Editor's note: Originally published on July 29, 2014.
This topic draws passionate opinions from seasoned investors on both sides of the argument. We have encountered short investors who cut losses if a stock moves 8% against them and we have encountered investors who will add to losing short positions if they still believe in their original thesis. The question is whether or not to use mechanical stop-losses on short positions. Our research shows that mechanical stop losses hurt more than they help.
We tested stop losses based on mechanical percentages of the entry price, using 15%, 20% and 30% as baselines. (These tests were done on our published short stock recommendations.) In addition, since fixed percentages ignore the relative volatility of different stocks, we test stops based on multiples of a stock's average true range (ATR). These tests used stops ranging from 2 to 5x a stock's recent ATR. In both cases, the stops were trailing stops, ie., they moved with the stock price if the price moved favorably.
The conclusion was that narrow stop losses forced an investor out of positions too early and didn't let the shorts "mature", resulting in lost opportunity. On the other hand, very wide stops proved ineffectual in reducing volatility. On net, there seemed to be little advantage in using stops at all, at least on the stocks from our recommended list.
In our opinion, risk control is better served when it is based on diversification and position sizing. Note that diversification should consider factor exposures such as momentum, market cap and growth profile, among others and not just diversification by sector or industry.
For further study: Would stop losses based on time perform better? The idea is that if a short position doesn't "work" within a certain defined time frame, the position should be covered. The concept is worth backtesting, but we can see problems already in that shorts based on an upcoming earnings miss and shorts based on declining secular business trends clearly have much different investment horizons.