Stop-Loss Doesn’t Work (Mean-Reversion And Stops)

Last Updated on April 19, 2022 by Quantified Trading

Do stop-losses really work? We are “brainwashed” to believe that you should always have a stop in place. But, perhaps surprisingly, stop-losses normally just make a strategy perform poorly.

Stop-loss doesn’t work for the majority of trading strategies. A better solution is to diversify into different markets and time frames. Moreover, always trade smaller than you’d like.

Many traders swear to stop-losses by default, but to me, it doesn’t make sense because:

  • If you buy something at 100, why should you just put in an arbitrary stop-loss level? If you bought something at 100, presumably it’s because of value or some quantified reasoning. If you use stop-loss it needs to make sense where to put your stop-loss level.
  • If you trade stock ETFs/futures, it means the risk premium in the market rises when the index falls. This means more likely higher returns in the future. Why sell at a lower price than you bought when the expected future return increases?
  • I have yet to see many strategies that perform better with a stop-loss. However, I’m mainly a mean-reversion trader, not a trend follower.

The fear of giving up profits

One of the first trading books I read was Trading For A Living by Gary Smith. I still remember one of his quotes:

Because of my fear of losing trading capital, I would either grab the smallest of profits or get scared out by the slightest of reactions.

I tend to agree with Mr. Smith. Moreover, in The Way Of The Turtle by Curtis Faith, he concludes the same as Smith. To give up part of the profits are normal, but this is such a painful experience for many traders that they chose to use stop-losses even if backtests don’t support it. If you are affected by losing money, you start fiddling both stop-losses and variables in your strategies.

How to make a stop-loss without having one

I prefer to arrange my risk in other ways than by using stop-losses. As a trader, you must protect your capital, but I suggest you look into other ways of creating “stop-losses” such as:

  • Trade as many strategies as possible.
  • Try to use/find non-correlated strategies.
  • Trade different markets/products.
  • Use different time horizons.
  • Use both long and short. Even though shorts might not work as well as long, they might serve a purpose as hedging and a way to hang on to your long strategies.
  • Trade lower size to avoid getting scared out of a position. Most traders underestimate the impact of potentially irrational decisions by trading too much size. A backtest never captures how you feel in the middle of a drawdown. Always trade smaller than you like.

Some weeks ago I published this simple mean-reverting strategy on Nasdaq. Without any stop-loss, the strategy returns 1.7% per trade, 12.34% annually, and has a maximum drawdown of 19.51%. The drawdown is big given you allocate 100% of your equity to this strategy, but as part of for example a portfolio of 10 strategies the drawdown is much more reasonable assuming the strategies are somewhat uncorrelated.

The table below shows the result for the same strategy using different level of stops (exit is done at close if the stop level is hit):

Stop level % per trade Annual returns Max drawdown
0.5% 0.79 8.17 -8.66
1% 0.92 8.99 -8.66
1.5% 1.05% 9.87 -8.66
2% 1.16 10.6 -8.81
2.5% 1.24 10.83 -8.81
3% 1.34 11.08 -8.81
5% 1.55 11.94 -10.95

At a 5% stop it seems the strategy does not deteriorate so much. I forgot to include the profit factor in the table, but that one also deteriorated with stop-losses.


Stop-losses might serve other purposes, for example, your faith in trading a system, but I suggest you look into the other options I have mentioned in this article.

I let Curtis Faith get the last words of this article. On page 145 he sums up his findings on stop-losses:

“Note that the zero case, which means no stop at all, has the best MAR ratio numbers. In fact, the test with no stops is better for all the metrics: CAGR%, MAR ratio, Sharpe ratio, drawdown, and length of drawdown – every single metric. The same things holds true for a test of the Triple Moving Average system: Every single measure was worse with any stops. The same test of stops applied to the Donchian Trend with time-based exit system yields similar results except that for very large stops of 10 ATR or more, the results are about the same as those for a test with no stops. This certainly goes against the common belief that one must always have a stop.”


Similar Posts