Glossary

Sharpe ratio

Q: What is the Sharpe ratio?

The Sharpe ratio is a strategy's average excess return divided by the standard deviation of those returns. Excess return means return above the risk-free rate. A Sharpe ratio of 1.0 means the strategy produced one unit of annualized excess return for every unit of annualized volatility. William F. Sharpe set out the modern form in the Journal of Portfolio Management 21(1), Fall 1994.

Q: How do you annualize a Sharpe ratio from daily returns?

Multiply the daily Sharpe ratio by the square root of 252, the conventional number of trading days in a year. The full expression is √252 × (mean daily excess return / standard deviation of daily returns). Use √52 for weekly returns and √12 for monthly returns. The scaling factor is always the square root of the number of return periods in a year.

Q: How do you annualize a Sharpe ratio for an intraday strategy?

Roll every trade up into daily profit and loss first, then compute the Sharpe ratio on that daily series, then multiply by √252. Do not compute a per-trade Sharpe ratio and scale it by √252. That is the most common error in this calculation, because the number of trades in a year is not 252, so the scaling factor does not apply to a per-trade figure.

Q: What is a good Sharpe ratio?

The practitioner convention is that above 1 is acceptable, above 2 is good and above 3 is exceptional. None of those bands has a primary academic source; they are trading-desk convention. An eFinancialCareers article quoting an anonymous trader is the origin of the widely repeated claims that most bank traders sit between 0 and 1.5 and that fewer than 5% of top-tier bank traders exceed 3. An EliteTrader thread quotes both sentences from it. One trader's estimate, not audited data.

Q: Is a backtest Sharpe ratio the same as a live Sharpe ratio?

No, and the gap is usually large. A backtest Sharpe ratio is computed on the one parameter set that survived a search, on the one historical path that happened, and often without full execution costs. A live Sharpe ratio includes slippage, fills, latency and regime change. Treat a backtest Sharpe ratio as an upper bound rather than as an estimate.

Q: Can a Sharpe ratio be negative?

Yes. A Sharpe ratio is negative whenever the strategy's average return is below the risk-free rate. The number is still arithmetically correct, but it stops behaving like a ranking. A negative Sharpe ratio should be read only as a pass or fail signal, never compared between strategies and never used as the numerator or denominator of another metric.

Q: Why does a negative Sharpe ratio rank the worse strategy higher?

Because dividing a negative number by a larger denominator moves it closer to zero. A strategy losing 5% a year at 10% volatility scores -0.50, while a strategy losing 8% a year at 20% volatility scores -0.40. The second loses more money with twice the volatility, yet ranks higher. When excess return is negative, adding volatility improves the Sharpe ratio.

Q: Why does Tradelyze show two different Sharpe ratios?

Because Sharpe (Bar) and Sharpe (Daily) sample at different rates. Sharpe (Bar) is measured on every bar of your data, annualized, then scaled down on a backtest shorter than a year. Sharpe (Daily) is measured on daily returns and annualized on the session count measured from your data, not a fixed 252. Neither subtracts a risk-free rate. The walk-forward and robustness cards show further Sharpe figures, each measured its own way.

Q: Does the Sharpe ratio here match TradingView's?

Not directly. TradingView subtracts a 2% annual risk-free rate and computes its Sharpe ratio from monthly periodic returns; Sharpe (Daily) subtracts nothing and samples per trading session, annualizing on a session count measured from your own data. A session-sampled figure matches an obsolete TradingView behavior rather than the current one. Sharpe (Bar) is further away still, because it is also scaled down on a backtest shorter than a year.

Q: Which Sharpe ratio do the robustness tests use?

Neither Sharpe tile on the Best Metrics card. Walk-forward efficiency, the deflated Sharpe ratio, minimum backtest length and parameter sensitivity use the annualized Sharpe ratio without the short-backtest cap. The Monte Carlo check and the permutation test measure the same trades on an evenly spaced grid, so read the Monte Carlo figures against MC Sharpe Original. Sharpe (Daily) is reported but used by none of the checks.

Last reviewed: 26 September 2026·Tradelyze

The Sharpe ratio is a strategy's average excess return (its return above a risk-free rate) divided by the standard deviation of its returns. It weighs profit against how much its returns swing around. A strategy beating cash by 1% a month, with returns swinging about 3% a month, scores 0.33 monthly, or about 1.15 a year. Higher means steadier profit.

In plain English

The Sharpe ratio asks how much profit a strategy made for how bumpy the ride was. Two strategies with the same profit can score very differently, and the one whose results jump around less gets the higher Sharpe ratio. A high Sharpe ratio on its own does not show that a backtest will hold up in live trading.

New to this? Start with what backtesting is.

What is the Sharpe ratio?

The Sharpe ratio divides a strategy's average excess return by the standard deviation of its returns. Excess return is the return above the risk-free rate. The risk-free rate is what cash, such as short-term government bills, would have paid with no risk taken.

Standard deviation measures how widely a set of numbers spreads around its own average. Monthly returns that all sit near 1% have a small standard deviation. Monthly returns that jump between losses and large gains have a big one.

Sharpe ratio = (mean return − risk-free rate) / standard deviation of returns

The standard deviation counts swings in both directions as risk. A large winning month raises it exactly as much as a large losing month does. William F. Sharpe set out the modern form of the measure in The Sharpe Ratio, Journal of Portfolio Management 21(1), Fall 1994, pages 49–58. That paper revised his earlier 1966 reward-to-variability ratio, and it is the version in general use.

The Sharpe ratio matters for your money because it separates a steady profit from a jagged one. Two strategies can make the same 20% in a year along very different paths. The one with the higher Sharpe ratio got there with smaller swings. Smaller swings tend to mean shallower losing stretches, and losing stretches are what break a prop firm drawdown limit.

Constructed illustration, not measured data. Both hypothetical strategies made 20% on the starting account in 12 months. The smooth one never fell more than 2 points below its high and scores a Sharpe ratio of 2.25. The bumpy one fell 7 points at worst and scores 0.95. Take a hypothetical rule that stops trading after a 5-point fall from the high. On these month-end figures, it would have stopped the bumpy strategy in month two. The smooth one never came close.

The Sharpe ratio does not measure drawdown, the fall from an account's high point, directly. Read it next to maximum drawdown.

Three properties of the Sharpe ratio cause most misreadings:

Leverage barely moves the Sharpe ratio. Leverage means trading a bigger position for the same account. Doubling position size roughly doubles both the average return and the standard deviation, so the ratio stays about the same. That helps when comparing strategies. It misleads anyone who reads a Sharpe ratio as a measure of how much money was made.
The Sharpe ratio depends on the return period. Daily returns, monthly returns and individual trades give three different Sharpe ratios for the same strategy. None converts into the others without knowing the period.
A negative Sharpe ratio breaks rankings. Once returns fall short of the risk-free rate, bigger swings produce a better-looking score. Why a negative Sharpe ratio is misleading works through the arithmetic.

How do you annualize a Sharpe ratio?

Annualizing a Sharpe ratio means converting a figure measured on short periods, such as days, into the yearly figure people quote. For daily returns, multiply the daily Sharpe ratio by the square root of 252. That is the conventional count of trading days in a year.

annualized Sharpe = √252 × (mean daily excess return / stdev of daily returns)

The multiplier is always the square root of the number of return periods in a year. That means √252 for daily returns, √52 for weekly and √12 for monthly. Returns grow in step with time, while standard deviation grows with the square root of time. So the ratio between them picks up a square-root factor.

Two constructed examples show the arithmetic. A strategy averaging 1% a month above cash, with a 3% monthly standard deviation, has a monthly Sharpe ratio of 0.333. That annualizes to 0.333 × √12 = 1.15. A daily Sharpe ratio of 0.10 annualizes to 0.10 × √252 = 1.59.

The square-root rule assumes each period's return has nothing to do with the last one. When returns tend to resemble the previous period's, the rule can overstate the Sharpe ratio. When square-root annualizing overstates a Sharpe ratio covers the research.

Tradelyze does not assume 252 trading days. Sharpe (Daily) annualizes on the number of trading sessions a year measured from your own data. That count differs between a crypto market that trades every day and a stock market that closes at weekends.

The most common annualizing mistake

An intraday strategy must roll its trades up into daily profit and loss first, and only then annualize. Compute the Sharpe ratio on the daily series, then multiply by √252.

What people do instead is compute a Sharpe ratio across individual trades and multiply that by √252. The result is meaningless. The multiplier √252 is for exactly 252 return periods a year, and a strategy's trade count has nothing to do with 252.

A strategy taking 4,000 trades a year and one taking 40 both get multiplied by the same 15.87. The figure ignores how often each strategy trades, so it understates the first and overstates the second. Two EliteTrader forum threads, listed in Sources, show traders running into exactly this.

The roll-up step is mechanical. Group every closed trade by its exit date, and sum the profit and loss within each date. Walk the calendar across every trading day between the first and the last. Include the days on which the strategy did nothing; each one counts as a return of zero.

Divide each day's profit and loss by the equity at the start of that day to get a daily return. Take the mean and standard deviation of that daily series. Divide the mean by the standard deviation, then multiply by √252.

Including flat days is not optional. A strategy that makes its whole year in six trades has a few large daily moves among roughly 246 zero-return days. Dropping the zeros inflates the average daily return and misdescribes how the strategy behaves over a year.

What is a good Sharpe ratio?

The practitioner bands are: above 1 acceptable, above 2 good, above 3 exceptional. These bands are trading-desk convention rather than a research finding. No primary academic source establishes them; they circulate because they are useful shorthand.

Sharpe ratio thresholds and where each one comes from
Value	Reading	Source
> 3	Exceptional, and rare enough that the first question should be how it was computed.	Practitioner convention. No primary source. The claim that fewer than 5% of top-tier bank traders exceed 3 comes from an anonymous trader quoted by eFinancialCareers (see Sources).
> 2	Good.	Practitioner convention. No primary source.
> 1	Acceptable.	Practitioner convention. No primary source.
0 to 1.5	Where most professional bank traders sit, on the same anonymous trader's account: “most guys in a bank are on 0-1.5”.	The same anonymous trader quoted by eFinancialCareers. One unnamed trader's estimate, not audited data.
< 0	Losing money on a risk-adjusted basis. Do not rank, compare or divide.	Direct reading. See negative Sharpe ratios.

Two caveats do more work than the bands themselves.

A backtest Sharpe ratio is not a live Sharpe ratio. The backtest figure comes from the one set of settings (parameters) that survived a search. It is measured on the one price history that actually happened, usually with optimistic assumptions about fills and costs. Each of those pushes the number up. Read a backtest Sharpe ratio of 2.4 as an upper bound. Expect the live figure to be lower, by an amount unknown until the strategy trades. See overfitting and sample size.

The bands assume an annualized Sharpe ratio computed on periodic returns, such as daily or monthly returns. A Sharpe ratio compared with these bands must say which period it used and whether it was annualized. Otherwise the comparison is meaningless. Both Sharpe tiles on the Tradelyze Best Metrics card are annualized. On a backtest shorter than a year, Sharpe (Bar) is deliberately scaled down, and neither figure deserves much weight. Which Sharpe ratio Tradelyze shows explains the difference.

Why is a negative Sharpe ratio misleading?

A Sharpe ratio is negative whenever the average return sits below the risk-free rate. The arithmetic is still correct. What stops working is the ranking. When excess return is negative, the ratio rewards higher volatility, meaning bigger swings in returns.

Dividing a negative number by a larger number moves the result closer to zero. Closer to zero reads as better on any chart or sort order that treats a higher Sharpe ratio as better. So the strategy with bigger swings scores higher.

strategy A: −5% return, 10% volatility → −0.50
strategy B: −8% return, 20% volatility → −0.40

B loses more money, with twice the volatility, and ranks higher.

Constructed example, not measured data. The rule is general, not a quirk of these two numbers: once excess return is below zero, adding volatility always improves the Sharpe ratio. So any sort or leaderboard ranking on Sharpe ratio puts the more volatile losing strategy on top — which is not always the worse one. A strategy losing 5% at 10% volatility scores −0.50, and one losing 2% at 20% volatility scores −0.10. In that pair, the higher-ranked strategy is the one that lost less. Either way the ordering reports volatility, not quality.

Three consequences follow for anyone reading a backtest report.

Do not compare two negative Sharpe ratios. The ordering between them carries no information about which strategy is better. It reports which one was more volatile.
Do not put a negative Sharpe ratio into another metric. Any ratio, average or score that consumes a Sharpe ratio inherits the inversion.
Do not average Sharpe ratios across a set that contains negatives. A mean built from a mix of signs is not interpretable, because the negative members are ordered backwards relative to the positive ones.

Where this does real damage

Dividing one negative Sharpe ratio by another cancels the signs. The result is a positive number that reads as a pass. Take some illustrative arithmetic, not a real run. The mean out-of-sample Sharpe ratio, measured on unseen test data, is −0.588. The mean in-sample Sharpe ratio, measured on the tuning data, is −0.457. Dividing the first by the second gives 1.287. That would read as a walk-forward efficiency of 128.7%, even though every tuning window and every test window lost money.

Tradelyze leaves its WF Efficiency tile blank whenever the average in-sample Sharpe ratio is zero or negative, so that false pass cannot appear there. Walk-forward efficiency covers the trap in full.

Which Sharpe ratio does Tradelyze show?

Tradelyze shows two Sharpe tiles on its Best Metrics card: Sharpe (Bar) and Sharpe (Daily). The walk-forward and robustness cards show further Sharpe figures. Tradelyze ignores the risk-free rate in all of them, so each figure divides the plain average return, not the excess return, by the standard deviation.

Leaving out the risk-free rate makes Tradelyze's figures read higher than the textbook Sharpe ratio whenever cash pays interest. The gap matters most for a strategy that barely beats cash: its textbook Sharpe ratio is near zero, while Tradelyze's figure is not.

If you only read one: use Sharpe (Daily) to compare with Sharpe ratios quoted elsewhere, and Sharpe (Bar) to compare trials inside Tradelyze. The other Sharpe figures belong to the walk-forward and robustness cards and should only be compared within their own card.

Which Sharpe label on my Tradelyze results means what?

Tradelyze shows several Sharpe ratios, measured in different ways. Compare each one only with the figures in its “Compare it with” cell. Setting two different measurements side by side can make a sound result look broken, or hide a real problem.

Every Sharpe ratio label in Tradelyze results, what it measures and what to compare it with
On-screen label	Where it appears	What it measures	Compare it with
Sharpe (Bar)	Best Metrics card	Profit earned per unit of the ups and downs the strategy put the account through, measured bar by bar. It is annualized on a rate worked out from your data's own bar spacing. It is then capped at the period actually observed, so a two-week backtest cannot claim a full year's worth of confidence.	Sharpe (Bar) from other runs, and the Sharpe column in Top Trials, which covers only the history the search saw.
Sharpe (Daily)	Best Metrics card	The same risk-adjusted return measured day by day over the trading sessions your data contains. It is annualized on the session count measured from that data, not a fixed 252. With fewer than 21 sessions of data it reads 0.	Published annualized Sharpe ratios, approximately. Not TradingView's Sharpe ratio, which uses monthly returns and a 2% risk-free rate.
Sharpe	Top Trials table	Each trial's own Sharpe (Bar): the same bar-by-bar, capped measurement, for that trial's settings, over the part of the history the search saw.	Other rows of Top Trials. When the held-out test runs, Best Metrics cover the full history, so the same settings can show a different Sharpe (Bar) there.
Mean IS Sharpe / Mean OOS Sharpe, or Mean Earlier Sharpe / Mean Later Sharpe	Walk-forward card	Averages, across the usable walk-forward windows, of each window's annualized Sharpe ratio with no short-backtest cap. On a run labeled Re-tuned each window, Mean IS Sharpe covers each window's tuning stretch (in-sample) and Mean OOS Sharpe its unseen test stretch (out-of-sample). WF Efficiency is the second divided by the first. On a run labeled One run, split by period, the default for eligible strategies, the tiles read Mean Earlier Sharpe and Mean Later Sharpe, and their ratio is the Retention Ratio. Those settings were chosen on the same history, so neither stretch was hidden from the search. A test stretch holding only a few trades can produce a large Sharpe ratio that says nothing about the strategy.	Each other, and the Sharpe (ann.) column in Per-Window Results. Not Sharpe (Bar).
Sharpe Ratio and Sharpe (ann.)	Per-Window Results table, walk-forward card	Figures for each window's tuning half and test half. Sharpe Ratio is the capped figure, built like Sharpe (Bar). Sharpe (ann.) is the annualized figure without the cap, the one WF Efficiency is calculated from.	Sharpe (ann.) with Mean IS Sharpe and Mean OOS Sharpe; the tuning half with the test half of the same window.
MC Sharpe Mean / MC Sharpe 5th–95th / MC Sharpe Original	Robustness card, Monte Carlo check	Tradelyze redraws your trade list at random, 1,000 times by default, so some trades appear twice and some not at all. It recalculates the Sharpe ratio for each draw. MC Sharpe Mean is the average across the draws. MC Sharpe 5th–95th is the band between two percentiles. Only 1 draw in 20 fell short of the 5th percentile, and only 1 in 20 reached the 95th. MC Sharpe Original is your real trade list measured the same way.	Each other only. Not Sharpe (Bar) or Sharpe (Daily). The Monte Carlo figures use an evenly spaced grid of your trades. The Best Metrics card uses the bars your trades really closed on.

Sharpe (Bar): measured on every bar, scaled down on short backtests

Sharpe (Bar) is the Sharpe ratio measured on every bar of your chart data and converted to a yearly figure. On a backtest shorter than a year it is then scaled down, so a short test cannot look as convincing as a long one. Tradelyze's tooltip calls a Sharpe (Bar) under 0.5 weak and one over 1.0 good.

Sharpe (Bar) is not a per-trade figure, whatever an older Tradelyze report may have called it. Tradelyze turns the backtest's profit and loss into a return on each bar and divides the average by the standard deviation. It annualizes on a rate worked out from your own bar spacing, so nothing is assumed about your market's opening hours. No risk-free rate is subtracted.

Sharpe (Bar) = annualized bar Sharpe × √min(years observed, 1)
risk-free rate = 0

On a backtest covering a year or more, the cap changes nothing. On a shorter one, Sharpe (Bar) is smaller than the straight annualized figure, which stretches a short sample out to a full year. The full tooltip bands are under 0.5 weak, 0.5 to 1.0 acceptable, 1.0 to 2.0 good and over 2.0 excellent. Tradelyze chose those bands, and they have no primary source. Tradelyze's optimizer does not rank trials on Sharpe (Bar) alone. It searches for higher profit, higher Sharpe (Bar) and lower drawdown together. Top Trials are then ranked, and each prop firm's recommended settings picked, on one score that combines the three and gives less credit to a result built on few trades.

This tile used to be labeled Sharpe (Trade)

The figure was once a per-trade ratio with no annualization, which let a handful of near-identical trades report an enormous Sharpe ratio. Tradelyze replaced it with the bar-by-bar construction and renamed the label Sharpe (Bar). Reports from before the change are not comparable, and neither are their old bands of 0.1, 0.3 and 0.5.

Sharpe (Daily): measured on daily returns and annualized

Sharpe (Daily) is the Sharpe ratio measured on your strategy's daily returns and then annualized, meaning converted to a yearly figure. Tradelyze's tooltip calls a Sharpe (Daily) under 0 losing, 0 to 1 below average, 1 to 2 good and over 2 excellent. Tradelyze chose those bands, and they have no primary source.

Tradelyze folds each bar's profit and loss into the trading sessions your data contains, and a session with no exits counts as a zero return. Tradelyze annualizes on the sessions a year measured from that data, not a fixed 252. That is why crypto and stock strategies are annualized differently. On Tradelyze's three reference datasets, the measured counts were 388.6, 319.4 and 252.3 sessions a year. No risk-free rate is subtracted, and with fewer than 21 sessions of data Sharpe (Daily) reads 0.

Sharpe (Daily) = √(sessions per year, measured) × mean(daily return) / stdev(daily return)

Sharpe (Bar) and Sharpe (Daily) legitimately differ. Folding bars into sessions smooths away movement within the day, and only Sharpe (Bar) is scaled down on short backtests. Use Sharpe (Daily) against published annualized Sharpe ratios or the good-Sharpe bands. Use Sharpe (Bar) to compare trials and runs inside Tradelyze.

Neither Sharpe tile matches TradingView's Sharpe ratio

TradingView computes its Sharpe ratio from monthly returns and subtracts a risk-free rate; Tradelyze does neither. TradingView's support article defines the return term as “Average return for a monthly trading period”, with a 2% annual risk-free rate. The 8 November 2024 notes for its RiskMetrics Pine library say it “now uses monthly periodic returns exclusively”.

No square-root-of-time factor converts one figure into the other. That scaling step is unreliable whenever returns resemble the previous period's, as when square-root annualizing overstates a Sharpe ratio explains.

Which Sharpe ratio the checks use

No walk-forward or robustness check reads either Sharpe tile on the Best Metrics card. Walk-forward efficiency, the deflated Sharpe ratio, minimum backtest length and parameter sensitivity use the annualized Sharpe ratio without the cap. The Monte Carlo check and the permutation test measure trades on an evenly spaced grid. So read the Monte Carlo band against MC Sharpe Original, as the Monte Carlo page explains. Sharpe (Daily) feeds none of the checks.

How does the Sharpe ratio compare to Sortino and Calmar?

The Sharpe ratio uses the standard deviation of all returns, so it treats upside and downside swings identically. A strategy that occasionally makes 20% in a month is penalized for it. The penalty is exactly the same as for a strategy that occasionally loses 20%.

The Sortino ratio swaps total deviation for downside deviation. Downside deviation is computed only from returns that fell short of a target, usually zero or the risk-free rate. Strategies with lumpy upside stop being punished for winning.

Trend-following is the standard example. Long stretches of small losses are broken up by a few very large gains. Those gains produce most of the standard deviation, which drags the Sharpe ratio down while describing nothing a trader would call risk.

Frank A. Sortino and Lee N. Price set out the downside-risk framework in Performance Measurement in a Downside Risk Framework, Journal of Investing 3(3), 1994, pages 59–64.

The Calmar ratio replaces volatility entirely. It divides annualized return by maximum drawdown, the largest fall from a high point, over a trailing window. That window is commonly three years by practitioner convention; no primary source is cited on this page for it. The Calmar ratio sets return against the worst observed loss rather than the spread of returns, which makes it sensitive to one extreme observation. See maximum drawdown for why that single observation is the least stable number on a backtest report.

Not available in Tradelyze

Tradelyze does not compute the Sortino ratio or the Calmar ratio. The Best Metrics card shows Profit, Sharpe (Bar), Sharpe (Daily), Max Drawdown, Win Rate, Profit Factor, Trade Count, First Trade and TV Comparable From. Sortino, Calmar and expectancy are not among them. For a downside-deviation figure, download a trial's trades with Download Trades in the Top Trials table and calculate it yourself.

What is the deflated Sharpe ratio?

The deflated Sharpe ratio lowers an observed Sharpe ratio to allow for how many strategy variants were tested before that one was picked. It turns the result into a probability that the edge is genuine, rather than the luckiest result of a large search.

Tradelyze reports the deflated Sharpe ratio as DSR Value on the robustness card and treats a value over 0.95 as Significant. A value of 0.95 or less fails the check, and so does a check refused for fewer than 5 trades; one that could not be computed does not. Since 26 September 2026, any failed check multiplies the robustness score by 0.69, so a near miss is no longer hidden inside a high total. As computed with Tradelyze's scoring code, a DSR Value of 0.94, with the other three checks perfect and 400 days of data, scores 68.7, which the card shows rounded down as 68, grade C+, verdict MARGINAL. The formula, the calculation and what a blank DSR Value means are on the robustness score page.

The practical rule: keep the number of trials that produced the winning settings next to the Sharpe ratio. Tradelyze shows that count as Trials run on the results page. In a constructed comparison, a Sharpe ratio of 2.6 from a 50,000-trial search can be weaker evidence than 1.4 from a 20-trial search. Why a large search inflates the best Sharpe ratio explains the mechanism, and overfitting and sample size covers the wider problem.

Going deeper

The sections below go deeper: when square-root annualizing overstates a Sharpe ratio, and why a large search inflates the best Sharpe ratio. You can skip them and still read your own report.

When does square-root annualizing overstate a Sharpe ratio?

Annualizing by the square root of time, such as multiplying a monthly Sharpe ratio by √12, assumes each period's return is independent of the last. Serial correlation, also called autocorrelation, breaks that assumption: one period's return tends to resemble the next.

Where the √T rule stops being safe

Square-root-of-time scaling assumes returns are independent from one period to the next. Andrew W. Lo showed in The Statistics of Sharpe Ratios (Financial Analysts Journal 58(4), July/August 2002, pages 36–52) that positive serial correlation in returns, where one period's return tends to resemble the next, makes the naive √T annualization overstate the Sharpe ratio, and that the size of the error depends on the autocorrelation structure.

The overstatement is not a rounding effect: in Lo's worked example, the naive annualization overstated one hedge fund's Sharpe ratio by as much as 65%. Strategies that hold positions across many bars, and anything with smoothed or infrequently marked pricing, are the cases where this bites.

The same assumption is why a Sharpe ratio measured on daily sessions cannot be rescaled into a Sharpe ratio measured on monthly returns, such as TradingView's, with a square-root factor.

Why does a large search inflate the best Sharpe ratio?

The problem the deflated Sharpe ratio solves is selection bias. Run enough setting combinations against random data and some will show a high Sharpe ratio by luck alone, and the more combinations tried, the higher the best lucky score climbs. David H. Bailey and Marcos López de Prado introduced the correction in The Deflated Sharpe Ratio: Correcting for Selection Bias, Backtest Overfitting and Non-Normality, Journal of Portfolio Management 40(5), 2014, available as SSRN 2460551.

Bailey and López de Prado's correction also allows for the length of the track record, and for returns that are lopsided or prone to extreme values. Statisticians call those two shapes skewness and kurtosis; they are the “non-normality” in the paper's title.

The same argument has been made from the significance-testing side. Campbell R. Harvey, Yan Liu and Heqing Zhu argue for a t-statistic hurdle of about 3.0 rather than the conventional 2.0, in … and the Cross-Section of Expected Returns, Review of Financial Studies 29(1), 2016, pages 5–68.

Harvey, Liu and Zhu's reasoning is that so many factors and strategies have already been tested across the literature and across industry that a threshold calibrated for a single independent test no longer controls the false discovery rate. A backtest clearing 2.0 after a large parameter search is weak evidence.

Stage 2 · step 8 of 18. Next in the learning path: In-sample vs out-of-sample

Check it on your own strategy

In a Tradelyze report, the Sharpe ratio appears as the Sharpe (Bar) and Sharpe (Daily) tiles in Best Metrics. The walk-forward and robustness cards show further Sharpe figures, each measured its own way. To judge the whole report, not one tile, use the pre-trade checklist. Tradelyze re-runs an uploaded TradingView Pine Script strategy on your price data and checks it against your exported trade list. It then runs parameter optimization, walk-forward analysis, a four-check robustness score and prop firm rule checks. It does not place trades, give financial advice or guarantee a challenge pass, and it is in beta.

Create an account. Already a user? Open your strategies.

Frequently asked questions about the Sharpe ratio

What is the Sharpe ratio?

The Sharpe ratio is a strategy's average excess return divided by the standard deviation of those returns. Excess return means return above the risk-free rate. A Sharpe ratio of 1.0 means the strategy produced one unit of annualized excess return for every unit of annualized volatility. William F. Sharpe set out the modern form in the Journal of Portfolio Management 21(1), Fall 1994.

How do you annualize a Sharpe ratio from daily returns?

Multiply the daily Sharpe ratio by the square root of 252, the conventional number of trading days in a year. The full expression is √252 × (mean daily excess return / standard deviation of daily returns). Use √52 for weekly returns and √12 for monthly returns. The scaling factor is always the square root of the number of return periods in a year.

How do you annualize a Sharpe ratio for an intraday strategy?

Roll every trade up into daily profit and loss first, then compute the Sharpe ratio on that daily series, then multiply by √252. Do not compute a per-trade Sharpe ratio and scale it by √252. That is the most common error in this calculation, because the number of trades in a year is not 252, so the scaling factor does not apply to a per-trade figure.

What is a good Sharpe ratio?

The practitioner convention is that above 1 is acceptable, above 2 is good and above 3 is exceptional. None of those bands has a primary academic source; they are trading-desk convention. An eFinancialCareers article quoting an anonymous trader is the origin of the widely repeated claims that most bank traders sit between 0 and 1.5 and that fewer than 5% of top-tier bank traders exceed 3. An EliteTrader thread quotes both sentences from it. One trader's estimate, not audited data.

Is a backtest Sharpe ratio the same as a live Sharpe ratio?

No, and the gap is usually large. A backtest Sharpe ratio is computed on the one parameter set that survived a search, on the one historical path that happened, and often without full execution costs. A live Sharpe ratio includes slippage, fills, latency and regime change. Treat a backtest Sharpe ratio as an upper bound rather than as an estimate.

Can a Sharpe ratio be negative?

Yes. A Sharpe ratio is negative whenever the strategy's average return is below the risk-free rate. The number is still arithmetically correct, but it stops behaving like a ranking. A negative Sharpe ratio should be read only as a pass or fail signal, never compared between strategies and never used as the numerator or denominator of another metric.

Why does a negative Sharpe ratio rank the worse strategy higher?

Because dividing a negative number by a larger denominator moves it closer to zero. A strategy losing 5% a year at 10% volatility scores -0.50, while a strategy losing 8% a year at 20% volatility scores -0.40. The second loses more money with twice the volatility, yet ranks higher. When excess return is negative, adding volatility improves the Sharpe ratio.

Why does Tradelyze show two different Sharpe ratios?

Because Sharpe (Bar) and Sharpe (Daily) sample at different rates. Sharpe (Bar) is measured on every bar of your data, annualized, then scaled down on a backtest shorter than a year. Sharpe (Daily) is measured on daily returns and annualized on the session count measured from your data, not a fixed 252. Neither subtracts a risk-free rate. The walk-forward and robustness cards show further Sharpe figures, each measured its own way.

Does the Sharpe ratio here match TradingView's?

Not directly. TradingView subtracts a 2% annual risk-free rate and computes its Sharpe ratio from monthly periodic returns; Sharpe (Daily) subtracts nothing and samples per trading session, annualizing on a session count measured from your own data. A session-sampled figure matches an obsolete TradingView behavior rather than the current one. Sharpe (Bar) is further away still, because it is also scaled down on a backtest shorter than a year.

Which Sharpe ratio do the robustness tests use?

Neither Sharpe tile on the Best Metrics card. Walk-forward efficiency, the deflated Sharpe ratio, minimum backtest length and parameter sensitivity use the annualized Sharpe ratio without the short-backtest cap. The Monte Carlo check and the permutation test measure the same trades on an evenly spaced grid, so read the Monte Carlo figures against MC Sharpe Original. Sharpe (Daily) is reported but used by none of the checks.

What is the difference between the Sharpe ratio and the Sortino ratio?

The Sortino ratio replaces the standard deviation of all returns with the deviation of returns that fell below a target, usually zero or the risk-free rate. The Sharpe ratio penalizes a large winning month exactly as much as a large losing month, because both raise the standard deviation. Trend-following strategies with lumpy upside are the clearest case where the two rank strategies differently.

Does Tradelyze compute the Sortino or Calmar ratio?

No. The Tradelyze Best Metrics card shows Profit, Sharpe (Bar), Sharpe (Daily), Max Drawdown, Win Rate, Profit Factor, Trade Count, First Trade and TV Comparable From. Sortino, Calmar and expectancy are not calculated. A downside-deviation figure such as the Sortino ratio has to be worked out by hand from a trial's trade list, which the Top Trials table offers through Download Trades.

What is the deflated Sharpe ratio?

The deflated Sharpe ratio adjusts an observed Sharpe ratio downward to account for how many strategy variants were tested before that one was selected. Bailey and López de Prado introduced it in the Journal of Portfolio Management 40(5), 2014, available as SSRN 2460551. Search hard enough over random data and some variant will show a high Sharpe ratio by chance alone. Tradelyze reports it as DSR Value on the robustness card.

What t-statistic should a backtest Sharpe ratio clear?

Harvey, Liu and Zhu argue for a hurdle of about 3.0 rather than the conventional 2.0, in the Review of Financial Studies 29(1), 2016. Their reasoning is that so many factors and strategies have already been tested that the conventional threshold no longer controls the false discovery rate. A backtest clearing 2.0 after a large parameter search is weak evidence.

Sources

William F. Sharpe, The Sharpe Ratio, Journal of Portfolio Management 21(1), Fall 1994, pages 49–58 — the modern definition of the measure.
Andrew W. Lo, The Statistics of Sharpe Ratios, Financial Analysts Journal 58(4), July/August 2002, pages 36–52 — why square-root-of-time annualization overstates the Sharpe ratio under serial correlation.
David H. Bailey and Marcos López de Prado, The Deflated Sharpe Ratio: Correcting for Selection Bias, Backtest Overfitting and Non-Normality, Journal of Portfolio Management 40(5), 2014, SSRN 2460551.
Campbell R. Harvey, Yan Liu and Heqing Zhu, … and the Cross-Section of Expected Returns, Review of Financial Studies 29(1), 2016, pages 5–68 — the argument for a t-statistic hurdle of about 3.0 rather than 2.0.
Frank A. Sortino and Lee N. Price, Performance Measurement in a Downside Risk Framework, Journal of Investing 3(3), 1994, pages 59–64 — the downside-deviation denominator.
TradingView, Risk/performance ratios: Sharpe ratio, support solution 43000681694 — the definition of MR as “Average return for a monthly trading period” and the 2% annual risk-free default. Retrieved 28 July 2026.
TradingView, RiskMetrics Pine library, script oOgZRqiM, version 2 release notes dated 8 November 2024 — “now uses monthly periodic returns exclusively”, replacing a scheme that used “returns from either daily or monthly periods”. Retrieved 28 July 2026.
eFinancialCareers article 312321, quoting an anonymous trader — the origin of the “most guys in a bank are on 0-1.5” range and the under-5%-above-3 figure. The article is about BlueCrest Capital Management. Deliberately not hyperlinked: the original article URL now redirects to that site's news index, checked 28 July 2026. Both sentences survive as quoted excerpts inside EliteTrader thread 325214, Fewer than 5% of top-tier bank traders have Sharpe ratios > 3, opened by the user pinetboltz, who links the article and quotes from it — a partial quotation with commentary rather than a full repost, re-checked 16 September 2026. “Most guys in a bank are on 0-1.5” is the trader's own wording; the under-5% figure is the article's summary of what he said. One unnamed trader's estimate, not audited data.
EliteTrader threads 261598 (calculate sharpe ratio for intraday trading) and 185617 (How to calculate Annualized Sharpe Ratio from Daily Returns) — the per-trade-times-√252 error. Forum discussion, not a study. Retrieved 28 July 2026.
Figures 1 and 2 on this page are constructed illustrations, not measured data. Figure 2's Sharpe ratios are computed from the twelve monthly results listed in its description, annualized with √12, with the risk-free rate set to zero.
Tradelyze implementation, reviewed 14 September 2026 — Sharpe (Bar) is the per-bar annualized Sharpe ratio multiplied by the square root of the years observed, capped at one; Sharpe (Daily) is annualized on the session count measured from the data and reads 0 with fewer than 21 sessions; no risk-free rate is subtracted from any Sharpe figure; the walk-forward averages, the deflated Sharpe ratio, minimum backtest length and parameter sensitivity use the uncapped annualized figure; the Monte Carlo band and the permutation test re-measure trades on an evenly spaced grid.
Tradelyze implementation, reviewed 26 September 2026 — the walk-forward method is chosen by the deployment, not the user. Runs labeled Re-tuned each window show Mean IS Sharpe, Mean OOS Sharpe and WF Efficiency. Runs labeled One run, split by period, the default for eligible strategies, show Mean Earlier Sharpe, Mean Later Sharpe and Retention Ratio, and their settings were chosen on the same history. Parameter stability, a third method, was retired on 26 September 2026. The search optimizes profit, Sharpe (Bar) and drawdown together; Top Trials are ranked, and each prop firm's recommended settings picked, on one score combining the three, with less credit for few trades. Top Trials are scored on the history the search saw, and Best Metrics on the full history when the held-out test runs. The deflated Sharpe check passes at a DSR Value over 0.95; it counts as failed when refused for fewer than 5 trades, but not when it could not be computed. Any failed scored check multiplies the robustness points by 0.69, so the score is at most 69, C+ and MARGINAL at best. The DSR 0.94 example on this page was computed with Tradelyze's scoring code and is given as the card shows it, rounded down.