F-0013 · first recorded 2026-08-21 · updated 2026-08-24
In four scan transitions, wallets holding free collateral disappeared more often; disappearance cannot be read as liquidation
PRELIMINARY
| Observation | across 4 deep-to-deep transitions, of wallets within 5% of liquidation: 62.7% of those WITH free collateral were absent at the next scan, against 45.1% of those with none. The sign is opposite to the hypothesis that free collateral aids survival. |
|---|---|
| Sample | 2,302 wallet-transitions within 5% (3,331 within 10%) across 5 deep scans of the frozen 5,395-wallet universe, Hyperliquid BTC, 2026-08-19 to 2026-08-21 |
| Method | track each wallet's BTC position between consecutive DEEP scans only; classify as absent, survived-improved or survived-worsened; split by whether withdrawable was above zero |
| What is not established | only 4 transitions, 12.9h apart, so wallets are counted repeatedly and these are not independent observations. One venue, one asset, 54 hours. The sign is clear only within this sample; neither its persistence nor its magnitude is established. |
| Who else publishes this | no provider publishes wallet-level survival at all, because none archives per-wallet account state over time |
| Evidence | product/calibration.py, ~/genesis-evidence/liqmap/snapshots-liq2.jsonl |
The proposed product was a base rate: “wallets in your situation escaped 3 times in 10.” It required that wallets which survive be distinguishable from wallets which do not.
They are not, and the reason is more interesting than the failure.
What was measured
| within 5% of liquidation | n | absent at next scan | improved | cut size |
|---|---|---|---|---|
| free collateral > 0 | 1,109 | 62.7% | 63.8% | 39.4% |
| free collateral = 0 | 1,193 | 45.1% | 70.8% | 37.1% |
Within these four transitions, wallets with the means to defend themselves disappeared more often than wallets without.
Why, and why it kills the base rate
“Absent” conflates two opposite events. A wallet holding free collateral can close its position and walk away. A wallet at zero free collateral is stuck — it cannot exit without realising the loss, and it cannot move its liquidation price either.
One plausible reading is that free collateral permits voluntary exit, while trapped wallets may persist because exiting realises a loss. The archive cannot establish that interpretation: its observation is disappearance, not the reason for disappearance.
Our archive cannot separate the two. A wallet is recorded only while it holds a BTC position, so a wallet that closed out and a wallet that was liquidated both vanish identically. More collection of the same state snapshots does not fix this, because the distinguishing event is not in those snapshots.
What it does not mean
It does not mean free collateral is useless — F-0001 stands, and the venue’s own figure is still the only correct way to compute it. It means presence in a later scan is the wrong outcome variable, and any product built on it would have been measuring exits and calling them escapes.
The remedy, and it is specific
Liquidation events are published, and 0xArchive’s free tier carries them alongside trades, funding and open interest. Joining their events to our states labels each disappearance as liquidation or exit, and the base rate becomes answerable.
That join is the next thing to try. Until it exists, the escape-rate product is not merely unproven — it is unmeasurable.
The methodological warning
The same analysis run on the fast tier gives 13.3% absent instead of 49.1%, because a wallet leaving the top-300 by notional looks identical to a wallet leaving the market. Had this been run on the hourly data — the obvious thing to do, since there is six times more of it — it would have produced a confident number that was almost entirely scan-set churn. F-0011 again, in a new place.