| Sharpe ratio | How it is usually described | What that reading assumes |
|---|---|---|
| Below 0 | Underperformed cash | Nothing. This one is safe to read directly. |
| 0 to 1 | Weak to acceptable | That the period is long enough for the estimate to mean anything. |
| 1 to 2 | Good | That the strategy was not chosen from many candidates on this same data. |
| 2 to 3 | Strong | The same, plus that costs, slippage and capacity are already in the number. |
| Above 3 | Excellent, and rare in public track records | In a backtest, most often that something is wrong: a search, a leak, or unmodeled costs. |
Where do the conventional bands come from?
From live, long-horizon track records of diversified portfolios, mostly institutional, measured over periods long enough that the estimate settles down. In that setting the number is a measurement: somebody ran a strategy, the returns happened, and the ratio describes them afterwards.
That is a different object from the Sharpe ratio on a backtest. A backtested Sharpe is the output of a selection process. You tried some number of rules, parameter values, instruments and periods, and you are looking at the one that came out highest. The bands were never calibrated for that, and nothing about them warns you.
Why is a backtested Sharpe ratio not comparable to a live one?
Because searching raises it for free. Test enough variants on the same data and some will show a high Sharpe ratio through chance alone, in the same way that some run of coin flips in a large enough sample looks like a hot hand. The winner of that search is what gets written down.
So the same figure carries wildly different evidence depending on a fact the figure does not contain:
- A Sharpe of 1.5 from the first idea somebody tested is a real, if modest, result worth investigating further.
- A Sharpe of 1.5 from the best of fifty thousand variants is close to what you would expect from noise, and is not evidence of anything.
Both are correctly calculated. Neither can be interpreted without the trial count, and the trial count is almost never reported — usually because it was never recorded.
What is a good Sharpe ratio for a backtested strategy?
The honest answer is that the question is not answerable in that form, and any source that gives you a threshold has quietly assumed the search away.
A more useful version: how high would this Sharpe ratio need to be before it is surprising, given how it was found? That question does have an answer, and it is what the deflated Sharpe ratio computes. It takes the observed Sharpe, the length of the sample, the skew and fat-tailedness of the returns and the number of trials, and returns a probability that the result exceeds a benchmark rather than being the best draw from a pile of noise.
Run that way, the picture usually changes. It is ordinary for a strategy showing an in-sample Sharpe above 2 to carry a deflated Sharpe near zero once a large search is accounted for. Both numbers are correct. The first describes the chart, the second describes what the chart is worth.
Backtested and hypothetical results are not indicative of future performance. Nothing here is investment advice.
Why is a very high Sharpe ratio a warning sign?
Because in a backtest the cheapest way to produce one is a mistake. Before treating a figure above 3 as a discovery, the ordinary explanations should be ruled out in this order, because they are ranked by how often they turn out to be the cause.
- Costs were left out or set optimistically. Spread, commission, slippage and financing come off the top, and a strategy trading often can be excellent before them and unprofitable after.
- The data leaked. A rule acting on a closing price it could not yet know, or on a fundamental figure published weeks later, produces performance no strategy could achieve.
- The search was large and uncounted. This is the most common cause and the least visible one, because nothing in the output records it.
- The sample is short. Sharpe ratios estimated over a few months are extremely noisy, and annualizing a short window makes a small run of luck look like a rate.
Only after those four are excluded does an unusually high figure become interesting, and at that point the interesting question is capacity: whether the edge survives at any size worth trading.
What does the Sharpe ratio miss?
It divides excess return by the standard deviation of returns, which builds in assumptions that do not hold for many strategies.
- It treats upside and downside variation as equally bad, so a strategy with occasional large gains is penalized like one with occasional large losses.
- It assumes returns are well described by their mean and standard deviation. Strategies that collect small premiums and occasionally lose a great deal — selling options, carry trades, most short-volatility exposure — report flattering Sharpe ratios right up until they do not.
- It says nothing about the shape of a drawdown, only its contribution to variance. Two strategies with the same Sharpe ratio can be very different to actually hold.
- It scales with the observation frequency it was computed at, so daily and monthly figures are not directly comparable unless both are annualized the same way.
None of this makes the Sharpe ratio useless. It makes it one summary statistic, best read alongside the drawdown profile, the trade count and how the result was found.
Common questions
Is a Sharpe ratio of 1 good?
For a live, multi-year track record it is generally treated as good. For a backtest it depends entirely on how many variants were tried to reach it. From a single tested idea it is a reasonable result; from a broad automated search it is roughly what noise produces.
What is a good Sharpe ratio for a day trading strategy?
Short-horizon strategies often report high figures because many small trades reduce measured volatility, and they are also the strategies most exposed to costs. Judge the number after realistic spread and slippage rather than before, and treat a high pre-cost Sharpe on a high-turnover rule as untested rather than strong.
Is a higher Sharpe ratio always better?
No. Above a certain point, in a backtest, a higher figure is better evidence that something is wrong than that something was found. It also ignores tail risk, so strategies that lose rarely and heavily can score well until the rare event arrives.
How many years of data do you need for a reliable Sharpe ratio?
There is no fixed number worth quoting, and trade count matters more than calendar time. What is safe to say is that estimates over short windows are very noisy, and that annualizing a few months of returns turns a short run of luck into something that reads like a rate.
What is the difference between the Sharpe ratio and the deflated Sharpe ratio?
The Sharpe ratio describes a track record. The deflated Sharpe ratio is a probability between 0 and 1 that the result is not simply the best outcome of the search that produced it, after accounting for trial count, sample length, skew and fat tails. They answer different questions and are not on the same scale.