selection answers whether a contract belongs. Weighting answers how much the admitted contract should influence the result. Combining those decisions creates a predictable failure: relaxing eligibility to fill a sparse basket can allow a weak contract to dominate the figure.
The Mensor selection framework begins with eligibility gates: subject fit, contract type, resolution text, tenor, venue listing status, price range, a two-sided book, an assignable direction where direction is required, and compatible rules for any cross-venue merge.
Eligible contracts are scored on depth, tradability, and relevance. Depth asks whether liquidity is both present and persistent. Tradability considers spread, executable size, quote persistence, fees, and tenor. Relevance considers subject fit, information content, attention, and whether a compatible second-venue leg exists. Membership follows the score; weight follows the book.
The operative selection document remains under formal review as of this paper. Until its approval status, implementation, public description, and observed publication behavior agree, the framework should be described as provisional.
The measurement
For a canonical binary event with compatible contracts across venues, the intended event probability is a weighted mean of fee-adjusted, de-vigged midpoints:
Events then compose into an instrument subject to concentration controls. A direction transform may be required: an escalation probability enters a risk meter directly, while a genuinely de-escalatory event enters inversely. That sign is economic data, not presentation metadata. A wrong sign changes the meaning of the index.
Not every instrument uses this engine in the same way. The Attention Index measures volume share and is intentionally volume-weighted. The Rate Path Index transforms a mutually exclusive probability ladder into an expectation. The Macro Index combines market-implied measures with public macroeconomic anchors. A suite-level methodology must state these differences rather than imply that one formula describes every product.
The instruments
Mensor Rate Path Index
The Rate Path Index expresses a market-implied expectation in its natural unit: the expected number of Federal Reserve cuts over a defined horizon. This is the most immediately interpretable professional concept in the suite. A trader can compare venue ladders, inspect repricing around macro releases, and test whether a local book is rich or cheap relative to a consolidated curve.
The next version is designed around a rolling horizon rather than a calendar year. A calendar-year expectation mechanically approaches zero as the year ends, confounding a change in outlook with the passage of time. A professional version must also preserve the full probability ladder, specify treatment of open-ended tail buckets, retain venue and bucket provenance, and publish a complete manifest.
Mensor Geopolitical Risk Index
The Geopolitical Risk Index is a meter of selected conflict and instability probabilities. Its strongest use is editorial and diagnostic: whether a broad collection of conflict contracts is repricing, which events caused the move, and where venues disagree.
Its limitation is equally important. Events with different horizons, severities, and economic consequences do not automatically form one hedgeable risk factor. One index point created by a distant invasion contract is not economically equivalent to one point created by an imminent ceasefire contract. The headline should therefore travel with its effective constituent count, largest-event contribution, venue concentration, horizon distribution, and event-level attribution.
Current coverage has been Polymarket-led. The instrument should not be described as an achieved cross-venue benchmark when its observed weight is concentrated on one venue.
Mensor Macro Index
The Macro Index is a composite monitor. It combines market-implied recession and rate-path measures with public macroeconomic anchors. Its purpose is to compress a morning macro dashboard into one reading and then show which components moved it.
It is not a pure prediction-market index. Live probabilities, market prices, and lagged or revisable government series have different clocks and different error structures. The useful product is therefore the decomposition: input timestamp, signed contribution, weight, revision policy, and comparison with a market-only version. The composite must demonstrate that it adds information beyond its dominant recession and rates components.
Mensor Attention Index
The Attention Index asks where prediction-market activity is concentrated by category. It is the one instrument in the suite for which volume is the measurement object rather than a weight on a probability claim.
This can become a valuable media and product-intelligence statistic. But the denominator is the product. Venue coverage, common units, duplicate handling, rolling-window semantics, category rules, and completeness must be proved before a share can be trusted. Incentive trading and artificial activity are still activity, but they may not represent organic attention; both interpretations should be measurable.
Corridor readings
Regional conflict readings are useful drill-downs only when they contain enough valid events to describe more than a single contract. A one-contract transformation is an indicator, not a diversified index. A corridor whose constituents do not contain the corridor in their resolution conditions is invalid and should be withdrawn or corrected rather than explained away with a footnote.
What the first audits found
The first adversarial reviews found problems that a credible index provider must be willing to publish.
- A geopolitical constituent had the wrong risk direction and was quarantined.
- A Gaza-labelled corridor contained contracts that did not measure Gaza.
- The audited geopolitical basket was single-venue in achieved coverage.
- Several corridor readings were one- or two-event products.
- The Macro Index displayed more components than its effective concentration implied.
- The Rate Path Index lacked a complete standard constituent trace and initially collapsed venue provenance.
- The Attention Index had incomplete source coverage, unverified cross-venue units, classification errors, and a query ceiling capable of silent truncation.
- Declared publication requirements existed before they were consistently enforced in the publication path.
These findings do not invalidate the thesis. They establish the difference between building an interesting calculation and operating a defensible benchmark.
The validation program
An instrument becomes useful to a professional desk only when its compression preserves information. The Mensor validation program therefore asks eight questions.
- Does the index add information beyond its largest constituent, an equal-weight basket, and simple venue baselines?
- What observable contract or asset P&L does one index point explain?
- What is the observed capture-to-publication latency distribution?
- How much historical movement came from prices, membership changes, and methodology seams?
- How much capital is required to move the index one point through price or displayed depth?
- How sensitive is the result to removing one event, changing caps, or changing thresholds?
- Can an independent calculation reproduce consecutive observations exactly from signed manifests?
- Do prospectively frozen signals demonstrate calibration, information content, or executable value after fees?
The reliability band currently used by parts of the system should be understood as a diagnostic band until it is calibrated against realised error. It is not a statistical confidence interval.
Governance, corrections, and rights
Benchmark quality is a property of the process, not the sophistication of the formula. The required operating model includes a methodology owner, a data and operations owner, and an independent reviewer; recorded approvals and recusals; a controlled hierarchy of documents; and a coded publication state for research, provisional, validated, production, degraded, suspended, withdrawn, and restated.
Wrong directions, incompatible mappings, incorrect corridors, source-attribution failures, incomplete capture, stale publication, and failed reproduction should automatically open a correction incident. Each incident should identify the affected range, corrected values or explicit non-restatement, downstream surfaces, root cause, remediation, and closure approval.
Cross-venue calculation also requires permission for the intended data use. A venue-derived figure should not be distributed commercially where the governing terms restrict index creation or redistribution and written permission is absent.
The current Mensor figures are research indicators. They should not be used as references for creating or settling a financial instrument. Any future partnership should state attribution, methodology version, venue coverage, permitted use, correction handling, and conflicts.
The opportunity
The opportunity is not to manufacture an S&P 500 analogue by declaration. It is to build the measurement layer that prediction markets lack: canonical events, comparable prices, venue disagreement, depth and provenance, transparent selection, and a public record of what changed.
The near-term product is a set of honest research indicators and decomposable dashboards. The longer-term product is benchmark infrastructure. The distance between the two is measurable, and this paper is intended to make that distance visible.