In equity markets, alpha has a workable definition. There is a market portfolio, you can measure your exposure to it, and whatever return remains after that exposure is accounted for is the part attributable to skill. The definition is imperfect, but it gives you something to subtract.
Currencies do not offer that. There is no long-only portfolio of foreign exchange that a passive investor would hold by default, because every position is a relative bet funded by a corresponding short. Long EUR/USD is not “being in the currency market”. It is a specific view on two economies, expressed as a pair.
This creates a measurement problem with real consequences. If there is no obvious benchmark to subtract, then any positive return can be presented as skill. And a great deal of what gets presented as skill in this market is something else entirely: exposure to well-documented risk premia that anyone can access systematically, dressed up as proprietary insight.
Four things that pay you without any skill involved
Decades of research and practice have identified a handful of persistent return sources in currencies. None of them are secret. All of them can be implemented mechanically.
Carry. Borrow in low-yielding currencies, hold higher-yielding ones, collect the interest differential. It works for long stretches and then loses a large portion of accumulated gains in short, violent episodes when risk appetite reverses. The pattern is so consistent that carry is often described as collecting nickels in front of a steamroller, and the description is fair.
Trend. Currency pairs exhibit medium-term momentum. Buy what has been rising over the past three to twelve months, sell what has been falling. Long flat periods, occasional strong runs, and a tendency to perform when carry is being punished, which is why the two are often paired.
Value. Currencies deviate from purchasing power parity for years and then partially revert. Extremely slow, requires patience most traders do not possess, but measurable.
Short volatility. Selling options, or any strategy whose payoff profile resembles selling options: frequent small gains, rare large losses. A great many strategies are short volatility without their operators realising it, including anything that averages into losing positions or that relies on mean reversion within a range.
A strategy whose returns are largely explained by these four things is not producing alpha. It is delivering beta to currency style factors, which has value, but it is cheap value and it should be priced as such. The important word is explained, not identical. A strategy can look nothing like a textbook carry basket and still derive most of its return from the same underlying exposure.
The decomposition you can actually run
You do not need an institutional risk system to get most of the answer.
Take your monthly returns for as long a period as you have. Alongside them, put three simple columns: the month’s return on a basic carry proxy, the month’s return on a simple trend signal applied to the same pairs, and the month’s move in a broad equity index as a proxy for risk appetite.
Then look at when your good and bad months occurred. If your strongest months cluster in calm, risk-on periods and your worst months coincide with volatility spikes, you have a short-volatility profile regardless of what your strategy description says. If your returns track the trend column, you are running a trend strategy with extra steps.
What remains after those relationships are accounted for is the candidate for genuine edge. It is usually much smaller than the headline figure, and sometimes it is negative.
The sample size problem nobody wants to discuss
Here is the part that changes how you read every track record you will ever see.
Suppose a strategy has genuinely produced an annualised Sharpe ratio of 1.0, which is a strong result. After three years of monthly data, the standard error on that estimate is roughly 0.59. The 95% confidence interval therefore runs from approximately -0.15 to 2.15.
Read that again. Three years of monthly data on a genuinely good strategy cannot statistically distinguish it from a strategy with no edge at all.
To separate an annualised Sharpe of 1.0 from zero at conventional confidence levels, you need something in the region of four years of monthly observations. For a Sharpe of 0.5, which is closer to what most real strategies deliver, the requirement scales with the inverse square of the ratio and lands somewhere around fifteen years.
Almost no track record you are shown will meet this bar. The sales pitch for alpha generation forex strategies rarely includes a confidence interval, and the reason is that the interval would swallow the claim whole.
This does not mean short records are worthless. It means they should be read as weak evidence about the future and strong evidence about something else entirely: the operator’s risk-taking, their behaviour in drawdown, their honesty about what they are doing. Those you can assess from three years of data.
Multiple testing, which makes the above worse
Now add the fact that most strategies are not discovered, they are searched for.
If you test two hundred combinations of parameters against historical data, the best of them will look excellent by chance alone, and the more combinations you test the better the winner will look. The reported Sharpe ratio of the survivor is not an unbiased estimate of anything. It is the maximum of a distribution, and the maximum of a distribution is always flattering.
The correction is conceptually simple even if the maths is not: the more variants that were tried, the higher the bar the winner must clear. Ask how many were tested. If the answer is “we don’t track that” or “just this one”, you have learned something either way.
What real currency edge tends to look like
When genuine edge exists in this market, it usually comes from one of three places, and none of them are indicators.
Structural advantage. Better execution, lower latency, access to more venues, tighter internalisation. Real, durable, and largely unavailable to retail participants.
Constrained counterparties. Someone who must transact regardless of price. Corporate hedging programmes, index rebalancing, month-end fixing flows, central bank operations. These produce predictable pressure at predictable times, and the effects have been documented in academic literature for years. They are small, and they are the closest thing to a free lunch in currencies.
Behavioural persistence. Patterns that survive publication because they are uncomfortable to trade rather than because they are hidden. Carry is the obvious example: everyone knows about it, and it persists precisely because the drawdowns are severe enough to shake most participants out.
Notice what is absent from that list. Proprietary indicators. Pattern recognition on price charts. Anything described as an algorithm without a stated economic reason for why the money should be there.
Questions worth asking
Before allocating to anything presented as a source of outperformance:
- What is the economic reason this return exists, stated in one sentence without jargon?
- Who is on the other side, and why are they willing to lose?
- What fraction of returns is explained by carry, trend, value and short volatility?
- How many variants were tested before this one was selected?
- What is the confidence interval on the reported ratio?
- What is the capacity, and what happens to the edge at ten times current size?
- What has to be true about the world for this to stop working?
A strategy that has good answers to these may still lose money. A strategy that has no answers is asking you to accept a number without a mechanism, which is a poor trade at any price.
Trading leveraged products carries a high level of risk and can result in losses that exceed your deposits. Past performance is not a reliable indicator of future results. This article is educational and is not investment advice.
| [donate]
| Help keep news FREE for our readersSupporting your local community newspaper/online news outlet is crucial now more than ever. If you believe in independent journalism,then consider making a valuable contribution by making a one-time or monthly donation. We operate in rural areas where providing unbiased news can be challenging. |
