F-0012 · first recorded 2026-08-21
The vulnerability figure Isobath publishes is very nearly a constant, so it cannot discriminate between clusters
MEASURED
| Observation | across 433 published clusters, cannot_defend_pct has median 100.0 and 71% sit at exactly 100%. Splitting at the median leaves 29% below and 0% above. thinly_defended_pct behaves the same: median 100.0, 27% below, 0% above. |
|---|---|
| Sample | 433 clusters published across 23 hourly and deep snapshots, Hyperliquid BTC, 2026-08-19 to 2026-08-21 |
| Method | contrast pre-condition P1 of market/CONTRACT-calibration.md, run before any outcome was examined |
| What is not established | three days, one venue, one asset, and drawn largely from the fast scan tier whose selection bias is F-0011. A market regime with materially lower leverage could restore the spread. What is not in doubt is that no spread exists in the period recorded. |
| Who else publishes this | no liquidation product publishes a defensibility figure at all, so none has had occasion to discover that theirs would not vary either |
| Evidence | market/CONTRACT-calibration.md §3, product/calibration.py, ~/genesis-evidence/calibration/calls.jsonl |
The number on the front of this product — the share of a cluster’s notional held by wallets with zero free collateral — is nearly always 100%. Not usually high. Nearly always the ceiling.
Why that matters more than it sounds
It is still true. Every one of those wallets genuinely cannot move its own liquidation price, and that remains an observed fact about the market worth publishing. F-0001 is untouched: the obvious arithmetic still misclassifies one wallet in five, and the venue’s own figure is still the only way to get it right.
But it cannot rank anything. A quantity that is 100 for almost every cluster cannot say which cluster is more vulnerable than another, which is what the word implies to a reader. And it cannot be calibrated — CAL-1 set out to test whether higher vulnerability predicts a larger reduction in a cluster, and there is no higher group to compare against.
It is a description of leverage, not a reading of today. If 96% is what BTC looks like on an ordinary Thursday, then presenting it as a live figure that changes invites a reader to infer something from its level. There is nothing to infer.
What this does not mean
It does not mean the product is wrong, and it does not mean the measurement should be removed. The honest version of the claim is narrower and still unusual:
Almost all forced exposure near liquidation is undefendable, almost all the time — and nobody else measures this, so nobody else knows.
That is a statement about the market. It is not a signal.
What would change it
A regime in which traders hold meaningful free collateral against near-liquidation positions would restore the spread and make CAL-1 answerable. Recording continues for exactly that reason: the record costs nothing to keep and the question becomes live the moment the market provides a contrast.
thinly_defended_pct was the obvious fallback and fails identically. The variables that do vary
— cluster notional, wallet count, distance — are not what the product claims to measure, and a
result on those must never be reported as though it answered the vulnerability question.
How it was found
By running the calibration contract’s own pre-condition before freezing the contract, on data already collected but never examined for outcomes. Written after the fact, this would have been a rationalisation; written before, it is a kill condition firing as designed.