Two checks before comparing prediction-market prices

A cross-venue comparison needs both a clock test and a contract test. Vultax audited 295 shared-family observations from 48 hourly collection cycles. Every response span was below five seconds, but only 100 were within one second. The median span was 1.105 seconds. Those figures describe the collector's observations, not how fast either exchange matches orders.

Passing a timing gate still leaves the harder question: do the prices refer to the same outcome? In a preserved October 2026 Fed example, Polymarket asks about the size of a rate change while the Kalshi contracts ask whether the resulting rate exceeds a threshold. Joining them by meeting name would create a comparison before proving economic equivalence.

For each shared family and cycle, subtract the earliest successful response time from the latest. All timestamps come from the same collector host. The maximum sampled span is 1.854 seconds. Tightening the admissible span from five seconds to one discards 195 of 295 family-cycles.

Download table figures:CSVJSON with sources

Download chart figures:CSVJSON with sources

The Polymarket 25-basis-point-cut outcome refers to the change in the upper target-range bound relative to its pre-meeting level. Kalshi ticker KXFED-26OCT-T3.75 resolves Yes when the resulting upper bound exceeds 3.75%. The No side therefore covers multiple lower or equal levels, rather than just a 25-basis-point cut.

Under an explicitly hypothetical 4.00% starting bound, a 25-basis-point cut and a 50-basis-point cut both satisfy No-above-3.75%. Only one satisfies the 25-basis-point-cut outcome. That counterexample is sufficient to reject the proposed one-to-one mapping. The second preserved Kalshi threshold, 4.00%, illustrates that a family can contain several level contracts with the same meeting label.

Download table figures:CSVJSON with sources

Keep the venue, exact condition or ticker, token side, outcome predicate, units, boundary inclusivity, scheduled decision, rule version, resolution source and exceptional conditions together. Add request-start and response-end timestamps, book quantity, fee schedule and the proposed equivalence test. A headline or family label is a discovery aid, not a primary key.

The fixture bundle preserves five Polymarket October outcome identifiers and two Kalshi threshold identifiers with retrieval times. It marks the cut-versus-threshold pair as rejected rather than silently treating both as a bet on the Fed. It also records the distinct closure and fallback provisions instead of assuming the same calendar label gives the same settlement risk.

The input is the first recorded book cycle in each hour on 17–18 September 2026. Only groups containing both venues enter the 295 shared-family count. Rolling crypto contracts contribute fewer shared cycles than the fixed event families. Repeated markets are dependent; this is not a random sample of all exchange traffic.

Response timestamps bound when our collector received data. They do not prove when every displayed order was last changed, clock accuracy at an exchange, or sub-second causal leadership. A minute-cadence archive also misses moves between samples. This article reports no arbitrage profit, venue superiority or executable spread.

cross-venue-timing.csv · contract-fixtures.json · 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.