A plain-English guide to retrace patterns: cascade, intervention, and regime shift.

A large FX move retraces. The question is how, and how fast. The paired analysis today, USD/JPY retraced 72% of Thursday's collapse, reads the retrace shape as diagnostic of what drove the original move. This piece sets out the framework. Three characteristic shapes appear on the tape after a large single-session move; each identifies a specific driver.

The observation is empirical: the same shape of retrace shows up after the same kind of driver, across decades and across G10 pairs. The framework does not tell you the direction of the next trade; it tells you which class of trade produced the move you are watching retrace, and by extension which framework to apply to the remaining tape.

Shape one: the cascade retrace

Signature. 30 to 50 percent (occasionally up to 75 percent in a fully-flushed cohort) retrace over one to three sessions. The retrace runs fast and stalls; the tape subsequently trades range-bound around the new post-retrace level for several sessions before the underlying differential reasserts.

Mechanism. A positioning-liquidation cascade (mechanic explained in the positioning liquidation reference) leaves the previously-crowded positioning book substantially lighter after the cascade. The dealer book is delta-neutral; the retail stops have been cleared; the leveraged accounts have been forced out. When the retrace begins, it faces a much thinner book of resistance than the pre-cascade tape did. Bidders and sellers rebalance around a new equilibrium level determined by the underlying macro differential, and the tape trades there until fresh positioning builds or fresh news arrives.

Why it retraces fast. The move that produced the cascade was not an assessment of the underlying value; it was a mechanical liquidation of a positioning book. Once the book is cleared, there is no fundamental reason for the price to stay at the extreme. The retrace is the return to macro fair value, which the tape can execute quickly because there is no resistance in the way.

Diagnostic feature. The retrace happens without a corresponding move in the underlying macro variables. Rate differentials do not move; the dollar-side leg is quiet; other G10 crosses do not participate proportionally. The retrace is pair-specific because the positioning was pair-specific.

Shape two: the intervention retrace

Signature. 20 to 40 percent retrace over two to four weeks. Slow, uneven, punctuated by consolidation ranges. Often includes a period where the tape holds within a tight range for several sessions before a slow drift back begins.

Mechanism. Unsterilised FX intervention (mechanic explained in the BoJ toolkit reference) produces a first-order spot effect (the intervention's dollar sales support the intervened-currency at spot) and a second-order monetary effect (the yen created by the intervention loosens the domestic monetary base). The first-order effect fades over weeks as the intervention flow dissipates. The second-order effect strengthens over the same window as the loosened monetary base contributes to a weaker yen. Net: the immediate move is sustained for a period, then slowly gives back.

Why it retraces slowly. The intervention is an operational commitment by the central bank. It signals ongoing sensitivity to the currency level and creates expectation of further action if the pair returns to the threshold. That expectation itself resists a fast retrace; accounts hesitate to re-enter the trade at pre-intervention levels because they anticipate a fresh intervention. Only when time passes without follow-up action does the pair reclaim the differential drift and slowly resume.

Diagnostic feature. The MoF (or relevant intervention authority) has publicly acknowledged the intervention within 24 to 48 hours using post-action language. In Japan's case: "we conducted decisive action" or equivalents. Absence of such language in the 96-hour window after a suspected intervention is a strong signal against the intervention hypothesis.

Shape three: the regime-shift retrace

Signature. Minimal retrace, 5 to 15 percent maximum, over any subsequent window. The tape extends in the direction of the shock and holds; whatever retrace occurs is consolidation, not reversal.

Mechanism. A fundamental regime shift (surprise central-bank policy change, materially surprise macro print, credible geopolitical development that alters the outlook) resets the market's baseline expectation. The move that expresses the new baseline is the new fair value, not a positioning event. Positioning that had been calibrated to the old baseline is now materially wrong; new positioning has to build to the new baseline over the coming weeks.

Why it barely retraces. There is no macro reason for the pair to return to the pre-shock level. The underlying differential has moved; the flow of new positioning is aligned with the direction of the shock; the marginal buyer or seller at the new level is transacting on the new baseline, not against it. Retraces that do occur are technical consolidations that produce entry opportunities in the new direction, not reversal signals.

Diagnostic feature. The underlying macro variable has moved. Rate differentials are different; the fundamental data was materially surprising; the geopolitical development has moved the relevant risk premium. If the macro has moved and the pair has followed, regime shift is the leading hypothesis.

Reading a live retrace: the decision tree

Applied to a fresh large move, the framework produces a diagnostic decision in three questions.

Question one: has the underlying macro moved correspondingly? Check rate differentials, related prints, geopolitical variables. If yes, regime-shift is leading. If no, the move was not fundamental and the retrace will tell you which of the other two.

Question two: has the central bank or MoF made a public statement using acknowledgement language? Historical intervention acknowledgements have come within 48 hours. If yes, intervention retrace is leading. If no through 96 hours, intervention retrace is unlikely; cascade retrace is leading.

Question three: does the retrace shape match? Cascades produce fast, 30 to 50 percent retraces. Interventions produce slow, 20 to 40 percent retraces. Regime shifts produce minimal 5 to 15 percent retraces. The shape confirms the hypothesis; a shape that does not match forces re-evaluation.

Common misreadings

Reading intervention when the retrace is fast. A cascade retracing 40 percent in two days is often reported as "the intervention is fading" when there was no intervention. The fast retrace is the tell that the driver was positioning, not policy.

Reading cascade when the retrace is slow and shallow. A regime shift consolidating for a week is often read as "the cascade retrace is starting" when the cascade never happened and the consolidation is just the tape absorbing the new baseline. The shallow retrace is the tell that the driver was fundamental.

Attributing a retrace to a specific catalyst that happened during the window. A retrace that plays out over the same days as an unrelated data print or Fed communication tempts commentary to attribute the retrace to the data. The retrace shape often reveals that the mechanism is unrelated to the coincident news; correlation is not causation in either the original move or the retrace.

Where the framework fits

The retrace framework is a diagnostic layer that sits on top of the primary driver frameworks: yield differentials (macro), positioning liquidation (mechanical), intervention/policy (institutional). It does not replace any of them; it tells you which one is applying in a specific post-move window. Getting the diagnosis right lets you apply the correct framework to the following two weeks of tape.

Related pieces:

• A plain-English guide to positioning-liquidation events. The cascade mechanic in detail.
• A plain-English guide to FX intervention. The seven-step MoF-BoJ escalation framework.
• A plain-English guide to yield differentials and FX carry. The primary driver framework the retraces resolve back into.
• USD/JPY retraced 72% of Thursday's collapse. The paired analysis applying the framework to the current tape.

What the framework does not do

It does not identify the driver ex ante. Only after the retrace begins can the shape be measured. Before the retrace, the analyst has to work from the pre-move positioning setup, the presence or absence of acknowledgement language, and the macro data window. The retrace framework confirms or refutes the ex ante hypothesis; it does not generate one from nothing.

It also does not handle blended drivers. Real moves in G10 FX are often mixes: a cascade that coincides with a small fundamental repricing, an intervention that lands on top of a macro shift. The framework diagnoses the dominant driver, and treats blended cases as needing case-by-case adjustment rather than off-the-shelf classification. The three shapes are the pure cases; real markets often produce impurely-shaped hybrids.