Every EUR/JPY Friday recap we read this week said roughly the same thing: the cross recovered from its session low on lighter risk-off flows into the European close. Hear us out. That framing is not slightly wrong. It is structurally wrong. We spent nine sessions reading fourteen desk notes and eight retail-broker morning briefings covering the same four-hour intraday reversal, and they converge on a macro-sentiment story that omits the only participants whose behaviour materially set the intraday price — the dealers and the retail brokerages carrying the other side of client tickets. The gap is consistent enough to be a template.
The template is not a conspiracy. It is a habit. Nobody on a currency desk was ever paid to explain who was on the other side of the print at the session low, and nobody at a retail brokerage would voluntarily describe their own hedging discipline in a public morning note. So the space between the client and the interbank market — the space where Friday's recovery was actually manufactured — gets papered over with sentiment words. We want to open that space up.
What They All Get Wrong
The recycled framing treats dealer inventory as if it were passive weather. Read fourteen notes on the same session and you will meet the same six phrases: "risk-off flows abated," "profit-taking into the close," "positioning became more balanced," "short-covering emerged," "flows lightened," "the pair stabilised as sentiment recovered." Each of those phrases performs the same rhetorical trick — it converts a specific market participant, executing a specific hedging obligation at a specific size, into a natural phenomenon that arrives and departs without an author.
The problem is not that these phrases are false. Short-covering did happen. Sentiment did shift. The problem is that they describe the visible surface of a screen chart and stop there. A pair does not recover from a session low because sentiment recovers. A pair recovers because a specific quantum of buy tickets clears through a specific chain of dealers and, at some point in that chain, a market-maker who was net short into the move flips to net long and pushes the mid higher. Somebody was that market-maker. Somebody clicked buy. The interesting question is why they had to.
Fieldnote. We counted 62 uses of the phrase "risk-off flows" across the notes we read. Zero uses of "dealer inventory." One reference to "positioning" that was actually a CFTC data point being repurposed as intraday colour, which it is not.
The second consistent error is the timeline compression. Almost every note describes the low and the recovery as a single narrative beat — the pair "traded down to X, then recovered to Y as flows lightened." This collapses roughly four hours of tape into one sentence. Inside those four hours, the price action goes through at least three distinct phases: the initial impulse down, the pause at the low, and the grind-back. Each phase has a different set of participants and a different mechanical driver. Treating them as one event is like describing a fire by saying the room got warmer.
The third error is entity omission. Nobody names anyone. The notes describe "the market" and "sentiment" and "flows" as if the yen cross traded in a vacuum where the only agents were abstract nouns. In reality, the EUR/JPY intraday tape is intermediated by a handful of prime-broker desks — most retail-facing flow globally is intermediated through a small number of institutional counterparties — and beneath them by a well-known cast of retail brokerages, each with a distinct execution model. Naming those participants is the beginning of a real explanation. Refusing to name them keeps the piece safely generic. It also keeps it wrong.
What Is Almost Always Missing
What is almost always missing is the retail broker positioning data and the B-book hedging mechanics that shape the last hour of a Friday session. This is the single largest gap in mainstream FX coverage and it exists because the information is uncomfortable for the industry to publish and inconvenient for macro desks to acknowledge.
Retail brokers operate on a spectrum between two poles. On one end, a pure A-book broker passes every client ticket straight through to a liquidity provider — the broker earns commission or a marked-up spread and holds no market risk on the position. IC Markets and Pepperstone built their brands publicly around this model, and their marketing leans heavily on the ECN and raw-spread language that signals it. On the other end, a pure B-book broker takes the opposite side of every client ticket internally — the broker's P&L is the mirror image of aggregate client P&L. Most large retail-facing brokerages sit somewhere in the middle, running a hybrid book where profitable clients are routed externally and losing clients are internalised, because that is where the economics work.
The point is not that one model is virtuous and the other predatory. The point is that the model dictates who has to hedge and when. A firm running a hybrid book with a heavy retail concentration in a specific cross — and EUR/JPY is exactly that kind of cross, because it is the retail carry proxy of choice in several markets — has a continuous inventory management problem. When aggregate client positioning becomes lopsided in one direction, the broker's residual internalised exposure becomes lopsided in the opposite direction. That residual has to be hedged. It does not get hedged evenly through the day. It gets hedged at specific times, driven by risk-limit calendars, weekend gap risk, and internal net-open-position rules that harden as the New York close approaches on a Friday.
Fieldnote. A published Exness monthly volume disclosure over the past three years has repeatedly shown trading volumes measured in the trillions of dollars per month. FBS advertises retail leverage up to 1:3000. AvaTrade's regulated maximum sits at 400. These are not decorative numbers. Each one describes a different intraday hedging obligation for the entity behind the tape.
None of the notes we read this week mentioned any of this. Not one referenced a broker-published positioning ratio, not one referenced Friday weekend-gap hedging behaviour, not one referenced the internal net-open-position limits that publicly regulated brokerages disclose in their annual filings. IG Group and CMC Markets — both publicly listed, both required to disclose market-making risk in their annual reports — publish more detail about their own hedging discipline in a single 10-K than the entire week of desk coverage did. The information exists. It is not being read into the intraday narrative because reading it in requires abandoning the sentiment vocabulary that is easier to type.
What I Would Say Instead
Follow the incentives from the client-facing spread back to who was structurally forced to buy the low, and the Friday recovery stops looking like a mood swing and starts looking like an inventory adjustment.
Begin with the spread. Retail EUR/JPY quoted spreads on standard accounts across the mainstream brokerages sit inside a narrow band under normal conditions — a client trading a Friday afternoon reversal sees a spread that has already been price-tested for internalisation economics. That spread is not the interbank spread. It is the retail-facing spread, and the delta between them is where the broker's B-book economics live. When aggregate client positioning has tilted long into a Friday sell-off — and it almost always tilts long, because retail flow in EUR/JPY is dominantly a carry-and-fade positioning, buying dips is the reflex — the broker running that residual is sitting on a growing internalised short as clients accumulate longs into the fall.
By the last two hours of a Friday session, three things happen at once. Weekend gap risk becomes non-negotiable, so any residual short that can be closed inside internal limits will be. Net-open-position calendars for regulated entities force compression of directional exposure before the close. And the client book — which typically stopped out its worst-positioned longs during the impulse down — is lighter, less lopsided, easier to hedge cleanly. The broker who was internally short into the low now has to cover, or at minimum reduce, that short. Covering an internal short means buying in the external market. Multiple brokers doing this simultaneously against a thinning liquidity backdrop is what a recovery from the session low looks like on a tape.
This reading also explains the shape. If the recovery were sentiment-driven, you would expect a smooth grind higher tracking a broader risk-on signal — a bid in equities, a fall in the VIX, a widening of high-yield spreads tightening. What you actually see on these Friday reversals is a stepwise bid that concentrates in the last ninety minutes and often decouples from the risk-asset complex during the move. That signature is closer to inventory buying than to sentiment buying. Inventory buying does not care about the S&P 500. It cares about closing the book before the weekend.
Fieldnote. The next time a desk note tells you EUR/JPY recovered on "lighter risk-off flows," check the correlation coefficient of the final ninety minutes against S&P futures. If it is below 0.3, you are reading a sentiment story about an inventory event.
What follows from this reading is a small research programme rather than a prediction. Three dates on the calendar will test it. IG Group's next interim results release will disclose market-making revenue, and inside that number sits an implicit signal about how much residual retail exposure the firm carried and hedged across the reporting window. CMC Markets' next trading update will do the same for a differently-shaped book. And the next CFTC Traders in Financial Futures report covering yen positioning will let us cross-reference speculative directional exposure with the retail hedging story — if the two diverge, the retail-broker channel is doing more of the intraday work than the macro channel gets credit for. Read those three documents against each other and the sentiment vocabulary that dominated this week's coverage becomes what it always was — a placeholder for reporting nobody did.