BTC Five-Minute Liquidity Observations: A Reproducible September 16 Sample

Collection method and denominator

On September 16, 2026, we made six paired UP/DOWN requests approximately 45 seconds apart through PolyCerno’s public read-only snapshot adapter. It normalizes Polymarket Gamma market metadata and CLOB asks. Requests were sequential, not simultaneous. The collection interval was 2026-09-16T04:46:37.034Z to 2026-09-16T04:50:30.622Z.

This convenience sample covers 2 market windows and 12 returned snapshots from 12 requests. 0 requests failed; 1 returned snapshots were excluded from cost summaries. Only non-empty, untruncated books with known fees and timestamps, marked VERIFIED by the adapter at capture, are usable. VERIFIED checks source context and timing; it is not a fill guarantee or an independent audit of every upstream value.

Observed cost distribution by total budget

Each budget uses the same 11 usable snapshots. Min, median and max describe this sample only; no population estimate or confidence claim is implied.

Complete means the supplied model supported the requested total budget; it does not mean a real order was filled. A zero median does not imply every observation had zero slippage.

Phase coverage, not a phase ranking

Phase groups are imbalanced and too small for comparative claims. A future edition needs a predeclared sampling window and substantially broader independent coverage before discussing typical behavior.

Observation and exclusion ledger

Download and independently recalculate

Download observed inputs (JSON) · Download calculated results (JSON)

Input file SHA-256: 8968097efcbbba3c45b6a66183b3ef2e14ed1a2a96d93e2ce6ebcf4b51829add. Values are frozen historical observations, not current quotes. The files contain public market data, not account or user records.

The companion recorded-book case walks through two inputs. Use the calculator to reproduce the arithmetic. To cite this edition, link this page and retain its date, 11-snapshot denominator and limitations.

Limitations and future editions

There are only 11 usable snapshots from 2 windows. Repeated observations are correlated, and UP/DOWN quotes were not fetched simultaneously. This is not a random sample, a long-term market benchmark, a latency study, a settlement reconstruction or evidence of a profitable strategy. The sample cannot establish which phase is normally more liquid.

We walk visible asks with fractional shares and the captured fee curve, using a total budget including estimated fees. Slippage is relative to the best ask in that same snapshot. Per-level fee rounding differs from possible real per-match rounding. Minimum size, tick/order constraints, cancellations, queue position and changing prices are not simulated. Unavailable values are excluded, never replaced by zero.

This edition is frozen. Future observations must have a new dated input artifact, documented capture interval and reviewed method. There is no automatic production collector or auto-publication job.