A market end date is not a payout clock

Vultax audited a retrieval slice of 121,468 Polymarket listings whose scheduled end dates fell from 11 through 18 September 2026. The open query was exhausted; the closed query was capped at 120,000 returned records, so this is not a platform census. After separating recurring up-or-down contracts, 121,432 records remained. Of those, 1,468 were still open when retrieved and 50,334 were marked closed before their listing end date.

The practical distinction is between a calendar field, a rule condition, a market closure and a redeemable or received payment. An end date can name an election day while the written rules wait for a person to assume office. An early close can reflect an outcome becoming certain before the stated date. Neither observation establishes an operational failure.

The closed-market cursor chain reached the stated retrieval cap; the returned listing slice is not exhaustive. We made 1215 cached, cursor-paginated requests and deduplicated market IDs. The snapshot spans 2026-09-19T23:22:50.611549+00:00 to 2026-09-19T23:33:14.309598+00:00; it is not an atomic snapshot.

Download table figures:CSVJSON with sources

Download chart figures:CSVJSON with sources

Among the 69,630 non-recurring closed records with a nonnegative measured gap, the median is 3.98 hours and the 90th percentile 5.14 hours. 678 of those gaps exceed 24 hours, and 251 exceed 72 hours. Those are conditional listing-to-closure gaps, not wait-time estimates for all users.

Open markets are excluded from that median because they have no observed closing timestamp. Reporting the conditional median without the open count can obscure unresolved observations. Setting open-market gaps to zero would be worse. Treating the listing end as a universally valid event-completion time would also impose a definition the rules do not guarantee.

The retrieved next-prime-minister-of-Sweden market carries a 13 September 2026 end date. Its written condition waits for the next individual to be officially appointed and assume office after the parliamentary election; an interim or caretaker prime minister does not count. The rule also has a later fallback date. That condition is not identical to the election day's midnight timestamp.

This is a reviewed example of a clock-definition mismatch, not an estimate of how common such mismatches are. It explains why reporters should read the rule before calling an open listing overdue. A genuine payout-delay study needs the time the resolution condition became knowable, proposal and challenge milestones, finality, and a claim or payment event.

The earlier Vultax article measured scheduled-end-to-closure gaps for a closed-market, nine-topic sample ending 10 September 2026. This follow-up uses a later date window and explicitly collects open records as well. It exposes the selection and clock limitations rather than presenting the earlier median as a direct measure of money reaching users.

The source is the Gamma markets keyset endpoint with separate closed=false and closed=true queries, limit=100, end-date bounds and tags enabled. Each cursor is checked for repetition, IDs are deduplicated, and returned end dates are validated against the window. Slugs containing updown or up-or-down are kept in a separate recurring group. Missing clocks remain missing. Category summaries in the export use an explicit priority order when tags overlap.

The retrieval can miss removed or unlisted markets. The capped closed query follows the endpoint's default ascending-ID order, rather than a random sample; its fractions and medians must not be generalized beyond the returned slice. Fields may change during pagination, and a market can transition from open to closed between the two queries. API timestamps have no independent claim-transaction verification in this study. The output is a reproducible field audit, not a service-level agreement or an accusation of withheld payments.

closure-market-results.csv · results.json · METHODS.md · Reproduction bundle. The bundle contains the calculation, frozen analysis inputs, results and a SHA-256 manifest.

Reuse the figures with a link to this study and the stated population and observation window. Vultax welcomes corrections and independent methods reviews. A citation or collaboration does not require a positive conclusion or a reciprocal link.

A publisher can choose a small set of upcoming events and define the reader's actual question before they finish. Vultax can record listing changes and finality milestones prospectively; the publisher can verify when the external event became known and document user experiences with consent. Publish all selected cases, including uneventful and unresolved ones. That would turn a convenient clock subtraction into a more useful settlement story.