What a stop-loss does to backtest results

A stop-loss is a rule that sells a position once its price falls a set amount below the purchase price or a recent high. Its purpose is to cap the loss on any one trade. In a backtest it changes far more than the size of the losses, and not always in the direction people expect.

This note explains the main effects conceptually. It gives no figures of ours, because the published cards do not report results with and without stop-losses.

The kinds of stop-loss

A fixed stop sells if the price falls a set percentage below the entry price, such as 10%. A trailing stop sells if the price falls a set percentage below the highest price since entry, so it rises with the trade. A time stop sells after a set period whatever the price. A volatility stop sets the distance using how much the price usually moves, so it is wider for jumpier stocks.

What it usually does to a backtest

It cuts the largest single-trade losses, which can reduce the maximum drawdown.

It increases the number of trades, because positions that would have recovered are closed and later re-entered. More trades mean more costs in real life, which a no-cost backtest leaves out.

It often lowers the win rate, because some trades that would have finished in profit are stopped out during a temporary dip first.

It can raise or lower the return. In markets with long, deep falls, stops can help. In markets that fall sharply and recover quickly, stops sell near the low and miss the recovery.

Gaps: the problem backtests miss

A stop-loss is an instruction to sell at the next available price once the stop level is reached, not a guarantee of that price. If a share closes at 100 and opens the next day at 80 after bad news, a stop at 90 sells near 80. Backtests built on daily closing prices can assume a fill at the stop level that would not have been available. Results that rely on tight stops are the most exposed to this.

Stops and buy-the-dip rules pull against each other

A rule that buys after a fall is betting that the fall reverses. A stop-loss sells if the fall continues. Combining the two can mean buying a dip, being stopped out lower, and missing the eventual recovery. Whether that helps depends heavily on the period and the stock, which is why it needs testing rather than assuming.

How to test a stop-loss fairly

Fix the stop rule before looking at the results. Test the same strategy with and without the stop over the same dates, and compare return per year, maximum drawdown, number of trades and time in the market. Try several stop distances, and be wary if only one distance works. Include realistic costs, or at least note that they are missing. And compare both versions with buying and holding.

Reading other people's stop-loss results

Check whether fills were assumed at the stop level or at the next available price, whether costs were included, and whether the stop distance was chosen after seeing the results. A stop that looks perfect on one stock over one period is often a fitted detail rather than a rule that travels.

Where to see our tested results

Our ticker pages show, for each published ticker, how often the strategies we test beat buying and holding, with the drawdown and return figures alongside. Our note on maximum drawdown explains how to read the drawdown figures, which are the ones a stop-loss is meant to improve.

Choosing a stop distance

A tight stop, say 3% below entry, will be hit often by normal daily movement, especially on volatile shares, so it creates many small losses and frequent re-entries. A wide stop, say 25%, is rarely hit, and when it is, the loss is large. Many traders set the distance relative to how much a share usually moves, so a calm fund gets a tighter stop than a volatile share.

Whatever the method, the distance should be chosen before the test and kept fixed across tickers. Picking a different distance for each ticker after seeing the results is a form of overfitting.

Stops in an ISA or pension

Stop-loss orders work the same way inside a tax wrapper as outside it, but not every platform offers them, and some charge for them. Some only offer stops on certain markets or during market hours. Check what your platform supports before building a plan around a type of order it may not have.

One more point is worth keeping in mind. A stop-loss is a decision made in advance about when to admit a trade was wrong. Its real value is often in removing the need to make that decision under pressure. That value does not show up in a backtest at all, and neither does the cost of abandoning the rule the first time it sells at the bottom.

This is research, not financial advice. Past results, real or simulated, do not tell you what will happen next.

Does a stop-loss improve returns?

Not reliably. It can reduce the largest losses, but it also sells positions that would have recovered, adds trades and costs, and can lower returns in markets that recover quickly.

Why do backtests overstate stop-losses?

Many assume a sale exactly at the stop price. When a share gaps down overnight, the real sale happens lower, so tight stops look better in a backtest than in practice.

What is a trailing stop?

A stop that sells if the price falls a set percentage below the highest price since the trade began, so the stop level rises as the trade gains.