Thursday 6 February 2020

What is the right way to set stop losses?

Stop losses are the most common method used by traders to control risk. However, they're often used inappropriately. In this post I'll quickly bust some of the myths around them, and explain how to use them properly.

This is the first of three posts aimed at answering three fundamental questions in trading:

  • How should we control risk (this post)
  • How much risk should we take? (next post)
  • How fast should we trade? (final post)
If you find my approach interesting, you may want to read some more about it in my latest book, "Leveraged Trading".

(Note to regular readers of the blog or any of my books: you won't find much new here. But I'm writing these posts in the hope that they will start appearing in Google searches to stop people making silly mistakes. Think of these three posts as the 'gateway drug' to the world of trading professionally).

What is a stop loss?

To make sure we're all on the same page, I'm defining a stop loss in the following way. I use trailing stop losses. 

Assuming you have a long position, you should sell when the price has fallen by more than X below it's high watermark. The high watermark is the highest price that a stock (or future, or whatever) has reached since you purchased it. Of course if the thing turned out to be a complete lemon, then the high watermark will be the purchase price.

If you have a short position, then you should buy (closing your short) when the price has risen by more than X above the low watermark. The low watermark, as you've probably guessed, is the lowest price that the stock has reached since you bet on it going down. I'll mostly frame the examples here in terms of long positions, since they're easier to get your head around, but everything I say is applicable to shorts once you've flipped the language around (long->short, fallen-> risen, high watermarket-> low watermark, highest price->lowest price).

Here's a real example. I bought shares in Go-Ahead group (a UK bus company) in September 2017. Go-ahead group has the ticker GOOG, and everyone I know that isn't a finance geek assumes this is the ticker for Google. But as any fool knows, Google nowadays prefers to be identified as Alphabet and hence has the ticker.... GOOGL. So no danger at all of confusing a modestly sized UK bus company with a global tech titan.

Let's have a look at the chart:

My acquisition price was £15.65 and I set my stop loss at 30% below that:

 (1-0.3)*15.65 = £10.96

(The whys and wherefores of where to set the stop loss will follow later in the post. For now, just take 30% as given).

If the price had subsequently fallen below £10.96 then I would have sold GOOG. But it didn't! In fact it went up, hitting a high water market (HWM) of just under £18 in November 2017. I would have adjusted my stop loss on every high, and it would have reached:

 (1-0.3)*17.89 = £12.52

Then some bad ju-ju hit GOOG and the price fell. Fell below £17, £16... kept falling until it was below the price I had acquired it at (£15.65, if you have a short attention span).

But it never reached my stop loss level of £12.52, so I hung on. In February 2018 the price bottomed out at £13.38. And then it rallied. And rallied some more. And by April it was making new highs, hitting a HWM of £18.10. So I adjusted my stop loss to:

 (1-0.3)*18.10 = £12.67

In fact it continued rallying some more all the way up to £19.64, which meant my stop loss was now £13.75.

I won't continue to bore you with the ups and downs of this stock, but to cut a long story short the most recent HWM was £22.78 set in November last year, and thus my current stop is £15.95. Notice that if the price falls to that level then, assuming I can sell exactly at my stop, I will 'lock in' a profit of £15.95 - £15.65 = £0.30. Not amazing after over 3 years, but better than nothing.

Some other kinds of stop losses

Notice some key features of the trailing stop loss:
  • The stop 'trails' the price, ratcheting upwards with every new HWM
  • Eventually, we're 'guaranteed'* not to lose more than our initial investment, and even to lock in a profit.
  • The most* we can lose at any time is 30% of our profits*
* Caveats! This assumes we can get out exactly at our stop

We can contrast this with a fixed stop loss, where I would have set the stop at £10.96 and left it there. A few scenarios can play out here. The price could plummet to £10.96, and you'd close: exactly the same as the trailing stop. If the price follows the path we've seen here, well it would still be holding the position, so again now change. 

But importantly, if the price now falls down to £10 then the trailing stop will prevent us losing more than 30% off the best price, whilst the fixed stop will mean we always lock in a loss on our initial investment when we close. 

Even if the price never gets down that far we could end up owning stocks forever when there are better opportunities out there; with a trailing stop set at the right level you will eventually close your position and have a chance to redeploy your capital.

A special case of a fixed stop is the breakeven stop. You initially set a stop below your purchase price, but once it's risen a bit your reset it permanently at your entry price. Guess what, there is nothing special about your entry level as far as the market is concerned. This is no better than any other kind of fixed stop.

Some people like to use stop profits eithier instead of, or in addition to, stop losses. A stop profit or profit target is a target price at which you will sell the stock. Clearly this doesn't provide you with any protection against losses, so you need to have a stop loss as well. Even then I have two issues with using profit targets. Firstly, they add unneccessary complexity to your trading. Secondly, I don't know where a stock might end up, so why set a target? 

Consider Intermediate Capital Group (ICP), which I bought at £4.26 in July 2016:
A typical setting for profit target is twice your stop loss, i.e. 60% of the purchase price or £6.82. ICP is now over £17, a rise of 300%! That's an awful lot of profit I would have missed out on. 

Note: For some kinds of mean reversion systems profit targets make sense, but they should be set using a proper analysis of the mean reversion process rather than a single figure. They make no sense at all if you're looking for trends.

Another stop loss strategy I am not keen on is the time based stop. You hold your position for 3 days or 3 months, and then close it if it hasn't reached some kind of profit target. Unless you're trading options where timing is inherent you shouldn't normally try and predict exactly when something is going up. It's hard enough predicting if it will go up. 

Of course any kind of stop loss is better than no stop at all. Without a stop loss I may well have sold GOOG at £18 to lock in my profits, missing out on the additional profits I've made subsequently. Or I could have panicked when it dropped below £13, and sold at a loss.

Where should we set our stop?

The figure of 30% above came out of thin air. But how in reality do we calculate this number? 

Here's an interesting quote:

"Each system has its unique and optimal betting percentage."

Is this really true? Do we have to risk calibrating/fitting the optimal percentage using a back-test? Or can we set the relevant figure with some simple rules? (Spoiler- the latter)

Let's review some facts:

Capital loss: A wider stop (more than 30%) means we'll lose more money. For example if we've invested half our account in a stock, then we're risking 30%*50% = 15% of our capital. Widening our stop will increase the likely damage to our account when the position is closed, and vice versa.

Time: The longer we hold a trade for, the more likely a stop will eventually be hit. GOOG is unusual; most of the trades I've done using this stop-loss last for about 6 months to a year.
The tighter a stop is, the quicker we'll hit it. For example if I set my stop for GOOG at 10% (i.e. initially at £14.09, then ratcheting up to 10% below £17.89=£16.10) I would have been stopped out in mid June 2018. Conversely, if I widen the stop enough then it might never be hit (a 99% stop will only be hit if a company is completely wiped out).

Volatility: If something is more volatile, it is more likely to hit a stop at a given level. Consider Bitcoin, here represented by an ETF:
It takes a few months -or even years - for most shares to move 30%, but 30% moves happen in a few days for this crazy POS asset.

Magic beans price level: We want to set our stop at the level where, if the price breaches it, it is guaranteed to keep falling. You may notice from the subheading that I am somewhat skeptical of this. Things like using fibbonaci numbers as 'key resistance' levels are pretty bogus. You may want to use a rule to tell you when the market is likely to fall, but it's better to drop the stop-loss entirely if you're going to use such a rule (this is discussed in chapter nine of "Leveraged Trading").

Let's start with the issue of capital loss. A common method for setting stops is to do so such that N% of your capital is at risk, eg N=1%. So for example if you bought 100 shares of Apple, that would be a position of $32,000 based on the current price of ~$320. If you had invested all your capital in Apple, then that would correspond to a 1% price move: $0.32. You're likely to be in the trade for less than a day.

This leads to traders saying stuff like this:

Yes, but I am trading low float stocks not something like forex where that 1% actually makes sense. If I put a stop loss at 1% on stocks 9/10 times it will probably just stop me out.

This trader appears to have a hobsons choice! They can set their stop appropriate to the volatility of the market at say 30% for low float stocks... but then they will be risking too much of their capital. Or they can set their stop appropriate to their capital, and risk being stopped out too quickly.

This makes no sense! Instead what you should do is:
  1. Set the size of your position according to your capital, and the ratio of your risk appetite and the risk of the underlying market.
  2.  Then set your stop according to volatility and time horizon.

Let's explore that. Suppose for example that you are happy to run your account at about the same level of risk as Apple shares, that you typically own two shares, and that you have $30,000 in capital. This implies that you should risk half your capital: $15,000 on your Apple stock position. At the current price of $320 that equates to 46 shares. 

Now you are free to set the stop-loss at a level which makes sense for the market. You should set the stop-loss at X * volatility of the market. Where does X come from? A larger X will mean wider stops, so you will hold positions for longer. A smaller value of X will result in holding positions for less time. You should set X according to how long you want to hold positions for.  I'll come back to that decision in in a future post. 

For now, let's use X=0.5, and assume we measure volatility as the annual standard deviation of the market which for Apple is around 20%.

Note: In subsequent posts I'll explain how you can use different measures of risk, such as the daily ATR, to calibrate your stop losses.

0.5 of 20% is 10%, so the stop loss for Apple should be set at 10% below your entry price. What proportion of our capital is at risk if we hit our stop loss? Well, it will be 10% of $15,000; $1500. Which is 5% of our capital.

Notice that the stop loss will have nothing to do with the size of your account. A small trader will have the same stop loss as a large trader, but a smaller position. A trader who has more appetite for risk will have a larger position, but the same stop loss. You can set your stop loss differently for FX and stocks without worrying about using too much capital. 

To see why this is a superior method, consider what will happen if you decide to invest in a Bitcoin ETF. Or maybe that should be 'invest'. Or to be accurate, invest in gamble on.

Suppose that Bitcoin is four times as volatile as Apple. We had a risk target for our account which was the same as Apple, 20% a year standard deviation of returns. But Bitcoin has four times the risk, 80% a year.

That means your position in Bitcoin will be a quarter of the size to compensate for the extra risk: $3,750. 

The stop loss will be set at 0.5 * 80% = 40% for Bitcoin. Much wider, to reflect the higher chance that Bitcoin will make a big move. 

How much of our capital is at risk? A 40% move on a $3750 position will cost us $1500. That's the same as for the Apple trade! Setting our position size and stop loss independently means we can target a particular % of our account at risk, whilst setting our stop differently according to market conditions.

Notice also that the same stop-loss can be used for different instruments, and it will adjust automatically if the volatlity of the market changes. A stop that made sense in the quiet days of 2006 would have been far too tight in the madness of late 2008!

Another benefit of this method is that you can switch to a trading system that doesn't have an explicit stop loss (like my automated futures trading account), and your risk will still be accurately sized.

Some unanswered questions

In this post we've learned that position size and stop levels should be set independently, and that stop levels should only depend on your expected holding period and the volatility of the market. 

But! There are hefty two elephants left in this particularly crowded room, namely:

  • How big should our positions be: how much risk should we take on our account?
  • What should the stop loss multiplier 'X' be: How fast should we trade
I will answer these questions in the next couple of posts (number two and number three). 


  1. Hi Rob, I think every trading approach should use different stop loss tactics. I remember the ATR (volatility) stop loss being use by the Turtles also as a position sizing technique. For example with overnight trading I use a fixed price based stop loss very large that triggers once in a while, like here:
    Marco Simioni
    Nightly Patterns

  2. Rob, I liked your Systematic Trading book.

    A question about a staunch systems trader:

    Suppose you have a daily trading system (say, 3 variants of EWMAC and 1 of Carry), you have a long position, and then you get stopped out. But then the next day, you get another combined long signal for some reason. Do you typically employ a "waiting period" such as a week to a month to allow an another position with that instrument or do you jump back in based on the combined long signal?

    1. Hi Andrew
      I don't use stops in the book "Systematic Trading". But in my third book "Leveraged Trading" I do. And in that book I say that you shouldn't go into a trade if you have been stopped out of a trade in the same direction; instead wait for the indicators to switch sign.

      I prefer not using stop losses; in which case you close your position when the indicators have flipped sign and you'd immediately and always open one up in the opposite direction (ignoring the issue of discrete positions of course).

    2. Hi Rob,
      I read your two trading books, they are great.
      About stop losses, if the "instrument risk" is updating every day with 20 days volatility lookback window(so does the stop loss gap which can grow large), how does each trade still risk the same portion of capital when "instrument risk" is changing while the exposure was fixed at entry?

    3. This is discussed at length in my third book Leveraged Trading; basically you shouldn't update the stop loss as vol changes if you aren't also going to change the position size, otherwise you will get the problem you describe.

  3. Hi Rob,
    I have a strategy that uses trailing stops. Its based on volatility stops so if there is a wick to the trailing line but close above it the trend doesnt flip.
    I am slightly stuck on if putting stops in and updating it everytime the trail gets closer is a better option or to not put hard stops in and wait for the trend to reverse then open a trade in opposite direction.

    There seem to be positive and negatives to both approaches as if i put hard stops in, then it can just wick to them and continue on. So would need a re-entry mechanism if I want to ride that part of trend.
    Or if I dont use any stops and wait till a reversal happens then it leaves me vulnerable to price shocks.

    Which approach would you recommend is better ?

    1. I don't know what a 'wick' is. Also don't know what you mean by 'hard stops'.

    2. I think what he is asking (which I am interested as well) is "should I set stops on closing price only?" (which prevents the problem of "wicks" taking out the stop - the price hits the stop intraday but closes inside of it, thereby taking the position out too early). If you look at a candlestick, a wick is the thin little line (up, down or both directions) that indicates price traced in a certain direction, then retraced intraday and closed away from the high/low.


Comments are moderated. So there will be a delay before they are published. Don't bother with spam, it wastes your time and mine.