A risk-adjusted return metric: excess return divided by the standard deviation of returns. It rewards a smooth equity curve, which is why it can flatter a martingale EA right up until it blows up.
also: Sharpe, Sharpe number
Updated
How much return did the strategy earn for how bumpy the ride was? It rewards a steady climb and punishes a jagged one, even when the jagged one ends higher. Read it as a smoothness score, not a profit score.
For an EA you intend to leave running unattended, the shape of the equity curve matters as much as where it ends, because a violent swing is what makes a trader intervene at the worst moment. Sharpe is the standard way to put that shape into a number — and the standard way to be misled about strategies whose risk has not been realised yet.
Sharpe ratio = (Mean return − Risk-free rate) ÷ Standard deviation of returns The denominator is the whole argument. Volatility scales with how often you measure it, so a Sharpe computed on per-trade returns, on daily returns and on monthly returns are three different numbers for the same strategy. Annualised figures conventionally come from monthly returns; MetaTrader 5's Strategy Tester does not compute it that way.
No single number describes a strategy. These metrics change how this one should be read.
Two EAs return 30% over the same year. The first climbs steadily; the second is flat and choppy for eleven months and then jumps. Sharpe ranks the first higher, because the ranking is about the spread of returns rather than their sum.
Nothing here says EA B is unprofitable — only that it delivered the same result through a ride three times as wide.
Calculation 30 ÷ 10 = 3.0 · 30 ÷ 30 = 1.0
Result Same profit, three-fold difference in risk-adjusted quality
A Sharpe figure carries no information unless you know the sampling frequency, the sample length and the tool that produced it. Establish those three, then read the number.
| Range | What it means |
|---|---|
| From MetaTrader 5's Strategy Tester | Computed from the tester's own return series, so it does not match a Sharpe calculated by hand from annualised monthly returns. The familiar 'above 2.0 is excellent' ladder does not map onto it. |
| Annualised from monthly returns | The conventional basis for the published benchmarks, and the only one those benchmarks should be read against. |
| Measured over less than about a year | Statistically noisy. A dazzling three-month figure routinely halves over a full year without anything changing in the strategy. |
| On a strategy that holds losers open | Actively misleading. Unrealised loss never enters the realised return series, so the curve looks smooth precisely because the risk has not been taken yet. |
mt5depot EA pages state which tester run each figure came from and over what window, so two published figures on this site are comparable with each other.
These strategies post near-vertical equity curves with tiny realised volatility because every losing position is held open and averaged down rather than closed. Sharpe sees only the smooth realised curve and rewards it, right up until the basket cannot be recovered and one close erases months of gains.
Volatility scales with measurement frequency, so a daily-returns Sharpe and a monthly-returns Sharpe are different animals. The strategy did not change; only the denominator did.
It is a statistic drawn from a sample. Three months of returns can produce almost any figure, and the shorter the window the more the number describes luck rather than design.
It says nothing about the worst single loss, nothing about how long a decline lasted and nothing about tail risk. It belongs next to drawdown and recovery factor, not instead of them.