Friday 2 Jan 2026. Nifty closed at 26,328.55. Monday 30 Mar 2026: 22,331.40, 15.2% lower. On 15 Sep 2026, the last close in the files, it stood at 23,118.60, 12.2% under the 2 Jan level, more than eight months on. From 1 Jul 2025 to 1 Jul 2026 it fell 6.0%, from 25,541.80 to 24,005.85. Three numbers describe one stretch: a return, a worst fall and a time under water. None of them tells the story alone, and a strategy record is read the same way. This lesson defines the five measures that matter, CAGR, Sharpe, Sortino, drawdown and expectancy, computes them on dated NSE closes and on two made up records, and says what each one hides. The closes are price levels without dividends. Nothing here is a strategy, a return claim or a view on where Nifty goes.
3By the end you can
- Compute CAGR, drawdown and the gain needed to recover from dated NSE closes.
- Compute Sharpe, Sortino, Calmar and expectancy from a record and say what each one ignores.
- Explain why one record is one draw, and why a high Sharpe ratio is not a safety rating.
The idea, plainly
What is CAGR?
CAGR is the constant yearly rate that takes a start value to an end value. Over exactly one year it is just the return, so the Nifty's fall of 6.0% from 1 Jul 2025 to 1 Jul 2026 is its CAGR for that year, with a minus sign. Over more years the rate compounds. A yearly 10% for two years gives 21%, not twice the rate. CAGR hides the road: two records that end at the same value after three years share a CAGR, however differently they got there. It also depends on the index it is read from. A price index leaves out dividends, and a total return index adds them, so a CAGR from price closes understates what the holder of the shares earned. A number should always say which one it used.
CAGR =
(end ÷ start)
^ (1 ÷ years) − 1
What are volatility and the Sharpe ratio?
Volatility is the standard deviation of returns, scaled to a year: the daily figure times the square root of the trading days in a year, about 252. The Sharpe ratio divides the return above a risk free rate by that volatility. It asks how much return came for each unit of bumpiness. The denominator is not a constant of the strategy. The 20 day realised volatility of Nifty ran from 5.2% to 29.5% in the files, so one strategy measured in a quiet window and in a stormy one gives two Sharpe ratios. Volatility also counts a jump up as risk as much as a jump down, and says nothing of a rare large loss.
Sharpe =
(return − risk free rate)
÷ volatility
The Sharpe ratio's denominator moved between 5.2% and 29.5%
What is the Sortino ratio?
Sortino keeps the top of the Sharpe ratio and changes the bottom. It divides by the downside deviation, which counts only returns below a chosen target, usually zero. A record whose bumps are mostly upward scores better on Sortino than on Sharpe. It still rests on past returns, and it still misses a loss that has not happened yet.
What are maximum drawdown and Calmar?
A drawdown is the fall from a peak to the next low. The maximum drawdown is the worst of them in the record. In the files it is the fall from 26,328.55 on 2 Jan 2026 to 22,331.40 on 30 Mar, 15.2%. Climbing back needs a bigger gain than the fall, 17.9 percent here, because the base is smaller. A fall of a quarter needs 33.3 percent and a fall of a half needs 100.0 percent. Calmar divides the CAGR by the maximum drawdown. Drawdown is the number a holder lives through and the one that ends strategies, since it says how deep and, with the dates, how long. It is also the worst fall in the sample, not the worst possible one, and a longer sample usually holds a deeper fall.
Drawdown =
(trough − peak) ÷ peak
Gain to recover =
1 ÷ (1 − fall) − 1
A fall of 15.2% from the peak, and still 12.2% under it at the last close
The deeper the fall, the steeper the climb back
What is expectancy, and why is win rate not enough?
Expectancy is what an average trade earns, in units of the amount risked, written R. It is the win rate times the average win, less the loss rate times the average loss. A record that wins 80% of its trades, earns 0.25 R on a win and loses one R on a loss has an expectancy of 0.00 R a trade, however often it wins. A record that wins 40% and earns 2 R on a win has 0.20 R a trade. The high win rate is the worse record. Profit factor, the gross profit over the gross loss on the same trades, tells it as 1.00 against 1.33. Buyer or seller shows the same split on option positions.
Expectancy =
win rate × average win
− loss rate × average loss
How do Sharpe, drawdown, Kelly, sizing and risk of ruin fit together?
Each answers a different question about a record, and each has a place where it misleads.
| Question | Measure | Where it misleads |
|---|---|---|
| How fast did it grow? | CAGR | Hides the road |
| How bumpy per unit of return? | Sharpe, Sortino | Depends on the window |
| How bad did it get, and how long? | Drawdown, Calmar | One episode |
| What does a trade earn? | Expectancy | Needs many trades |
| How much to risk? | Kelly, risk of ruin | Needs the odds known |
The first four describe a past record. The last turns a believed edge into a size, and it is only as good as the odds put into it. Position sizing and risk of ruin works that through, and the Kelly rule is in Kelly and sizing for option strategies.
2 terms this section turns on
Worked example: two made up records and one model
Records A and B are made up. Both have a CAGR of 15% over three years, so Rs 10,00,000 becomes Rs 15,20,875 in each. The risk free rate is assumed at 5%. They differ in how they got there.
| Made up record | A | B |
|---|---|---|
| CAGR | 15% | 15% |
| Volatility | 10% | 20% |
| Downside deviation | 6% | 14% |
| Maximum drawdown | 8% | 30% |
| Sharpe | 1.00 | 0.50 |
| Sortino | 1.67 | 0.71 |
| Calmar | 1.88 | 0.50 |
Model, not market data
One edge, many paths: the weakest twentieth ended near 0.93 times the start
Read 2 more paragraphs, 138 words
The same CAGR hides a record that was twice as bumpy and fell nearly four times as far. Sharpe is 1.00 against 0.50, and Calmar is 1.88 against 0.50. A holder of B had to sit through a fall of 30%, which is the number that decides whether anyone stays.
One record is also one draw. Take a model with a win rate of 45%, an average win of 1.5 times the average loss and 250 trades, risking 2% of the account on each. The expectancy is 0.125 R a trade and Kelly is 8.3%. Over 2,000 simulated paths the median account ends at 1.77 times its start. The weakest twentieth end at 0.93 or less, and the strongest twentieth at 3.22 or more. Every one of those paths has the same edge.
The common mistake
The belief: A high Sharpe ratio means a safe strategy.
Sharpe measures bumpiness in the window it was computed on. A seller of far options earns small amounts in nearly every period and meets the tail in a few, as the lesson on backtest biases shows with a fair priced model. Computed on a stretch without the tail, the volatility is tiny and the ratio looks superb. The ratio ranks records. The drawdown says what the holder lived through. Before trusting any metric, ask over what window, on what data, from how many trades and after how many versions were tried. SEBI lets only registered advisers, research analysts and empanelled algo providers state past performance, and only as risk and return figures verified by a recognised agency. A screenshot of a ratio is not that.
Key takeaways
- CAGR is the constant yearly rate that compounds to the end value.
- Sharpe divides excess return by volatility, and depends on the window.
- Sortino counts only downside returns. Drawdown says how deep and how long.
- A fall needs a bigger gain to recover, because the base is smaller.
- Expectancy, not win rate, says what an average trade earns.
- A single record is one draw from a wide band.
Before you trade. Derivatives trading carries a high risk of loss. SEBI's study "Profitability of Individual Traders in the Equity Derivatives Segment (FY25 to FY26)" (August 2026) found that 87.7% of individual traders in equity F&O made net losses in FY26, with aggregate net losses of Rs 91,685 crore. Read the SEBI study