Historical replay · paper

Protection Lab

We don't ask you to trust the words "risk management". Here are the largest catastrophes in crypto history — pick any one and see what would have happened to a virtual $100,000 under our protection versus a passive holder of the same basket. Including the cases where protection failed or made things worse: this is research, not a showcase.

How to read this

This is a deterministic replay over sourced historical data (generated 2026-08-22) — not live results, a forecast, or investment advice. The portfolio is virtual ($100,000, paper). Replay decisions are made by the system's real risk code; no look-ahead: a day-T decision sees the market as of end of day T−1. An exit from a frozen venue is recorded as an execution failure, not as protection. Mapping assumptions for historical venues are documented per scenario in the public repo.

historical crises in the library (2014–2026), each with sources

11/14

replays without a hard (−10%) drawdown of the protected basket

−13.5%

worst protected drawdown — credit/freeze channels with no price signal

risk channels the lab honestly names as not covered

Historical crises

Pick a catastrophe

Synthetic stresses

Scenarios that have not happened yet

Failure families generated by the parametric engine — including ones where protection is knowingly powerless or harmful. The same deterministic code as the historical replays.

Breakdown

Method

How this is computed

Two runs of one basket: "Passive" does nothing; "Protected" is the daily cycle, with decisions made by the system's real risk code (the two-tier kill-switch ladder and the depeg gate), not by a copy of its thresholds.
The accounting does not flatter: exit haircuts and gas are paid, exits execute at shock-day prices, and "capital saved" can be negative — we show those cases first.
The uncovered channels are named: credit default with no price signal; a depeg of the cash asset itself; withdrawal unavailability while NAV is intact; intraday cascade speed; locking losses on transient discounts.
Provenance: every historical scenario carries a UTC timeline, ≥3 sources and mapping assumptions in the public repo; contested figures are flagged inside the scenario itself.