How to test a trading rule before you trust it

Every trading rule looks good in the moment you think of it. The point of testing is to find out whether it still looks good once the market has had a say. This is the order we use, and each step catches problems the one before it can’t.

1. Write the rule down first

Before you look at any results, write down what the rule does and what would count as a pass. “Buy when the 20-day average crosses above the 50-day, sell on the opposite cross” is a rule. “Buy when it feels oversold” is not.

Write the pass mark down too, for example: profit after fees, on data the rule wasn’t tuned on, with a drawdown you could live with. If you decide what counts as success after you see the numbers, you will almost always find a way to call it a success.

2. Backtest it, with fees

A backtest runs the rule on past prices and shows what it would have done. Three things make it honest:

• Count every cost. Fees and slippage are paid on every trade. A rule that wins most of its trades can still lose money if each win is smaller than a round trip costs. The total costs line is the one to read first.
• Keep some data back. Set the rule on one stretch of history and check it on another it has never seen. A rule tuned on all of the data will always look better than it is.
• Look at the bad days, not only the total. Max drawdown tells you how far the account fell from its peak. Ask whether you would have kept the rule running through it.

If you try many versions and keep the best one, remember that the best of many tests is partly luck. Re-check the winner on data you haven’t used yet. Reading backtest results covers the rest of the report.

3. Run it on paper

A backtest can’t show you how the rule behaves on prices that haven’t happened yet. Paper trading can. The rule runs on live market data with simulated money, so you see real fills, real timing and real gaps without risking anything.

Give it enough time to see a few different kinds of market. A week of calm prices tells you very little about a crash.

4. Go live small, with limits

When you do move to real money, start with an amount you can afford to lose, and set the limits before you start: a stop-loss on each position and daily risk gates that stop the bot after a bad day. Stops can slip in fast markets, so size the position as if they might.

Then compare live results with the paper ones. If they drift apart, find out why before adding money.

What testing can’t do

No test proves a rule will work in future. A rule that passed every step has survived the test so far, and markets change. Testing is how you find the bad rules cheaply, before they cost you real money.