Evidence · Methodology · Reproducibility

Every claim is reconcilable to a documented event.

147 cascade events. $23.96B liquidation notional. 91.0% cross-venue confirmation. All audit-grade. All reproducible.

Cascade Events Documented
Real, audit-grade, reproducible
Liquidation Notional Analyzed
Across 8 deep-archive markets
Cross-Venue Confirmation
As of — · archive expanding live
Marquee Events

The cascades we flagged before they hit.

Documented, cross-venue confirmed, reproducible during evaluation. Each flagged by CRI with material lead time before forced selling accelerated.

4 Jun 2026 · BTC · headline event
$1.575B
Largest documented cascade in the archive. Flagged by CRI structural pressure 214 minutes before forced selling accelerated — a 3.5-hour operational lead.
214 MIN CRI LEAD
2 Jun 2026 · ETH
$1.218B
Second-largest documented cascade. Flagged 230 minutes early — the longest lead time in the marquee set.
230 MIN CRI LEAD
17 May 2026 · ETH
$1.043B
Largest single-asset ETH cascade from the May contagion window. Cross-venue confirmed.
130 MIN CRI LEAD

Also documented · BTC $1.041B (163-min lead, 17 Apr 2026) · BTC $957.2M (170-min lead, 7 Jun 2026). Most recent · ETH $445.9M (135-min lead, 15 Jun 2026 · cross-venue confirmation pending daily enrichment).

Concentration

The BTC + ETH duopoly.

BTC 42.4%
ETH 28.2%
Other 29.4%
BTC 42.4%
ETH 28.2%
All other markets 29.4%

Two markets drive 70.6% of all liquidation notional.

Across the full archive, BTC and ETH together account for 70.6% of liquidation notional analyzed. Cascade risk in perpetual futures is concentrated — which is exactly why deep-archive coverage of the majors matters more than breadth across thin markets.

It is also why cross-asset contagion is the second-order risk: when the two dominant markets move together, the system moves with them.

Lead-Time Distribution

Warning measured in hours, not seconds.

Distribution of CRI lead time across documented cascades. The majority land in the 60–160 minute window — enough time for a risk officer to act before forced selling accelerates.

CRI lead time before cascade onset

60–160 min · highlighted
0-303%30-609%60-9019%90-12027%120-15022%150-18013%180+6%
Cross-Asset Contagion · Statistical Proof

Contagion is measured, not asserted.

Permutation-Tested Cluster Detection

When one market breaks, we measure whether the others are about to.

Two layers of evidence. Real-time detection — the terminal catches contagion as it fires; most recently on 15 June 2026, four assets (ETH, SOL, BTC, BNB) shared a cascade onset inside a 14-minute window. Statistical validation — cross-asset clusters are tested against a null model built by randomly shuffling cascade onset times, repeated 10,000 times on the calibrated archive. Four clusters meet strict criteria (2+ symbols, 20-minute peak window, legs ≥$100M each); the validated clustering occurs roughly 600× more often than chance, at p<0.0001.

Method · permutation test · 10,000 iterations · shuffled-onset null · 4 validated of 11 detected · validated through 28 May 2026 · p<0.0001
4
Validated clusters · 11 detected
p<0.0001
Significance threshold
~600×
Over random expectation
10,000
Permutation iterations
Audit-Grade Reasoning

Every event carries its own proof.

01

Source-hashed

Every cascade event has a source_hash and inputs_json capturing the exact data that produced it.

02

Reproducible

Re-run any event during the evaluation period and get the same result. Deterministic, manifest-backed.

03

Reconcilable

Every published figure traces to SQL. Built so a risk team can verify any claim against its underlying data.

Verify any claim during a paid evaluation. Reconcile any event to its source data.
Request institutional briefing →