EdgeWatchFXby ErgodicLabs

The RecordMethod

Sharpe ratio in forex: what it shows and where it misleads

The Sharpe ratio is the most quoted risk-adjusted figure in trading. It is also easy to inflate: a smooth balance curve, a short record or a strategy with rare large losses can all produce an excellent number. Here is how it is calculated and how to read it.

4 min read

In short: the Sharpe ratio is the average return divided by how much returns varied, usually scaled to a year. It rewards steady results. Read it with three questions: was it measured on equity or on the closed balance, how long is the record, and does the strategy have rare large losses that the record has not met yet.

How it is calculated

Take the account’s daily returns. Divide their average by their standard deviation. That is the daily Sharpe ratio. To make it a yearly figure, multiply by the square root of the number of trading days in a year, 252 by convention. Forex trades about 260 weekdays a year, which changes the result very little.

Sharpe ratio (yearly) = average daily return / standard deviation of daily returns x √252

An account with an average daily return of 0.08% and a daily standard deviation of 0.6% has a daily Sharpe ratio of 0.133, or 2.1 a year. The original definition subtracts a risk-free rate first; for a trading account over a short period it changes little and is often left out.

What it shows

Two accounts with the same return can carry very different risk. The Sharpe ratio separates them: the one that got there with smaller daily swings has the higher figure. It works on any account size and any leverage, because doubling the position size doubles both the return and the swings.

What is a good Sharpe ratio in forex

As a rough reading, a yearly Sharpe ratio around 1 over several years is considered good for a fund, and above 2 sustained for years is rare. A short trading record with a figure of 5 or 10 is more likely to be telling you about the measurement than about the strategy. The Sortino ratio, which counts only the downside swings, is read the same way and is usually higher than the Sharpe ratio for the same record.

Where it misleads

Measured on the closed balance

The balance only moves when trades close. A strategy that holds losing positions open shows a calm balance with small daily changes, which makes the standard deviation small and the Sharpe ratio large.

Take the account above, with the same average return, but measured on the closed balance where the daily standard deviation is only 0.15%. Its Sharpe ratio becomes 8.5 a year. Nothing about the trading changed; the swings were simply left in the open positions. Measured on equity, with open profit and loss included, the figure goes back to about 2.1. See equity drawdown vs balance drawdown.

A Sharpe ratio measured on closed trades rewards whatever kept the losses open.

A short record

The Sharpe ratio is itself an estimate, and on a short record a rough one. With 120 trading days and a true yearly Sharpe ratio of 2.1, the standard error of the measured figure is about 1.5. A record that shows 2.1 could come from a strategy whose true figure is well under 1. See how many trades before a track record means something.

Rare large losses

The standard deviation treats a large loss and a large gain the same way, and it assumes nothing about what has not happened yet. A martingale or a grid can show months of small, steady gains and an excellent Sharpe ratio before the one losing run it was always exposed to. Until that run is in the record, the figure cannot see it. See reading a martingale from its statement.

Skewness and fat tails

Trading returns are rarely bell-shaped. Many strategies have negative skew, many small gains and a few large losses, and fat tails, more extreme days than a normal distribution would give. The same Sharpe ratio is less convincing for such a strategy than for one with symmetric returns.

The probabilistic Sharpe ratio

Bailey and Lopez de Prado proposed the probabilistic Sharpe ratio in The Sharpe Ratio Efficient Frontier (2012) to deal with a short record and with skewness and fat tails. It cannot see losses the record has not met. It asks: given the length of the record, its skewness and its kurtosis, how likely is it that the true Sharpe ratio is above zero, or above a chosen level?

For the account above, 120 trading days with a daily Sharpe ratio of 0.133, the probability that its true Sharpe ratio was above zero is about 93% with symmetric returns. With a skewness of -1 and a kurtosis of 6, which is closer to many trading records, it falls to about 91%, and the record needs to be longer to reach the same confidence. The same paper gives the minimum track record length: how many observations a record with a given Sharpe ratio, skewness and kurtosis needs before its returns can be told apart from zero at a chosen confidence. For this account at 95%, it is about 155 trading days with symmetric returns and about 177 with a skewness of -1 and a kurtosis of 6.

How to read a Sharpe ratio you are shown

  • Ask whether it was measured on equity or on the closed balance.
  • Ask over how many days or weeks, and on daily or monthly returns.
  • Treat a very high figure on a short record as a reason to look at how the trades were placed.
  • Read it beside the equity drawdown and the losing streaks, not instead of them.

How EdgeWatchFX uses it

The Sharpe ratio on EdgeWatchFX is measured on the daily equity index, open positions included, with deposits and withdrawals taken out so that money moving in or out does not count as return. The Evidence readout uses the probabilistic Sharpe ratio, with the record’s own skewness and kurtosis, to say how long the record needs to be. A yearly Sharpe ratio of the daily equity far above what funds have reported over years is treated as a reason for a closer look, not as a strength. The methods are listed on how EdgeWatchFX measures an account.

Figures in this note describe recorded results. None of it is a forecast or investment advice.

More from The Record

What changes when an EA goes live

A backtest answers whether a strategy could have worked. A live account answers a different question: is it still behaving like itself? Here is how to tell a…

Method2 min read