A track record shows up with a Sharpe ratio printed beside it, and that one figure does most of the selling. It looks like the only line on the page that cannot be spun, because a formula produced it rather than a marketing team.
It can absolutely be spun. The ratio is a measurement of an equity curve, and it inherits every weakness in the curve underneath it: how long the record ran, how often it was sampled, and whether the trades inside it were ever fillable.
Here is what you will be able to do by the end: ask three questions about any Sharpe ratio you are shown, and know which metric to reach for when the answers come back thin.
Key Findings
- Definition: the Sharpe ratio is average return above a cash benchmark divided by the standard deviation of those returns, so it scores the smoothness of the ride rather than the size of the profit.
- Origin: William F. Sharpe introduced it as the reward-to-variability ratio in 1966 and, revisiting it in the Journal of Portfolio Management in 1994, stressed that the published figure is an estimate whose reliability rests on the length of the record.
- Main distortion: scaling a short record up to an annual figure by the square root of time overstates it when one period's return resembles the last, which Andrew Lo demonstrated in the Financial Analysts Journal in 2002.
- Prerequisite: the ratio is only as honest as the fills behind it, so an equity curve built from prices that were never actually available describes an account nobody could have held.
What does the Sharpe ratio actually measure?
Return per unit of volatility, measured against doing nothing. Take the account’s returns over regular periods. Subtract what cash would have paid you over those same periods. Average what is left, then divide by the standard deviation of those period returns.
That denominator is the whole argument. It grows whenever results scatter, so the ratio rewards a calm climb and punishes the same profit delivered in lurches.
I think that emphasis is right, and most traders underrate it. The account you can stay invested in is the one that does not make you quit in month four. A curve that doubles and halves on the way to a good year is a curve most people abandon before the good year arrives.
How is it calculated on a trading account?
Four decisions, and every one of them changes the answer.
Pick the sampling interval first: daily, weekly or monthly returns on the account balance. Pick the benchmark next, normally the yield on a short-term government bill for the same period, because that is what your capital could have earned without risk. Average the excess returns. Divide by their standard deviation.
The fourth decision is the one people skip. Pick the window and disclose it, because a ratio without a start date and an end date is not a measurement, it is a headline.
Why does a short record flatter the number?
Because the ratio you see is almost never the raw one. A monthly figure gets scaled up to an annual figure by multiplying by the square root of twelve, and that step carries an assumption: that each period’s return is independent of the one before it.
Trading returns often are not. Trends persist, losing streaks cluster, and a strategy that leans on one market condition produces months that resemble their neighbours. Andrew Lo worked through the consequences in the Financial Analysts Journal in 2002 and showed that the square-root rule can overstate an annualised Sharpe ratio substantially when returns are serially correlated.
Sharpe made the quieter version of the same point himself in 1994, when he described the published ratio as an estimate rather than a fact, with an error band that shrinks only as the record lengthens. A short sample does not give you a precise answer about a short period. It gives you an imprecise answer, dressed as a precise one.
Sharpe, Sortino, profit factor or drawdown: which one answers your question?
Different metrics fail in different places, which is the argument for reading more than one.
| Metric | What it measures | Best question for it | Where it misleads |
|---|---|---|---|
| Sharpe ratio | Excess return per unit of total volatility | Was the climb smooth enough to hold? | Counts big winning periods as risk; inflates on short, annualised samples |
| Sortino ratio | Excess return per unit of downside volatility only | How rough was the bad side specifically? | Uses fewer observations, so it needs an even longer record |
| Profit factor | Gross wins divided by gross losses | Did the trade set make money overall? | Ignores the order of trades, so it hides the worst stretch |
| Maximum drawdown | Deepest peak-to-trough fall | What was the worst it ever felt? | A single historical event; says nothing about the next one |
| Risk of ruin | Chance of losing the account at current sizing | Can this survive a normal losing run? | Depends entirely on inputs you have to estimate honestly |
Read down that last column and the point makes itself: every one of these numbers has a blind spot, and the blind spots do not overlap. No single headline figure should carry a decision alone.
RelicusRoad Pro
Have you been trading for a while but have never made consistent profits or are you new to FOREX trading and want to get a head start? Try RelicusRoad and you'll never look back.
Get RelicusRoad ProCan a Sharpe ratio be made to look better than the strategy?
Yes, and mostly without anybody lying.
Three mechanisms do the work. Whoever publishes the figure chooses the window, so a start date after the last bad quarter is a legitimate choice with an illegitimate effect. They also choose the sampling interval, and a strategy that lurches intraday can look serene on monthly snapshots that never see the middle of the month.
The third is structural. Any approach that collects small gains repeatedly and surrenders them rarely but deeply will post a strong ratio for as long as the rare event stays outside the window. Option selling behaves this way. So does grid trading , where positions stack against the move and the drawdown stays unrealised until it isn’t. The volatility of the losses is real; it just hasn’t been sampled yet.
What makes the ratio worthless before you even compute it?
Bad inputs, and the worst of them sits at the entry price.
Every return in the calculation comes from a trade, and every trade started at a price your platform recorded. If that price shifted after the candle closed, the return series is fiction and the ratio is arithmetic performed on fiction. This is the same defect that makes repainted backtests look profitable , carried one step further: a repainted history does not just inflate the average return, it smooths the variance too, because the losing trades that were removed by hindsight were the volatile ones.
Two other input problems are worth naming. A record that spans a rule change has blended two strategies into one average. A record drawn from a single market regime describes that regime, and a proper backtesting process exists largely to stop you mistaking one for the other.
Where does the indicator on your chart come into this?
It has one job here, and accuracy is not it. The job is that the signal you logged last Tuesday is the signal still sitting there today.
RelicusRoad Pro locks its arrows at candle close and leaves them alone, on MetaTrader and on TradingView alike, so the entry prices feeding your return series are prices you could have been filled at. That does not raise your Sharpe ratio. It makes the one you calculate mean something, which is a different and more useful claim.
Everything above it still belongs to you. No chart tool sets your position size, your holding period, or whether you take the trade after two losses. What a fixed signal history buys you is the right to measure your own results and act on the measurement, and pairing that with consistent risk per trade does more for a ratio than any entry filter will.
Frequently asked questions
What is the Sharpe ratio in trading? It is a measure of how much return an account produced for the amount of volatility it put you through. Take the account’s returns over regular periods, subtract what risk-free cash would have paid over the same periods, average the result, then divide by the standard deviation of those period returns. A smooth climb scores well; the same profit delivered in lurches scores badly, because the denominator grows with every swing in either direction.
What counts as a good Sharpe ratio for a trading strategy? You will see round thresholds quoted, usually that above 1 is respectable and above 2 is excellent. Those are conventions borrowed from asset management, where the figures normally rest on years of monthly data, and no regulator sets them. On a retail record covering a few months the threshold matters far less than the sample behind it. A ratio of 3 across eleven weeks tells you less than a ratio of 0.8 across four years.
Is the Sortino ratio better than the Sharpe ratio for traders? It answers a different complaint. Standard deviation treats a large winning month as risk, which most traders find odd, so Frank Sortino and Lee Price proposed in 1994 that the denominator should count only returns below a chosen minimum. That fits a trader’s intuition better. It also uses fewer observations, so it needs a longer record before it settles. Read both when you have them.
Can a Sharpe ratio be manipulated? It can be presented selectively without anyone lying. Whoever publishes it chooses the start date, the end date and the sampling interval, and all three move the result. Strategies that collect small gains repeatedly and give them back in rare, deep losses also score well for as long as the rare loss stays outside the window.
Does the Sharpe ratio work for discretionary traders? It works mechanically on any equity curve, but it tells you more about a rule-based system than about a method that is still changing. If you altered sizing or added a setup partway through the record, the ratio has averaged two strategies into one figure that describes neither. Start the sample where the current rules started, accept the shorter record, and say so when you quote the number.
Pull your own equity curve, note the window and the sampling interval before you calculate anything, and if you want an entry history steady enough to measure, start with RelicusRoad Pro .