Backtesting

In simple terms

A backtest asks: if I had followed these exact rules in the past, what would have happened? It is the main way to check a chart-based idea before risking money on it.

A backtest can be honest research or a sales pitch, as the misleading backtests lesson explains. This lesson describes how to do the honest kind. The next lessons cover the biases that distort backtests and the problem of overfitting.

Step 1: write precise rules

"Buy breakouts with good volume" cannot be tested. A testable rule specifies everything:

• Universe: which shares, chosen how, and when (for example, all shares in a given index at each date).
• Entry: for example, a daily close at least 2% above the highest close of the previous 60 sessions, with volume at least 1.5 times its 20-day average.
• Exit: a stop at two ATRs below entry, a target, a trailing rule or a time limit.
• Position size: for example, risking 1% of capital per trade, as the position size lesson explains.
• Execution: at what price you assume trades happen, for example the next day's open, not the close that triggered the signal.

If two people following your rules on the same data would get different trades, the rules are not precise enough.

Step 2: use clean, adjusted data

Use prices adjusted for splits and bonuses, as the adjusted prices lesson explains, so that corporate actions do not appear as crashes. Include shares that were later delisted or removed from the index, which the biases lesson explains.

Step 3: include every cost

The trading costs lesson explains the charges. Rates change, so take current ones from official sources when you test.

Step 4: measure the results

Look at several measures, not just total profit:

• Number of trades: too few means results are mostly luck.
• Win rate, average win, average loss and expectancy after costs.
• Total and yearly return on capital.
• Maximum drawdown: the largest fall from a peak in capital.
• Longest losing streak.
• Exposure: how much of the time money was invested.

Step 5: compare with a benchmark

A strategy that made 9% a year sounds good, unless simply holding an index fund over the same period made 12% with less effort and similar risk. Always compare with simple alternatives over the same period:

• holding a broad index;
• holding the same shares the strategy traded;
• random entries with the same exits, which shows whether the entry rule adds anything.

A test the idea can fail

Decide in advance what result would make you reject the idea, for example net expectancy below zero, or a return below the benchmark. If you would find a reason to accept any result, the backtest is not a test.

Why it matters

Backtesting is how traders and fund managers check ideas before risking money. Done honestly, it often shows that an appealing chart idea does not survive costs or beat a simple benchmark, which saves money. Done carelessly, it produces confident numbers that lead to losses.

Limitations

Even a carefully built backtest describes the past. Markets change, costs change, and a rule that worked for a period may stop working. Backtests also cannot capture everything about real trading, such as the emotional difficulty of following rules through a losing streak.

Common mistakes

• Vague rules that leave room for judgement after the fact.
• Ignoring costs and slippage.
• Assuming trades happen at the signal price rather than the next available price.
• Not comparing with a benchmark.
• Testing on too few trades.

Try it yourself

1. Write a moving-average crossover rule precisely: universe, entry, exit, size and execution price.
2. In a spreadsheet, apply it to five years of adjusted daily prices for one share.
3. Count the trades, and calculate the result before and after an assumed cost per trade.
4. Compare the net result with simply holding the share for the five years.