Listing and admission policy
Which scientific hypotheses we list, which we refuse, and why. You can hold us to it.
1. The admission standardFalsifiable, with the settlement rule fixed in advance.
A hypothesis is listed only if it is falsifiable with a resolution criterion fixed in advance. Someone can state, before listing, the finding that would count against it: the measurement, the population, the threshold and the protocol that would settle it. A hypothesis no conceivable result could contradict can never be corrected by evidence, so it is refused.
Falsifiability is about testable content. It does not mean the position expires on a date, and an evidence cutoff for one weekly report does not close the market to later discoveries.
- Population, intervention, comparator, outcome. A hypothesis states these where they apply, so neither side reads the resolution rule differently.
- Fixed in advance. Wording and criterion are locked at listing. A materially different hypothesis is a different market, listed beside the old one.
- Refuting evidence named. The listing record says what observation would count against the hypothesis.
2. What we do not listFive refusals: issuer, patient, unlawful, violence, vague.
These exclusions are policy, not examples.
- Single-issuer clinical, FDA, financing, or solvency milestones. A contract that turns on one company's trial result, regulatory decision, financing, or solvency concentrates risk in a single actor whose insiders can see the answer first. Those listings are out. Diversifying the wording or inserting an AI intermediary does not change that.
- An identifiable patient's health, death, or private medical outcome. Excluded on privacy, adverse-incentive and public-interest grounds. We use public, lawful sources, and unpublished patient-level data is not accepted as evidence.
- Harmful or unlawful experiments. Excluded. Activity that is illegal or dangerous stays that way, and no scientific framing changes it.
- Assassination, terrorism, or war-linked events. Excluded for the same reason. A hypothesis that turns on violence against a person or a conflict outcome is not a scientific question this product lists.
- Vague or untestable hypotheses. Excluded at admission. If no result would count against the hypothesis, there is nothing for a round to report.
3. Candidate hypothesis types
The first column describes hypothesis types, not approved products. A type that can apply still has to pass admission with a specific wording and a fixed resolution rule.
| Candidate hypothesis type | Posture | Reason |
|---|---|---|
| Independent materials or energy benchmark with measurable commercial relevance | Eligible to apply | Measurable reference and a potentially demonstrable economic use |
| Broad scientific reproducibility index | Eligible to apply | Diversification may reduce single-actor influence; publication selection still matters |
| Specific preregistered replication | Conditional | A clear protocol helps, but a small investigator group can influence timing and outcome |
| A general theory judged only by an internal AI opinion | High uncertainty | Weak external verification and a possibly weak economic connection |
| Single issuer's trial, FDA, financing, or solvency milestone | Excluded | Concentration in one issuer, with insiders able to see the answer first |
| An identifiable patient's health, death, or private medical outcome | Excluded | Privacy, adverse incentives, control, and public-interest concerns |
| Harmful or unlawful experiments | Excluded | Statutory event-contract restrictions and other laws may apply |
| Assassination, terrorism, or war-linked events | Excluded | Statutory event-contract restrictions and other laws may apply |
| Vague or otherwise untestable hypotheses | Excluded | No round could ever say anything about them |
4. Independence and manipulation resistanceThree providers are not three administrators.
The panel estimate is the main listing risk. Three commercial model providers are not three independent administrators when one operator controls the prompts, the evidence selection, the retries and the publication. An average smooths differences, but shared evidence or a shared framing can produce the same error three times.
- Separate the administrator from the maker. Whoever runs the panel stays independent of proprietary trading, with real information barriers. Different wallets are not a barrier when one person controls all of them.
- Publish the methodology. What is averaged, admissible evidence, the cutoff and the publication round are stated before listing.
- Fix the valid-run rules. Averaging two, swapping a provider, dropping an outlier or rerunning an answer changes the method unless a disclosed contingency allows it.
- Keep the record. Provenance, arithmetic and rule following are recorded per round and published with it.
5. What every listed market carriesA short record per hypothesis.
The template alone does not justify a listing. Each market keeps its own short record:
- The exact proposition, its resolution criterion, and the evidence cutoff rules.
- The admission review and what it turned on.
- The commercial rationale and the economic-use case.
- Admissible sources, and how contrary evidence, retractions and corrections are handled.
- Who could affect the referenced outcome, and the trading restrictions that follow.
- The jurisdictions admitted for that hypothesis, where restrictions vary.
If no conceivable finding would count against a hypothesis, it is not listed. A hypothesis that turns on a single issuer, a private patient outcome, unlawful conduct, or violence is refused regardless of how testable it looks.