A little understanding goes a long way
What does a backtest
actually tell you?
It asks a historical question: if these exact rules had been followed, with this data and these assumptions, what would the simulated result have been?
Start with rules you can explain.
“Buy when it feels right” cannot be reproduced consistently. A rule such as buying when a completed daily close moves above a 20-period moving average is more specific. You still need to define the indicator calculation, evaluation time, execution timing and sell condition.
Compare the same dates.
A strategy’s return needs a reference. Buy and hold means investing in the same asset at the start of the test period and keeping it through the end, under the stated benchmark assumptions. Compare both over the same dates and with consistent cost assumptions. A return difference is measured in percentage points.
If a strategy returned 12% and holding returned 8% over the same dates, the strategy finished 4 percentage points ahead. That describes that simulation, not what will happen next.
Costs change the story.
Commissions, spreads and slippage can reduce simulated returns. Position sizes, available liquidity and execution timing matter too. A result is only as useful as the assumptions it includes. A percentage return does not make those assumptions disappear.
Look at the journey as well as the ending.
A profitable ending can include large temporary declines. Drawdown measures the fall from a previous account high. Trade counts, the time spent invested and the length of the test help you interpret how much evidence supports the result.
History is evidence, not a promise.
Changing rules after seeing the results can make a strategy look better on the past without making it more reliable. Historical data can have gaps or revisions. Real orders may fill differently. Keep the rules, test dates, engine version, dataset version and settings attached to the report.