Transparent by design

Methodology

Information asymmetry in an event contract cannot be observed directly, but the structural properties that permit it are written in the contract's own text: who produces the outcome, how many people know it early, whether they can trade on it, how it gets disclosed, and whether it can be manufactured. This taxonomy follows the outcome-maker classification that the Anti-Corruption Data Collective validated against 435,000+ settled Polymarket markets and $54B of volume (April 2026, https://acdatacollective.org/work/insider-risks-in-polymarket-political-markets/). Related quantitative work: the Information Leakage Score framework (arXiv:2605.00493) and the CFTC's proposed public-interest factors for event contracts (RIN 3038-AF65).

Modelclassifies against public anchors
Codeverifies quotes and computes every number
Chainattests the result and method version

The five dimensions

D1 · Outcome Control

weight 30%

What produces the outcome of this event?

Classify by the most concentrated locus of control over the OUTCOME itself. Work down this ladder and stop at the first match: (1) can a single person settle the outcome by their own discretionary or unilateral act? -> level 5; (2) does a bounded body or organization decide it by formal action (vote, order, ruling, corporate act)? -> level 4; (3) is it the aggregate of mass public behavior, or of an open competitive contest among many independent participants? -> level 3; (4) is it an official statistical procedure measuring something external? -> level 2; (5) otherwise a natural or market process -> level 1. Controlling the procedure, timing, or disclosure around an outcome is NOT controlling the outcome — those belong to D4 and D5. If the contract text is genuinely ambiguous between two levels, pick the higher (riskier) one.

Anchor levels for Outcome Control
LevelLabelDefinitionReference examples
1natural_processThe outcome is produced by a natural or physical process, or by large-scale market data that no identifiable party controls.
  • Will BTC close at or above $120,000 on Dec 31? (exchange market price)
  • Will the recorded high temperature in NYC exceed 100°F in August?
2statistical_procedureThe outcome is produced by an official statistical procedure following a predefined, published methodology.
  • Will August CPI year-over-year exceed 3.0%? (BLS methodology)
  • Will the unemployment rate fall below 4.0%?
3public_collectiveThe outcome is the aggregate of behavior by a large public (mass voting, mass purchasing), or of an open competitive contest among many independent participants. No single participant meaningfully controls it.
  • Who wins the presidential election?
  • Will the film gross over $500M in its opening month? (or a viewership/ratings figure)
  • Will Club Alpha win the league fixture? (a live competitive contest)
4institutional_decisionThe outcome is produced by a decision of an institution, committee, or organization: a vote, an order, a corporate action, a ruling.
  • Will the FOMC cut rates at its September meeting?
  • Will Company X announce layoffs before Q3 ends?
  • Will country X conduct a military strike before the deadline?
5individual_willThe outcome is decided by a single person's discretionary choice — what they say, post, wear, or announce — or by a unilateral official act that one person alone can perform (a pardon, a veto, a resignation).
  • Will person X tweet the word 'Y' this week?
  • Will the couple publicly announce a breakup this month?
  • Will the president pardon person X? Will the CEO resign?

D2 · Knowledge Circle

weight 20%

Between the moment the outcome is effectively determined and the moment it is publicly disclosed, how many people already know it?

Estimate the size of the early-knowledge window population from the structure of the event, not from any specific organization's current staffing. The outcome is "determined" when the person or body who controls it has settled on it, NOT when it is executed: a discretionary act the actor has already decided to perform has an early-knowledge window of at least one person (classify D2 at level 3 or higher for individual-will outcomes). Only when no party can know the outcome before it is public — it is produced and published in the same instant by a process nobody steers — is the circle empty (level 1).

Anchor levels for Knowledge Circle
LevelLabelDefinitionReference examples
1noneNo early-knowledge window exists: the outcome becomes known to everyone at the moment it is produced.
  • Live sports scores; on-chain or exchange price closes
2embargoed_fewA small number of people know early inside a formal embargo/lockup regime with access controls.
  • CPI figures inside the BLS before the 8:30 ET release
3handfulA handful of people know early with no formal control regime: the decision maker and their immediate circle.
  • A celebrity's personal announcement, known to the person and confidants
4dozensDozens of people know early: a department, a planning group, an editorial or product team.
  • A corporate layoff plan weeks before announcement
  • Google's 'Year in Search' list, known to its editorial team pre-publication (traded on by an insider; CFTC charges, May 2026)
5large_or_unboundedHundreds of people or an unbounded, unknowable set know early: multi-agency operations, long planning chains.
  • Military operation preparation chains (ACDC found 52% longshot win rates in military/defense markets vs a 35% price-implied ceiling)

D3 · Insider Tradability

weight 25%

Are the people who know the outcome early effectively prevented from trading on it?

Judge the constraint regime that applies to the likely knowers on the venue where this contract trades. The venue appears in the contract data when known; when it is unspecified, assume an offshore venue that most legal regimes do not clearly reach (level 3 or higher unless the knowers are independently constrained). Cross-dimension rule: if D2 is level 1 AND D5 is level 2 or below, classify D3 as level 1 — with no early-knowledge window and no realistic manufacturer, there is no insider to constrain. If D5 is level 3 or higher, a party can hold foreknowledge WITHOUT any disclosure lag (they intend to produce the outcome); classify D3 on the constraint regime facing that party. This rule is also enforced deterministically by the scoring engine.

Anchor levels for Insider Tradability
LevelLabelDefinitionReference examples
1prohibited_enforcedClear legal prohibition, an enforcement track record, and a regulated venue. Or no early-knowledge window exists at all (see guidance).
  • Corporate insiders trading their own securities on a regulated exchange
2prohibited_weakly_enforcedA legal prohibition arguably applies but enforcement is rare or novel.
  • US-regulated event contracts after the first CFTC insider prosecutions (2026)
3policy_onlyOnly employer NDAs or platform terms constrain the knowers; no specific law clearly covers trading this contract on this venue.
  • The Google 'Year in Search' insider: bound by employer confidentiality only, traded $1.2M in profits across 23 markets before CFTC fraud charges
4ambiguousThe applicable law or jurisdiction is unclear; constraints exist on paper somewhere but plausibly do not reach the knowers or the venue.
  • Most offshore prediction-market events with foreign-official knowers
5unconstrainedNothing prevents the knowers — including the outcome's decision maker — from trading the contract themselves.
  • A celebrity betting on their own announcement timing

D4 · Disclosure Synchronicity

weight 15%

Is there an official moment when everyone learns the outcome at the same time?

Classify the predictability of THIS outcome's release moment, not the reliability of the channel: an official channel with a fixed calendar does not make a discretionary act's timing scheduled — if the disclosing party chooses when the channel fires, that is level 3 or 5, not level 1. Use the contract's own resolution source when one is named.

Anchor levels for Disclosure Synchronicity
LevelLabelDefinitionReference examples
1fixed_scheduleA fixed, pre-announced release time on an official channel; everyone learns simultaneously.
  • BLS releases at 8:30 ET; FOMC statement at 14:00 ET
2official_roughly_scheduledAn official channel with an approximately known release window.
  • Quarterly earnings during a scheduled reporting season
3official_unscheduledAn official channel exists but the timing is at the discloser's convenience.
  • A company that 'will announce in the coming weeks'
4media_diffusionNo single official channel; the outcome diffuses through media reporting and confirmation over hours or days.
  • Geopolitical events confirmed progressively by wire services
5discloser_controlledThe disclosure moment itself is controlled by a party with a stake in the outcome.
  • A celebrity choosing when to post their announcement

D5 · Outcome Manufacturability

weight 10%

Could someone cause the outcome to happen in order to win a bet on it?

This dimension is an amplifier: it asks about deliberately producing the event, not merely knowing it early. Consider cost and coordination required.

Anchor levels for Outcome Manufacturability
LevelLabelDefinitionReference examples
1impossibleNo party can plausibly cause the outcome.
  • Macroeconomic statistics; weather
2prohibitively_costlyTheoretically possible but the cost dwarfs any plausible bet size.
  • Moving the global BTC price to cross a strike
3multipartyManufacturing the outcome requires coordination across multiple parties.
  • Steering an institutional decision
4small_groupA small group acting together can produce or time the outcome.
  • A department accelerating or delaying an announcement
5unilateral_cheapOne party can produce the outcome unilaterally at trivial cost.
  • Posting a tweet; using a phrase in a speech (a White House teleprompter operator profited $100k+ on speech-word markets; referred to CFTC, 2026)

Scoring

Each dimension's matched level maps to a 0–100 score via a fixed weighted formula, then to a band:

Score bands
BandScore range
low024
moderate2549
elevated5074
high75100

Circuit breakers

A circuit breaker overrides the weighted score with a floor when a specific, high-risk structural combination is matched.

CB-1 — floor 80

Condition: D1 == 5 && D3 >= 3 (graduated floor: D3=3 -> 80, D3=4 -> 90, D3=5 -> 95)

A person who decides the outcome and is not clearly barred from trading it makes every other dimension nearly irrelevant. ACDC observed exactly this: individual-decided markets look unremarkable on average while hiding a small set of perfect-win wallets. The floor is graduated by D3 because the policy_only/ambiguous boundary is the most subjective call in the library — a single hard threshold would put a 20+ point cliff on it.

CB-2 — floor 75

Condition: D5 == 5 && (D3 >= 3 (graduated: 75/85/90) || D2 == 1 (floor 85))

If the outcome can be manufactured unilaterally by someone not clearly barred from trading it, the market is an invitation to trade against its own author. The D2 == 1 alternative exists because a manufacturer needs no disclosure lag: they hold foreknowledge of an outcome they intend to produce, which the knowledge-circle dimension cannot see.

Conservative default

A dimension that defaults to level 4 is always flagged as a caveat on the result. When a dimension cannot be determined from the contract text, we never guess low. Ambiguity resolves toward higher risk, and the result is always flagged.

What this is not

  • Not a prediction of any outcome, and not trading advice.
  • Not live insider detection: the score reflects the structural risk of the event type, not activity in a specific market right now.
  • Not an accusation against any person or account.