Choosing where to put a trailing stop-loss to avoid being stopped out "unnecessarily"!

19th Dec '16 by Nick Ray

28 comments 19385 reads

If you wanted to put a trailing stop under your stocks to allow them room to be volatile but to get you out before real trouble appears, where would you put it?

For many people -20% seems to be a 'safe' choice. It is away from the day-to-day noise and by the time you are -20% down it feels as if any further would start to feel like a long way to come back.

But what if we measure typical drawdowns? Can we apply a bit of science to find some sweet spot which will be less likely to take us out on spurious noise but is a good breakpoint before disaster strikes?

After being surprised by some drops in stocks which I thought were clearly high quality (for example GB (LON:GBG) fell by 38% on what was barely even bad news) I decided to measure how big the drawdown over 256 trading day periods (approximately a year) really is in typical cases.

As they say in clickbait articles, the results will surprise you...

Here is a histogram of the 256-day drawdown for all shares with an MCAP>£100M and spread<5% over about the last three years:

The red line shows the 'median' drawdown is at -29%. That is lower than I expected. In my mind stocks didn't draw down more than -20% unless they were heading for serious trouble.

But maybe good, high-Q stocks perform significantly better than this?

Well, it looks like they are less likely to go bust but the median is now -28%. Hardly an improvement. And why are there little peaks at about -25% and -30%. Are these stocks triggering stops and bouncing back? Let's try Q>85 and see what happens:

The median is now -27%. It has barely budged. Again there are mild hints that these "good" stocks bust stops at -20%, -25% and -30% and then bounce back. High-Q stocks seem to be less likely to go and have a total collapse, but their high quality does not stop them having very similar drawdowns to any other stock.

Now you might argue that in fact what was happening was that some high-Q stocks were going 'bad' and that is why the drawdown histogram looks like it does. So I simulated 3000 stocks with a daily mean return of…

Unlock with Email

Unlock with Google

Already have an account?
Login here

Nick Ray

19th Dec '16

Hi Ramridge.

I agree with you about illiquid stocks. In this analysis I only used stocks with MCAP>£100M and Spread<5% specifically to exclude that issue.

The limits you are using are much tighter than would work for me. I suspect that you are happier with shorter holding times than I am and that is why it works for you.

Redrichmond

21st Dec '16

Hi there, I have done some reading on this from recommended stockopedia books

One being "Trade like a stock Market genius" by mark minervini

He suggests (pg 300-311)

Tighten up stop losses. If you normally cut losses at 7 to 8 percent,cut them at 5-6

If you get stopped out at 20% , that means that stock has to go up at 22% + to just get your money back (you still are not in profit (approx spread included-broker commission)

You then need it to go up another 10-15% to make worthwhile money

So stop loss at 20% means it needs to go up 35% for you to make a decent profit..Sometimes this will take years

I tend to look for trending shares that are not so volatile (BURFORD CAPITAL) is one, and has a huge market cap,still black swan events can effect it

I set the stop loss at 5% - recently I got stopped out but thats fine as I made 22% profit in a few months

I use charts and moving averages and williams r% to time the moment to go back in

Over the years of my mistakes I am realising that its the human element thats lost me money

Personally I think physiology of buying and selling is a huge factor to my profit ..and loss

iwright7

Red,

I personally think 5% stop loss is too tight and I would chose -15% or maybe -20% before bailing out of what I previously believed was a good stock. In the case of Burford Capital (LON:BUR) staying put has gained me +145% to date. Agree with you about buying trending shares, but it normally pays to run the winners.

The stop loss sell exception would be companies where the outlook or circumstances have materially worsened, or where a share price has been flat for several months and I think there is better potential gain elsewhere.

Rusty2

I agree, with a 5% stop loss you will be getting stopped out all the time. Spread and dealing cost/stamp duty often around 3%. I do hold Burford Capital (LON:BUR) as well, it's up over 200% but I did top slice at 585p, I find it better to sell after a big rise than after a fall. Could have sold out at 460p a few weeks ago after the 10% fall. Yes best to run your winners and top up on dips, top slice after big rises. I think the reference to 7-8% stop loss was on US shares in a bull market. I looked at NAPS 2016 with a 10% trailing stop loss 75% of the shares would have been sold, overall you would have lost money, if you did nothing you would have made over 3%.

PhilHofGrowYourDough

Here's an ichimoku Chart for Burford.

No Close below the cloud in the chart which starts back in Feb 2016

Cheers
Phil

The 256-day drawdown on Burford Capital (LON:BUR) is -12.1% which is very low. (The drawdowns over 10% for Burford Capital (LON:BUR) occur roughly where the price touches the cloud in PhilH's chart above).

If you are trading short term (holding for a few days or weeks) then probably 5% stop-losses are needed. But if you are trying to operate some version of the NAPS strategy I would say that 5% is too close.

In more recent times for stocks with big price surges I've been experimenting with shifting my trailing stop loss from the base of the cloud to the Tenkan Sen (Fast Moving Average). I use the '%50daMA' and 'price change % 1m' fields to determine when I should (de)activate this rule. I've been using this approach to take some profits rather than sell out completely.

22nd Dec '16

Earlier in the thread someone asked whether stocks which suffer a significant drawdown tend to recover from it or not.

If a stock recovers from a drawdown, that means that the stock achieves at some time in the future a price higher than its previous 256-day high (or whatever ever period you choose). Regardless of the size of the drawdown the only question becomes whether the stock achieves higher values in the future. The size of the drawdown then simply characterises the depth of volatility likely to be experienced. High-Q stocks seem to have a similar median drawdown compared with other shares, but less likely to experience very deep (>60%) drawdowns. However when that happens it is likely that the fundamentals have also changed so it may not make sense to ask whether high-Q stocks which experience -60% drawdowns recover without asking why the fall happened. (Whereas falls near the median we might surmise are not unusual.)

Someone else mentioned that if a stock falls by -20% it needs to recover by +22% to get back to where it started. (And I have seen this using more dramatic moves: for example a -33% fall needs +50% to recover and -50% needs +100%.)

So I decided to look at the opposite of a draw-down - a "gain-up" you might call it. Just to revise for a moment, a drawdown is the biggest fall that the stock has suffered from its highest price over the previous 'n' days. (I am using 256 working days in these histograms, and I look at the biggest drawdown seen in the last 256 days).

Similarly, the "gain-up" can be defined as the price increase from the previous lowest price in the last 256 days, and we can compute the greatest "gain-up" seen over the last 256 days. It is important to stress that I am not talking about the %-change over 256 days here. We first find the lowest price in the last 256 days and then look at the gain from there. This is the exact mirror image of the definition of drawdown. So using the examples above, a drawdown of -33% followed by a gain-up of +50% would exactly cancel out, as would a -50% drawdown followed by a +100% gain-up.

Here are the drawdown and gain-up plots for the allshare and Q>75 case over three years. The median gain-up is 54% for all shares and 57% for Q>75 shares. It is not far off the reciprocal of the drawdown which we previously saw was -29% (all-share) and -28% (Q>75). It looks as if recovering from a -29% drawdown is not uncommon. (Bear in mind that the "gain-up" is defined as the improvement from the lowest point so a gain of +57% is not as surprising in that context.)

This does not mean of course that the stocks which fall the most are also the same ones which rise the most. But there is some kind of balance in aggregate. (Does this suggest that in a NAPS portfolio if you top-slice at +50% you should stop out at -33% (or any other reciprocal pair) because that harvests the positive "drift" component from the portfolio? Food for thought there.)

As ever, this kind of analysis is really only useful if your style involves holding for about a year or more.

Nick de Peyster

Wouldn't you want to scale it on a stock-by-stock basis, using recent ATR or VAR?

Nick de Peyster
http://undervaluestocks.info/

mpat89

23rd Dec '16

Nick, I think that should only be considered if you used recent ATR or VAR to determine the position size.

14th Feb '17

Hi Nick,

Are you able to assess the speed of the gains vs the speed of the drawdowns.

If the gains occur more rapidly then there maybe an argument to cut your losses sooner in order to maximise gains?

Regards
Phil

Hi Phil

Not quite sure what you mean by 'speed' in this context, but I think in general trying to create a strategy which works by reacting faster than the market is impossible. The usual pattern seems to be slowly building gains and the occasional sharp gap down.

These days I tend to focus on the performance of a portfolio as a whole and not worry too much about individual stocks within it between rebalancing points. The stock-specific risks tend to cancel but you cannot get rid of market risk and that is the real problem. If the market decides to make a correction, by the time it is happening you are too late to beat it. If you could reliably predict when it was about to happen, so could everyone else and it would happen that extra day sooner.

My thoughts aren't well formed here but let's see if this makes any sense ...

From you charts it seems that the median drawdown is 33%. So doesn't that mean if I'm currently sat on a 5% drawdown that it is more likely to increase to 33%

Conversely if I'm sat on a 5% gain then that is more likey to convert into a 50% gain?

Therfore I'd be better off selling the loser and backing the winner?

This would be even more important if the creep towards a 33% loss took longer than a gain of 50% because I'd miss the quick gains?

Maybe this makes no sense?

15th Feb '17

If you use a model of stock prices similar to Geometric Brownian Motion (which is a common assumption) then the stock price movements have no memory. So the fact that it has fallen by -5% does not of itself influence where it goes next. The median drawdown can emerge even without memory, as it did for the simulated stocks in my example. (It's a bit like the fact that if I collect how often each number comes up in the lottery I get a binomial distribution, but the fact that a number has not come up very often so far does not mean that it tries to catch up going into the future.)

Now, the big question is of course: is GBM a reasonable assumption for stocks? Perhaps they do have memory since their prices are determined by intelligent agents who definitely do have memory. Certainly there seems to be evidence that odd things happen near round numbers for drawdowns for example. Trend following or momentum are also suggestive of intelligent agents using memory.

Personally I doubt that a 5% movement can be used to reliably predict a bigger move. However if you were to wait (say) six months and then partition your portfolio into those which are above 0% and those which are below 0% then even assuming a GBM model the stocks which are above 0% are more likely to perform better than the below 0% stocks going forwards. The reason being that after six months the volatility effects are smaller compared with any underlying trend.

(The advanced version of this partitioning is to divide the excess return (over a minimum acceptable return rather than 0%) by the volatility (i.e. compute Sharpe Ratio) and only keep those with the highest ratio. The very advanced version is to optimise the whole portfolio return divided by volatility, taking into account covariance between the stocks in the portfolio.)

But surely if a stock has been hitting yearly highs and then draws down 5% then your research suggests that more often than not it is likely to continue to drawdown towards -33%?

Conversely if a stock has been hitting yearly lows and gains by 5% your research suggests that it is more likely to continue to improve towards a 50% gain?

I think inadvertently my strategy harnesses the effect I'm describing.
Firstly all of my stocks are high quality.

I buy on breakouts, in effect a 5% gain. I'm looking to ride the wave up to 50% and well beyond.
I don't sell on a % drawdown but rather look for a very definite switch to a bearish perspective on a stock by stock basis.

Why bother hanging on to a stock that is definitely in a strong drawdown when it could languish and deteriorate to on average -33% off the highs whilst I could be investing in stocks that are gaining towards an average of a 50% gain?

It's a Naked Trader strategy really, buy winners that are breaking out and cut them quickly if they fail.

Your comments caused me to do a bit of googling to see what the "mean-variance" gang have to say about drawdowns. It turns out that the seminal paper is "Optimal Investment Strategies for Controlling Draw-downs" by Grossman and Zhou in 1992 but since then the ideas have been refined considerably.

The short version is that you choose a "maximum drawdown" you are prepared to accept and then you scale out of your position as the price falls rather than having a stop-loss which takes you out right at arguably the worst possible point. Where it gets tricky is scaling back in again and controlling the costs associated with constantly adjusting your position.

Cuthbert C

Redrichmond

In post no 10 above you say "...If you get stopped out at 20% , that means that stock has to go up at 22% + to just get your money back…."

I don't understand what you mean.

If you are stoped out at -20% then you lose 20%. You do not get that lost money back if the stock out of which you were stopped goes back up.

If you lose 20% and keep hold of the shares then it will take a 25% increase in their residual value to get you back to your starting point.

T

Ah good to see my thumbs down troll has woken up.
I've nick named it Donald!

Choosing where to put a trailing stop-loss to avoid being stopped out "unnecessarily"!

Unlock the rest of this article with a 14 day trial

About the Author

Nick Ray

28 comments

Unlock the metrics for all these stocks and more...

Unlock the metrics for all these stocks and more...