CAGR and maximum drawdown: read the path behind the return
Two strategies turn 100 into 150. One falls to 95 along the way; the other falls to 60. They share an endpoint but not the experience of holding them. A single annualized return hides that difference.
CAGR is an equivalent growth rate
An invented balance grows from 100 to 144 over two years: total return is 44%, while CAGR is √1.44 − 1 = 20%. That calculation uses endpoints and elapsed time, not the path between them.
A first-year loss of 20% followed by an 80% gain also takes 100 through 80 to 144. Its CAGR is still 20%. CAGR does not mean a steady 20% return each year.
Drawdown starts from the preceding peak
A balance rises from 100 to 120 and then falls to 90. It is down 10% from inception, but the drawdown from its peak is 25%. Maximum drawdown is the deepest observed peak-to-subsequent-trough decline.
Bitcoin’s distance from its price high and a strategy’s equity drawdown concern different curves. Similar arithmetic does not make them the same risk record.
Recovery is asymmetric
A fall from 100 to 80 loses 20%, but recovering from 80 to 100 requires 25%. From 50, recovery to 100 requires 100%. The denominator changes, so equal percentages do not cancel.
Observed maximum drawdown is not a ceiling on future losses. End-of-day equity samples can also miss deeper intraday declines.
Short samples make annualization fragile
In an invented example, a 10% gain over 30 days annualizes to about 219% using 1.1^(365/30) − 1. The actual gain remains 10%; the larger figure assumes the short period can repeat.
For a sample shorter than a year, read the actual period, total return and number of trades together. A complete record across changing market conditions is a more useful comparison than a selected strong window.
Check costs and evidence before ranking results
Match the period, allocation and leverage before comparing fees, slippage, funding and valuation. Costs affect outcomes; a gross research result and a net backtest are not directly comparable.
Win rate also needs payoff sizes. In an invented sequence, nine gains of 1 and one loss of 12 give a 90% win rate and a total loss of 3 before costs. Missing trade statistics should remain missing rather than being inferred from a smooth curve.
Concept references and examples
Prices, amounts and scenarios are invented teaching examples, not historical market observations, backtest results or return promises. These references support the concepts; they do not identify the supplier of this site’s market series or imply endorsement.
Sources checked: 2026-10-03Now read the number in its context.
The guide explains the reading; the dashboard provides dated observations. Put them together and make your own decision about participation.
Explore strategy backtests →