BL-SD-020: Cascade Exhaustion, Bybit

Status: RELEASED

Classification: RESTRICTED

Revision: 1

Date: 2026-09-08

Venue: Bybit Programme: I Verdict: OPEN Blocking axis: Coverage Transferability: -

---

1. Identification

The pre-registration is a premise gate, not a strategy kill sheet. Its purpose is to test whether exhaustion, measured from realised decay of forced flow, carries information beyond the fact that a cascade is simply in progress, before any profit-and-loss construction is attempted. That ordering, testing the mechanism before building the strategy around it, is the design's own requirement and is why step 1 produced a coverage finding rather than a market verdict.

---

2. Economic thesis

During a liquidation cascade, the participants generating order flow are the venue's own liquidation engine acting against price-insensitive positions, not discretionary traders responding to information. That flow has to end: once the positions being forced out are gone, the selling or buying pressure that had nothing to do with information stops, and price has no structural reason to continue in that direction. The thesis is that a dislocation measured at the point exhaustion is realised, rather than at the point the cascade is first detected, is a better predictor of normalisation back toward the pre-cascade level than a naive fade entered at detection.

Who is structurally forced to lose. Two distinct counterparties. The liquidated positions themselves, whose exit is executed by the venue engine as a taker order regardless of price, which is the definition of a forced, price-insensitive seller or buyer. And late entrants who join the cascade direction after the forced flow has already been exhausted, who are trading a move that no longer has the fuel that started it.

This venue differs from the sibling design in one structural respect that the thesis has to carry as a cost rather than assume away: Bybit does not expose a map of third-party positions the way the venue behind the sibling design does. Exhaustion here cannot be read from a position map ex ante; it has to be inferred from realised decay in liquidation intensity, open-interest flush, and order-book refill, after the fact. The thesis absorbs that disadvantage through tighter validation gates rather than through a weaker mechanism claim.

---

3. Falsifiable claim

Withheld under BL-FM-004 section 4. This is an open lead, and the pre-registered claim, its kill thresholds, and the amendment history are held rather than published, consistent with the classification policy that an unresolved hypothesis is withheld in substance so as not to hand over a live research position while giving a reader nothing they could act on with confidence. The row remains visible in the Kill Ledger under code B1 with its status stated openly.

---

4. Mathematical specification

Withheld under BL-FM-004 section 4, for the same reason as section 3. The signal, entry, exit, sizing, and cost terms that would appear here define the exhaustion detector this hypothesis is still testing, and publishing them before the premise gate resolves would publish an open research position rather than a closed finding.

---

5. Operating envelope

Withheld under BL-FM-004 section 4. The envelope parameters that would appear here, including the round-trip cost assumption and the execution channel this hypothesis is being measured against, are held pending a resolved verdict.

---

6. Instrument

Withheld under BL-FM-004 section 4. The instrument that measured this hypothesis is in-repository and carries test coverage, consistent with the programme's evidentiary standard; its identity and test count are held rather than published while the hypothesis remains open.

---

7. Results

Withheld under BL-FM-004 section 4, with one exception carried into this section because it is a fact about instrument coverage rather than about the market and is disclosed publicly elsewhere in the record: over a measurement window of 1.94 days, the pre-registered detector identified 7 cascade-and- exhaustion events across the fixed three-instrument universe, a rate of approximately 3.6 events per day. Of those 7 events, the fraction with sufficient price coverage after the exhaustion point to evaluate the pre-registered reversion metric was 29 percent. Every other quantity that would normally appear in this section, including the reversion measurement itself and any comparison against the null, is withheld, because the coverage finding below is what determined that no verdict could yet be drawn from what those numbers would show.

---

8. Verdict, blocking axis, and transferability

Verdict: OPEN. Axis: coverage. Transferability: not applicable.

This is not a kill and it is not a validation. The pre-registered decision rule anticipated exactly this outcome: below a minimum event count or below a minimum fraction of events with usable post-exhaustion price coverage, the result is indicative rather than a settled verdict, and premise evaluation does not proceed. Both conditions were met on the step-1 sample. The event count, 7 against a pre-registered floor in the tens, is underpowered on its own terms. The coverage fraction, 29 percent, is the more consequential of the two, because it is not a market finding at all. It is a property of the recording instrument: snapshots of the order book and trade tape were captured only at the moment of a liquidation event, so once a cascade's intensity decays, the events that would otherwise trigger a fresh snapshot become rare, and the price series needed to evaluate reversion at a fixed horizon after exhaustion goes dark for the majority of cases.

The distinction drawn in section 6 of BL-IN-004, between a system that is running and a system that is measuring, applies again here on a different family. The recorder is healthy. It is not yet capable of answering the question the premise gate asks. That is a statement about instrumentation, and it is why the axis of death, provisionally, is coverage rather than idea, cost, or data contamination. No transferability classification applies, because nothing has closed.

---

9. Cost and capacity analysis

Withheld under BL-FM-004 section 4.

---

10. Envelope sensitivity and reopening conditions

Withheld under BL-FM-004 section 4. What can be stated openly, because it follows directly from the coverage finding in section 8 rather than from any withheld figure, is the shape of the reopening path: this hypothesis requires an instrumentation change, a scheduled post-event capture cadence rather than an event-triggered one, before the premise gate can be evaluated at the pre-registered power. That is a recorder decision, not a signal or cost decision, and it sits upstream of anything section 10 would normally report.

---

11. Reference implementation

Withheld under BL-FM-004 section 4.

---

12. Evidence register

Withheld under BL-FM-004 section 4.

---

13. Lessons extracted

The coverage-before-verdict discipline applied here is the same discipline BL-IN-004 produced from the withdrawn dislocation entry: measure the reach of the recording instrument before trusting anything it reports, rather than after. Here that discipline was applied prospectively, at pre-registration, rather than recovered after a false verdict had already been issued, which is the improvement this entry represents over its predecessor incident. The cost of applying it early was small: one characterisation run that spent no research runway on a profit-and-loss construction that the coverage gap would have made unreliable regardless of what it showed.

---

14. Cross-references

BL-IN-004 for the coverage-before-verdict discipline this entry applies. BL-FM-004 section 4 for the classification policy governing what this dossier withholds. BL-D-004 section 5 on verdict vocabulary, specifically the distinction between an underpowered or coverage-limited result and a settled verdict in either direction. Ledger row B1, status OPEN.

---

Copyright (c) 2026 Bijective Labs. All rights reserved.