How to read a quant strategy backtest

Short answer: Do not judge a backtest by return alone. First identify the market period, then examine maximum drawdown, trading costs, out-of-sample behavior, and the worst regime. A backtest is useful only when both return and risk are explainable.

What a backtest actually answers

A backtest asks: “What would have happened if a defined rule had run on a historical dataset?” It is not a future promise and does not prove that a strategy has survived live execution.

Its best use is to understand behavior:

• when the strategy participates;
• how wins and losses are distributed;
• how long losing sequences can last;
• how far the equity curve may fall;
• how sensitive results are to turnover and costs.

Check one: period and market regime

A test covering only a rising market says little about sideways or falling conditions. Ask:

1. When does the test start and end?
2. Does it include high volatility, low volatility, trends, and ranges?
3. Which assets, sessions, and data frequencies are used?
4. Could missing data, survivorship bias, or hindsight selection affect the result?

The date range is not a decorative field. It defines what the result can support.

Check two: maximum drawdown

Maximum drawdown measures the largest fall from a prior equity peak to a later trough. It often describes the real user experience better than total return.

Two strategies may both end up 20%, while one falls 6% on the way and the other falls 35%. They require different capital, risk tolerance, and stopping rules.

Review:

• drawdown depth;
• drawdown duration;
• time required to recover;
• market regimes in which drawdowns cluster.

Check three: costs and execution assumptions

A high-turnover strategy can look much better when fees, spread, slippage, and fill constraints are omitted.

A credible description should explain the fee model, how market and limit orders are simulated, whether large orders are assumed to fill at one price, and whether signal-to-fill delay is represented.

Check four: in-sample and out-of-sample behavior

In-sample data helps form or tune the strategy. Out-of-sample data tests whether behavior persists in a period that did not participate in tuning. Strong in-sample results alone may describe overfitting.

A more robust process reserves unseen data and uses rolling or segmented tests to examine parameter stability.

Check five: worst periods and stopping conditions

Find the worst month, longest losing sequence, and most difficult regime. Then convert those observations into operating rules:

• when to reduce allocation;
• when to pause;
• when a new backtest is required;
• which deviations suggest the original assumption no longer holds.

Backtest checklist

Using backtests in QANTERION

Backtests should sit beside allocatable capital, strategy thresholds, and account risk — not act as a standalone leaderboard. Continue with How to choose a strategy by risk budget and review the risk disclosure.