Japan’s retail FX market turned over ¥1.227 quadrillion in September—105 times its month-end open positions

Japan’s over-the-counter retail foreign-exchange margin market reported September 2026 trading volume of 12,271,914 units of ¥100 million. Written out in ordinary currency, that is ¥1,227,191,400,000,000: approximately ¥1.227 quadrillion in one month. The month-end value of long and short open positions was ¥11.716 trillion. Gross turnover was therefore about 104.7 times the positions still open at the reporting date.[1]

The ratio is striking but easy to misread. It does not mean traders held 105 times their account equity, nor that brokers earned fees on a new ¥1.227 quadrillion pool of capital. Trading volume counts repeated transactions over a month; open positions are a snapshot at the end. A position can be opened, reduced, reversed and closed many times before the final date, contributing repeatedly to turnover while leaving no exposure in the snapshot.[1]

The first investigative step is checking the unit

The Financial Futures Association of Japan publishes the figures in hundreds of millions of yen. A single reported unit is ¥100,000,000. Multiplying 12,271,914 by that unit produces the quadrillion-yen total. Omitting two zeros would turn the market into a hundredth of its actual reported size; treating the published number as yen would make it smaller by eight orders of magnitude. The conversion is part of the finding, not a cosmetic presentation choice.[1]

The same conversion applies to positions. The association reported 73,799 units of long positions, equivalent to ¥7.3799 trillion, and 43,358 units of short positions, equivalent to ¥4.3358 trillion. Its published total was 117,158 units, or ¥11.7158 trillion. The long and short components add to 117,157, a one-unit difference equal to ¥100 million that is consistent with aggregation or rounding in a dataset measured at that scale.[1]

The market ended the month net long by 30,441 units, or ¥3.0441 trillion. Long positions were approximately 63.0 per cent of the published total and short positions approximately 37.0 per cent. That is a directional snapshot of the contracts outstanding, not a forecast and not proof that retail customers as a group profited or lost when exchange rates moved.[1]

Turnover measures activity, not the amount at risk at one moment

A simple example shows why the monthly multiple can be high. A trader opens a ¥10 million USD/JPY position in the morning and closes it in the afternoon. Repeating that round trip on ten days creates far more gross trading volume than the position visible at month-end, which may be zero. Across an industry, high-frequency strategies, short holding periods, hedging and frequent reversals can produce enormous flow from a comparatively small stock of open contracts.[1]

The association’s 104.7-times ratio should therefore be read as an activity-to-snapshot comparison. It does not reveal an average holding period because the denominator is only the final day and the numerator aggregates the whole month. It does not reveal unique customer capital because notional value is larger than margin. And it does not reveal broker revenue because spreads, commissions, financing and internalisation differ by product and firm.[1]

Nor should the monthly volume simply be multiplied by twelve and presented as a forecast of annual trading. Currency volatility, policy decisions, holidays and customer behaviour can move retail activity sharply from month to month. An annualised number would be an extrapolation, not an observed fact. The defensible measurement is the September total and its relationship to the September-end position stock.[1]

Leverage rules set a ceiling, not a measurement of actual leverage

Japan requires a minimum margin of 4 per cent for individual customers in retail FX margin transactions, equivalent to a maximum leverage ratio of 25 to one. That rule helps interpret the product but cannot be used to calculate actual customer margin from the published notional positions. Some accounts may hold more collateral, different currency pairs can generate different risk, and brokers may impose requirements above the regulatory minimum.[2]

Applying the maximum mechanically to the ¥11.716 trillion position total would yield a theoretical minimum margin of about ¥468.6 billion. That is not a reported customer-balance figure and is intentionally not used as a headline. It assumes every position is margined at the minimum, ignores offsets and firm-specific treatment, and combines long and short positions that may belong to different customers. The exercise illustrates why notional exposure and customer money must not be conflated.[1][2]

The same discipline applies to the quadrillion-yen turnover. Notional volume is useful for understanding market activity and infrastructure demand. It is not equivalent to the amount customers could lose, because gains and losses depend on price changes, position direction, leverage, close-out rules and available collateral. A one-basis-point move on a notional position is economically different from the entire notional amount changing hands as cash.[1][2]

The long bias is meaningful, but only within its limits

The ¥3.0441 trillion net-long position indicates that, in aggregate, reported long notional exceeded short notional at month-end. The data page does not by itself identify the currency pairs responsible. A long USD/JPY trade is long dollars and short yen; a long EUR/USD trade is long euros and short dollars. Summing “long” contracts across products therefore describes contract direction within the reporting convention, not a single macroeconomic bet on one currency.[1]

A robust interpretation would pair the aggregate with pair-level tables, customer counts and distribution across firms. Those data could answer whether the flow was concentrated in USD/JPY, whether a small group of accounts generated most turnover and whether the long-short balance reflected many modest positions or a few large books. The flash report establishes scale but not that deeper distribution.[1]

The number of reporting operators was 46. That count is also not a customer count or a measure of competition. One operator may have vastly more accounts or volume than another. Without firm-level shares, the data cannot show whether the market’s infrastructure or counterparty exposure is concentrated.[1]

A quadrillion-yen market can still be described precisely

The useful conclusion is narrower than the eye-catching number and stronger because of that restraint. Japan’s reported retail OTC FX market processed ¥1.227 quadrillion of gross notional trading in September while carrying ¥11.716 trillion of positions at month-end. Turnover was 104.7 times the closing stock, and the position book was roughly 63 per cent long by reported notional.[1]

The turnover-to-position ratio is not a conventional velocity measure

Calling the 104.7-times figure “velocity” would imply a stock measured consistently through the month. The denominator here is only the open position on the final reporting date. If customers happened to reduce exposure before month-end, the ratio would rise even with unchanged monthly turnover; if they carried unusually large positions across the date, it would fall. The multiple is a useful comparison of flow with a closing snapshot, not an estimate that the same yen of exposure was recycled exactly 105 times.[1]

A stronger turnover measure would compare monthly volume with average daily open positions, ideally by currency pair and broker. The flash report does not provide that denominator. The article therefore preserves the date-specific language rather than converting the ratio into an average holding period or customer-churn statistic.[1]

Large gross flow changes the operational burden even when exposure closes

Transactions that disappear from the month-end book still have to be priced, margined, confirmed, recorded and monitored while they are open. A market processing ¥1.227 quadrillion of monthly notional flow imposes substantial demands on order systems, price feeds, risk limits, reconciliation and regulatory reporting even if the final position stock is much smaller.[1][2]

That operational significance is different from credit exposure. A broker’s risk depends on whether it internalises customer flow, hedges externally, how quickly positions move and the collateral available during adverse price changes. Gross turnover indicates the amount of activity passing through the infrastructure; it does not tell us the residual market risk retained by the 46 reporting operators.[1][2]

The aggregate long-short split cannot identify who holds the risk

A 63 per cent long share may be produced by many small customers or a narrow group of large accounts. It may also be distributed unevenly among brokers, with some firms carrying net-long books and others net-short books. The flash report’s aggregate does not reveal concentration, offsetting positions within individual accounts or the extent to which dealers hedged the customer imbalance in wholesale markets.[1]

Those omissions matter for stress analysis. A ¥3.0441 trillion aggregate net-long position is not automatically a ¥3.0441 trillion unhedged exposure for Japanese brokers. The customer side, broker side and external hedge book are separate layers. The reported figure establishes customer-contract direction within the association’s dataset, not the industry’s final directional risk.[1]

Precision at this scale depends on retaining the reporting convention

The one-unit discrepancy between the stated total and the sum of long and short positions is ¥100 million—large in ordinary terms but roughly 0.00085 per cent of the ¥11.716 trillion total. Reporting it openly is preferable to silently forcing the components to match. The source’s unit and aggregation method determine the apparent inconsistency; the difference is far too small to support a claim of a material reporting failure.[1]

Those figures portray a market built on rapid reuse of exposure rather than a static pool of contracts. They also show why broker analysis must keep four concepts separate: turnover, open notional, customer margin and broker revenue. The FFAJ report supplies the first two. Converting them accurately—and refusing to invent the other two—is the investigation.[1][2]

Report an error or submit an update