Currency Momentum (Cross-Sectional FX Trend)

Thesis (edge)

Currency markets move in long, slow swings. A dollar bull market can run for years. A yen weakening trend can persist for a decade. The idea behind currency momentum is simply to notice which currencies have been winning and to keep backing them, while shorting the ones that have been losing.

Why would this work at all in a market dominated by sophisticated banks and central banks? The usual explanations are behavioural and structural. Investors are slow to update their beliefs about a country's economic path, so news gets priced in gradually rather than instantly. Real money flows, such as pension funds rebalancing their currency hedges, take months to complete and push in the same direction the whole time. And central banks, which are the largest players, are explicitly not profit-maximizing. They lean against moves gently and predictably, which leaves a trend in place rather than snapping it back.

The result is that past FX returns carry a little information about future FX returns. Not a lot. But enough, historically, to build a diversifying portfolio around.

Where it works (regimes)

The strategy works when macro conditions are diverging clearly. One central bank is hiking while another is cutting, one economy is booming while another stalls, and the currency market grinds steadily in one direction for months. The 2014 to 2015 dollar surge and the 2022 tightening cycle are the archetypes.

It works badly when everything is pinned. Between roughly 2010 and 2021 most developed central banks held rates near zero simultaneously, macro divergence collapsed, and currency trends became short and choppy. Momentum in FX made very little money in that decade, which is a useful and honest data point: this is not a strategy that works in all weather.

It also fails at turning points. The moment a trend exhausts, momentum is by construction maximally long the thing about to fall. The transition from a trending regime to a reversing one is where the losses are concentrated.

Signals

• Past return: the core signal is total return over the last 3 to 12 months. Total return means the spot move plus the interest earned or paid. This matters more than beginners expect. A currency can be flat in spot but paying 5 percent interest, which makes it a clear winner even though the chart looks like nothing happened.
• Cross-sectional ranking: each month, sort the universe by past return. Buy the top group and sell the bottom group. Because the book is long some currencies and short others, it is roughly neutral to the dollar itself, which removes a large source of noise.
• Time-series confirmation: optionally require that a currency is also above its own longer-term trend versus the dollar before going long. This filters out cases where a currency is merely the least bad in a universe where everything is falling.
• Volatility normalization: divide the signal, or the position size, by recent volatility so that a naturally jumpy currency does not swamp the portfolio.

Portfolio construction

Rank the universe monthly. Take equal risk, not equal notional, in the long and short legs. A long basket of three and a short basket of three is typical for the ten major currencies, and going much finer than that just spreads the same signal across more transaction costs.

Because currency momentum and currency carry frequently point at the same names, check the overlap explicitly. A high yielding currency that is also strengthening will show up in both books. If you run both strategies you may believe you have two positions when you really have one, doubled.

Rebalance monthly. Weekly rebalancing adds turnover without adding much information, because the underlying macro signal simply does not change that fast.

Risk model

The characteristic loss is a sharp trend reversal. Currency momentum tends to have negative skew of its own, and it can crash in the same weeks that carry crashes, because a global risk-off event simultaneously ends the trend and unwinds the funding.

Central bank intervention is a discrete jump risk that has no equivalent in equity momentum. A finance ministry can decide on a Tuesday to defend its currency, and a trend that has been intact for a year can reverse ten percent before you can act. The Swiss National Bank removing its euro floor in 2015 is the canonical example. Any position sizing model built purely on historical volatility will have said that move was essentially impossible.

Set a hard limit on single-currency risk, use volatility targeting at the portfolio level, and accept that stop losses in FX can be gapped straight through.

Costs & implementation

G10 FX is one of the cheapest markets in the world to trade, which is the strategy's main practical advantage. Spreads are tight, liquidity is deep, and the trade can be expressed through forwards, futures or spot with a rolling hedge.

The real implementation cost is the monthly roll of forwards, and the discipline of computing returns correctly including the interest component. Emerging currencies are a different story entirely: wider spreads, thinner books, and settlement complications that can eat the whole edge.

Turnover is low. Capacity is high. This is one of the few systematic strategies where a large book is not obviously self-defeating.

Failure modes

• Ignoring carry in the return calculation. This is the single most common mistake and it corrupts the ranking entirely.
• Overfitting the lookback window. If 9 months works beautifully and 6 and 12 months do not, you have found a quirk of your sample, not a market truth.
• Believing the strategy is uncorrelated with carry. Very often it is the same trade with a different label.
• Assuming the historical Sharpe ratio still applies. The evidence of decay since 2008 is real, and the strategy went nearly nowhere for a decade. Any honest expectation should be lower than the published backtests.
• Sizing on quiet-period volatility ahead of a policy shock or an intervention.

Our Notes & Suggestions

The strongest case for currency momentum is not as a standalone strategy. On its own its return is modest and its drought periods are long enough to test anyone's patience. Its real value is as a partner to carry. Momentum tends to cut exposure to a high yielder that has already started falling, which is exactly the situation where the carry book is most exposed. Combining the two signals, rather than running them separately, historically produced a better shape of returns than either alone.

Start with the ten major currencies only. Add emerging markets later, if at all, and only once you are honest about whether the extra return is compensation for genuine skill or simply for illiquidity and crash risk.

And set expectations properly before allocating capital. This is a strategy that can be flat for years and still be working as designed. If you cannot tolerate that, you will abandon it precisely at the point where the regime turns back in its favour.