How to read a backtest: drawdown, CAGR, and Sharpe

A backtest shows several numbers at once. This page explains them in the order worth reading them.

Start with the chart#

The growth chart shows how $1 would have grown, next to the benchmark. Before any number, look for the stretches where the strategy fell or went nowhere, and how long they lasted. Ask yourself whether you would have kept following the rules through them.

Drawdown: read it first#

Maximum drawdown is the biggest drop from a peak to a later low. It matters more than it looks, because losses need bigger gains to recover:

The drawdown chart also shows how long each recovery took. A strategy that spends years below its old high is hard to stick with, whatever its final return.

Return: total and yearly#

Total return is the growth over the whole backtest. CAGR is the steady yearly rate that would give the same result. CAGR hides the path: two strategies with the same CAGR can feel completely different to live through.

Volatility and Sharpe#

Volatility measures how much returns swing from day to day, scaled to a year. Sharpe divides the average yearly return by volatility: higher means more return for each bump along the way.

Tactfolio’s main Sharpe doesn’t subtract a risk-free interest rate. The figure under it, over T-bills, subtracts the Treasury bill return first; use that one to compare with other sites.

Compare with a fair benchmark#

Compare with what you would otherwise hold. A strategy that holds tech stocks can look brilliant next to SPY simply because tech did well; that’s why the starter patterns compare with QQQ instead.

Then compare every number, not just return. Less return with a much smaller drawdown can be a good trade. More return with a deeper drawdown isn’t automatically better. The at same risk row of the Reality checks table, such as SPY at same risk, makes this comparison for you: the benchmark held at the strategy’s volatility, the rest in Treasury bills.

Turnover and costs#

Turnover is how much of the portfolio is traded each year. The more a strategy trades, the more costs matter. Tactfolio charges a small cost on every trade (slippage), but not taxes or commissions. The Costs at 20 bps and 40 bps rows show the same trades at higher costs; if the result falls apart there, it depends on cheap trading.

A checklist#

• Does the backtest include at least one bad market, such as 2008, 2020, or 2022?
• How deep was the worst drop, and how long did it take to recover?
• Does the strategy beat its benchmark on drawdown or Sharpe, not only on return?
• Does it survive higher costs?
• Does it hold up on dates you didn’t tune it on? See Why past results mislead.