No single metric decides the question. A strategy can show a high win rate and lose money, or a high return that came from one lucky month. Read the report as a set: what the rules earned per trade after costs, the worst fall a trader would have sat through, how steady the path was, and whether the result holds up on days the rules were not built on.
Backtest metrics at a glance
| Group | Metric | What it answers |
|---|---|---|
| Return | Net profit, total return, CAGR | How much did the rules make? |
| Return | Expectancy | How much does an average trade make after costs? |
| Risk | Maximum drawdown and its length | What is the worst fall I would have lived through, and for how long? |
| Risk | Largest loss, longest losing streak | What will test my nerve and my margin? |
| Risk-adjusted | Sharpe, Sortino, Calmar | Was the return worth the ride? |
| Trades | Win rate, average win and loss, profit factor | How is the result made? |
| Trades | Number of trades, time in market | Is the sample large enough, and how much capital was used? |
| Robustness | Result on unseen days, by month, by weekday | Does the edge last, or did it come from one period? |
A short history of performance metrics
The idea of judging a return against its risk is younger than most traders assume. In 1952 Harry Markowitz described risk as the variability of returns, which gave investors a number to set against return.1 In 1966 William Sharpe proposed a "reward-to-variability ratio" for comparing mutual funds: the return above a risk-free rate, divided by the standard deviation of returns. It became known as the Sharpe ratio, and Sharpe revised its definition in 1994.2
Traders soon objected that Sharpe's ratio punishes large gains as much as large losses. In 1991 Frank Sortino and Robert van der Meer set out a measure of downside risk that counts only returns below a target, the basis of what is now called the Sortino ratio.3 The same year, the Calmar ratio appeared in the managed-futures world: the annual return divided by the maximum drawdown, a measure built for traders, who feel a drawdown more than they feel a standard deviation.
Trade-level measures such as expectancy and profit factor came from the systematic trading community of the 1980s and 1990s, where traders testing rules on futures needed a way to compare strategies trade by trade. More recently, research has turned to the trap inside the metrics themselves. In 2002 Andrew Lo showed how to put a confidence range around a Sharpe ratio and how serial correlation in returns distorts the usual annualised figure.4 In 2014 David Bailey and Marcos López de Prado proposed the deflated Sharpe ratio, which corrects for the number of strategy versions a researcher tried before reporting the best one.5
A worked example
The metrics below refer to this example: a Nifty options strategy backtested over three years with ₹10 lakh of starting capital.
| Metric | Value |
|---|---|
| Starting capital | ₹10,00,000 |
| Ending capital | ₹15,20,000 |
| Total return | 52% |
| CAGR | 15.0% |
| Maximum drawdown | ₹1,85,000, or 14.2% |
| Trades | 312 |
| Win rate | 58% |
| Average winning trade | ₹6,200 |
| Average losing trade | −₹4,600 |
| Profit factor | 1.86 |
| Expectancy per trade | ₹1,665 |
| Sharpe ratio | 1.06 |
| Sortino ratio | 1.65 |
| Calmar ratio | 1.06 |
| Largest losing trade | −₹38,000 |
| Longest losing streak | 7 trades |
| Time in market | 41% |
The figures are illustrative, built to be consistent with each other: 181 winning trades at ₹6,200 make ₹11.22 lakh, 131 losing trades at ₹4,600 lose ₹6.03 lakh, and the difference is the ₹5.2 lakh net profit.
Return metrics
Total return and CAGR
Total return is the overall gain as a percentage of starting capital. CAGR, the compound annual growth rate, converts it to a yearly rate so that strategies tested over different periods can be compared.
In the example, (15.2 ÷ 10)^(1 ÷ 3) − 1 = 15.0% a year. A CAGR means little on its own: 15% with a 14% drawdown is a different proposition from 15% with a 40% drawdown.
Expectancy
Expectancy is the average profit per trade after costs. It combines the win rate with the size of wins and losses.
In the example, 0.58 × 6,200 − 0.42 × 4,600 = ₹1,664 per trade, or ₹1,665 when worked from the exact trade counts. A positive expectancy after costs is the minimum requirement for any strategy. Multiply it by the number of trades you expect in a year and you have a rough yearly figure to set against the drawdown: at about 104 trades a year, ₹1,665 a trade comes to about ₹1.73 lakh. Divided by the average loss, it gives expectancy in units of risk: here about 0.36, meaning each trade earns on average 36% of a typical loss.
Risk metrics
Maximum drawdown
Maximum drawdown is the largest fall from a peak in the equity curve to the following low, before a new peak. It is the worst experience a trader following the strategy would have had.
In the example, the account fell from a peak of ₹13.0 lakh to ₹11.15 lakh at one point, a drawdown of 14.2%. Ask whether you would have kept trading through that fall. Most traders overestimate their tolerance; a live drawdown that reaches half of what they said they could bear often ends the strategy. The max drawdown calculator works it out from a series of account values.
Drawdown duration matters too: how long the account took to make a new high. A strategy that spends a year below its peak is hard to follow even if the fall was shallow.
Volatility of returns
The standard deviation of periodic returns, usually daily or monthly, measures how bumpy the path was. It is the bottom half of the Sharpe ratio.
Largest loss and losing streaks
The largest single losing trade and the longest run of losses show what the strategy demands of your nerve and your margin. With a 42% chance of losing any one trade, a run of seven or more losses somewhere in 312 trades has about a one-in-three chance of happening by luck alone, and a run of ten about a 3% chance. A streak of seven in the example is normal, not a sign of failure. A streak of twelve would be.
Risk-adjusted ratios
Sharpe ratio
The Sharpe ratio is the return above the risk-free rate divided by the volatility of returns.
In India the risk-free rate is usually taken as the 91-day Treasury bill yield, or close to the RBI's repo rate. A strategy earning 15% a year with 8% volatility, against a risk-free rate of 6.5%, has a Sharpe of (15 − 6.5) ÷ 8 = 1.06. As a rough guide, a Sharpe above 1 after costs is good for a retail strategy, and one above 2 over several years is rare enough to suggest overfitting. The Sharpe ratio calculator does the arithmetic.
Annualising. If you work from monthly returns, multiply the monthly ratio by the square root of 12; from daily returns, by the square root of about 250 trading days. A strategy that averages 1.2% a month with a monthly standard deviation of 2.6% has an annual Sharpe of about (1.2 − 0.54) ÷ 2.6 × √12 = 0.88, using a monthly risk-free rate of 0.54%. Daily figures from strategies that hold positions overnight can overstate the ratio when returns follow each other, which is the distortion Lo described.
Sortino ratio
The Sortino ratio is the same as Sharpe but counts only downside volatility, since upside swings are not a problem for most traders. It is higher than Sharpe for strategies whose large moves are mostly gains, as with the example's 1.65.
Calmar ratio
Calmar divides CAGR by maximum drawdown: 15.0 ÷ 14.2 = 1.06. It answers the question most traders ask first: how much return for each unit of the worst pain. A Calmar near or above 1 over three or more years is a reasonable bar.
Recovery factor and Ulcer index
Two more measures appear in some reports. The recovery factor divides net profit by the maximum drawdown in rupees: in the first example, ₹5,20,000 ÷ ₹1,85,000 = 2.8, meaning the strategy earned its worst fall back almost three times over. The Ulcer index, proposed by Peter Martin in 1987, measures both how deep and how long the drawdowns were, by averaging the squared percentage falls below each peak. Both reward a smoother path, which is what most traders actually want.
Trade metrics
Win rate
The share of trades that made money. It is the most quoted metric and the least useful alone. Option-selling strategies often win 70% to 85% of the time and still lose money, because the losses are much larger than the wins. Trend-following strategies may win 35% of the time and make money because the wins are large.
Profit factor
Gross profit divided by gross loss. In the example, ₹11.22 lakh ÷ ₹6.03 lakh = 1.86. A profit factor below about 1.2 leaves little room for costs and slippage that the backtest may have understated.
Average win and average loss
The reward-to-risk ratio of the strategy as traded, not as planned. A strategy designed with a target twice the size of its stop often shows nearer 1.3 to 1 in the backtest, because many trades exit early on time or get stopped at a worse price.
Number of trades
Confidence depends on the sample. Thirty trades tell you little; three hundred, as in the example, give some confidence that the edge is not luck. Fewer than 100 trades is a warning sign, especially for a strategy with many settings.
Time in market
The share of time the strategy held a position. A strategy that is in the market 41% of the time earned its return with capital free the rest of the time, which matters when comparing it with simply holding an index.
A second example: the high win rate that loses
Take an option-selling strategy over 100 trades:
| Metric | Value |
|---|---|
| Win rate | 82% |
| Average winning trade | ₹1,500 |
| Average losing trade | −₹8,200 |
| Expectancy | 0.82 × 1,500 − 0.18 × 8,200 = −₹246 per trade |
| Profit factor | (82 × 1,500) ÷ (18 × 8,200) = 0.83 |
Eight trades in ten make money, and the strategy still loses about ₹24,600 over the hundred trades. A trader who looks only at the win rate sees a winner; the expectancy and the profit factor show a loser.
Compare a trend-following rule with a 35% win rate, an average win of ₹9,000 and an average loss of ₹3,500. Its expectancy is 0.35 × 9,000 − 0.65 × 3,500 = +₹875 per trade, and its profit factor is 1.38. It loses on two trades in three, and it is the one that makes money.
How many trades are enough?
Every metric in a report is an estimate from a sample, and small samples mislead. A rough way to see this is to put a margin of error around the expectancy.
In the first example, the trade results swing widely: wins of ₹6,200 and losses of ₹4,600 give a standard deviation per trade of about ₹5,330. The margin of error on the average trade shrinks with the square root of the number of trades:
| Trades | Standard error of expectancy | Expectancy ÷ standard error |
|---|---|---|
| 30 | 5,330 ÷ √30 = about ₹973 | 1,665 ÷ 973 = 1.7 |
| 100 | 5,330 ÷ √100 = about ₹533 | 3.1 |
| 312 | 5,330 ÷ √312 = about ₹302 | 5.5 |
A ratio of about 2 or more is the usual rough bar for an average that is unlikely to be luck. With 30 trades, the same strategy's edge cannot be told apart from zero; with 312, it can. The arithmetic is a simplification, because trades are not truly independent, but it shows why a report with few trades proves little, however good its numbers look.
Comparing two strategies
Suppose you must choose between two backtested strategies over the same three years:
| Metric | Strategy A | Strategy B |
|---|---|---|
| CAGR | 22% | 14% |
| Maximum drawdown | 31% | 9% |
| Calmar | 0.71 | 1.56 |
| Sharpe | 0.9 | 1.3 |
| Trades | 140 | 420 |
| Longest time below a peak | 11 months | 3 months |
Strategy A makes more. Strategy B is the better strategy by every other measure, and it has three times the trades behind it. Because drawdown and return both scale roughly with position size, B at one and a half times its size would aim for about 21% a year with a drawdown nearer 14%, still far shallower than A's 31%. Most traders who pick A for its return abandon it during an 11-month stretch below its peak.
The habit to build is simple: compare return per unit of drawdown, not return alone, and prefer the strategy with more trades and a shorter time under water.
Metrics for option strategies and for futures
The same metric means different things for different kinds of strategy:
- Short-premium option strategies (straddles, strangles, iron condors) usually show a high win rate, small average wins and occasional large losses. For these, read the largest loss, the worst day and the profit factor first. A win rate above 80% is normal and tells you almost nothing.
- Bought-option strategies show a low win rate and large average wins. Read the longest losing streak and the drawdown length: the strategy may lose for weeks before one large win.
- Futures trend-following sits in between. Read the drawdown length and the results by year, because trend rules make their money in a few strong years.
- Intraday strategies of any kind trade often, so costs dominate. Read net against gross profit and the charges line.
A report that suits one type poorly can still be fine for another. Judge each against what its kind of strategy is expected to look like.
Reading the equity curve
Numbers summarise; the equity curve shows. Look for:
- Steadiness. A curve that rises in steps across different markets is more trustworthy than one that made all its money in a single quarter.
- Dependence on conditions. Compare trending and ranging years, and calm and volatile periods.
- Recent performance. A strategy whose last year was flat may be losing its edge, whatever the three-year total says.
- Unseen periods. If the report separates the period used to design the strategy from the period after it, the second part is the one that counts. Walk-forward testing repeats that split across many windows.
Robustness checks inside the report
A good report does not stop at the totals. It splits the result in ways that expose luck:
- By month. If two months out of thirty-six produced most of the profit, the strategy depends on those conditions returning.
- By weekday. For option strategies in India, results often differ sharply on expiry day. A rule that makes all its money on Tuesdays is an expiry-day rule, whether or not you meant it to be.
- By exit reason. How many trades closed on the stop, the target, the overall limit or the exit time? A "target" strategy that exits mostly on time is really a time strategy.
- On unseen days. The share of the period held back from design. If the held-back part is much weaker, the rules may fit only the earlier market.
Costs and slippage in the numbers
Every metric above should be read after costs. In India, the costs of an F&O trade include brokerage, securities transaction tax (raised on futures and options from 1 April 2026), exchange transaction charges, SEBI fees, stamp duty and GST.6 For strategies that trade several legs every day, costs can take a large share of the gross profit.
A useful test: run the same backtest with slippage at zero, then at the level you expect, then at twice that level. If the result turns negative at twice your expected slippage, the edge is thin, and live trading will find it out.
The metrics that decide
For most retail strategies, three numbers settle the question:
- Expectancy after realistic costs must be clearly positive, with at least 100 trades behind it.
- Maximum drawdown must be one you would actually sit through, with room to spare.
- Calmar or Sharpe must hold up on days the rules were not built on.
If all three pass, a forward test on live prices for a few weeks checks that live fills match the backtest's assumptions before any capital is at risk.
Common mistakes when reading a backtest report
- Leading with the win rate. It hides the size of losses.
- Reading totals before trades. Open the trade list and check that a few trades follow the rules as you meant them.
- Comparing returns without drawdowns. 20% with a 35% drawdown is worse for most traders than 12% with a 10% drawdown.
- Trusting a high Sharpe from few trades. Thirty trades can produce any ratio by chance.
- Ignoring how many versions were tried. The best of fifty tests is more likely to be luck than the only one.
- Forgetting costs and slippage. A thin edge before costs is usually a loss after them.
A ten-point checklist for any backtest report
Before you trust a report, from any software, go through these in order:
- Costs. Are brokerage, STT, exchange charges, stamp duty and GST included, at the rates for the period tested?
- Slippage. Is there a slippage setting, and does the result survive twice that setting?
- Trades. Are there at least 100, and does the trade list show entries and exits that follow the rules?
- Expectancy. Is the average trade clearly positive after costs?
- Drawdown. Would you have kept trading through the deepest fall and the longest time below a peak?
- Largest loss. Could your margin and your nerve have taken it?
- Spread of results. Is the profit spread across months, or did a few months make it all?
- Unseen days. Does the part of the period not used for design hold up?
- Versions tried. How many variations did you test before this one? The more, the less this one proves.
- Rule changes. Do lot sizes, expiry days and taxes in the period match how the strategy would trade today?
If any answer is "no" or "I don't know", fix that before you look at the return.
From backtest metrics to live metrics
Once a strategy runs, keep the same metrics for its live trades and compare them with the backtest every month. Three comparisons tell you most:
- Expectancy per trade. If live expectancy runs well below the backtest, look at fills and costs first; that is where most of the gap usually comes from.
- Win rate and average loss. A win rate close to the backtest with a larger average loss points to slower or worse exits than the test assumed.
- Drawdown. A live drawdown beyond the backtest's maximum is the signal you wrote down in advance to pause and review.
Two or three months of live trades are still a small sample, so look for large gaps rather than small ones, and keep Forward test running alongside live trading when you change a rule.
Metrics in QuantPower's backtest reports
QuantPower runs two kinds of backtest, and both report in rupees after costs.7
- Rule backtests test a condition on one instrument over up to 2,000 past candles from your broker, from 5-minute to daily, with a stop-loss and target in per cent, an exit rule and your slippage setting. The report shows net profit, the trade list, the equity curve, win rate, average win and loss, expectancy, reward to risk, maximum drawdown and how long it lasted, losing streaks and return against maximum drawdown.
- Option strategy backtests test a set of option legs minute by minute through each trading day. The report shows days tested, winning and losing days, win rate, net profit, gross profit and charges, average winning and losing day, best and worst day with dates, maximum drawdown with its start and end dates, the longest losing streak, results by month and by weekday, the reason each leg closed, the equity curve, and return and drawdown in per cent when you enter your capital.
- The unseen-days check. The option report holds back the last 30% of trading days and compares their average result with the first 70%, with a plain verdict such as "held up on unseen days" or "weaker on unseen days".
- Plain warnings. The report adds notes written from the numbers themselves, for example when the test covers fewer than 60 days, when one bad day took back more than half of the total profit, or when charges took more than 40% of the gross profit.
A backtest costs 2 tokens a run, and the first 10 runs each week are free.
Start with 500 free tokens
Connect your broker, build a strategy and test it on paper before a single live order goes out.
Where performance measurement is going
Two changes stand out. The first is honesty about the number of trials: with software that runs hundreds of variations in minutes, a report that shows only the best result means little unless it also shows how many were tried. Measures such as the deflated Sharpe ratio, and simple guards such as holding back unseen days, are moving from research papers into ordinary tools. The second is attention to the path, not only the total: drawdown length, losing streaks and results by month say more about whether a trader will stick with a strategy than any single annual figure.
Frequently asked questions
Which single backtest metric matters most?
There is no single one, but expectancy after costs comes closest: if the average trade does not make money after brokerage, taxes and slippage, nothing else in the report can rescue the strategy.
Can two strategies with the same CAGR carry different risk?
Yes. One may reach 15% a year with a 10% drawdown, another with a 35% drawdown and a losing streak of three months. That is why CAGR is read beside the drawdown, the Calmar ratio and the monthly results, never alone.
Can I compare Sharpe ratios from different periods?
Only with care. A Sharpe ratio from a calm year and one from a volatile year are not measuring the same conditions, and the risk-free rate used may differ. Compare strategies over the same dates, with the same rate and the same costs.
Why look at results by weekday?
Because an average can hide where the profit comes from. In India, an option strategy's result often depends on how close each day is to expiry, and expiry falls on Tuesday on the NSE and Thursday on the BSE. If most of the profit comes from one weekday, the strategy is really an expiry-day strategy, and should be judged and sized as one.
What is a reasonable maximum drawdown?
One you can live through without abandoning the strategy. Many traders set a limit of 15% to 20% of capital, and pause a strategy whose live drawdown reaches about one and a half times its backtested maximum.
Related guides
- Backtest trading strategies: the method, step by step
- How to read backtest results: the report screen in the app
- Look ahead bias: when a backtest uses information it could not have had
References
- H. Markowitz, "Portfolio Selection", *Journal of Finance* 7(1), 77–91 (1952)
- W. F. Sharpe, "Mutual Fund Performance", *Journal of Business* 39(1), 119–138 (1966); and "The Sharpe Ratio", *Journal of Portfolio Management* 21(1), 49–58 (1994)
- F. A. Sortino and R. van der Meer, "Downside Risk", *Journal of Portfolio Management* 17(4), 27–31 (1991)
- A. W. Lo, "The Statistics of Sharpe Ratios", *Financial Analysts Journal* 58(4), 36–52 (2002)
- D. H. Bailey and M. López de Prado, "The Deflated Sharpe Ratio: Correcting for Selection Bias, Backtest Overfitting and Non-Normality", *Journal of Portfolio Management* 40(5), 94–107 (2014)
- Union Budget 2026-27: securities transaction tax on futures and options raised from 1 April 2026 (indiabudget.gov.in)
- QuantPower: backtest engines and report measures, from the app's code, checked 12 November 2026