Testing Different Trend Following Stops

A number of students on my trend following course have mentioned a couple of different trailing stop ideas, so I thought I would put them to the test and see how they perform. Following is an exploration into two different methods; an idea from Nick Radge and a chandelier stop from Charles Le Beau.

Note: the full video tutorial, code and back-test results are included in the course itself.

Nick Radge Stop

So, a couple of students mentioned an idea from Nick Radge of The Chartist, where the trailing stop stays at 40% during a bull market and then tightens to 10% during a bear market.

I don’t have access to Nick’s system but I’m told Nick bases this on the broader S&P 500 Index. When the 10 week MA for the S&P 500 is trending up, the stop loss for a stock stays at 40% and when the 10 week MA is trending down, the stop loss changes to 10%.

So what I’m going to do first is run the original system and then I’m going to change the trailing stop slightly and see the effects.

So running the original system between 1/1/1998 and 1/1/2012 returns 20.01% with a drawdown of 28%. This is with a fixed percentage trailing stop of 35%.

Original system
CAR: 20.01%
MDD: 28%

Now, I’m going to see what happens if we change the trailing stop to 10% whenever the S&P 500 drops below the 90 day moving average.

And this is easy to code up in Amibroker.

All we need to do is add in an extra line, so this says that when the S&P is in bullmarket mode (above the 90 day MA), we will use a 35% trailing stop and when the market is not in bull market mode we will use a 10% stop.

Trailpercent = IIF(bullmarket, 35, 10);

We also need to make sure our applystop line has this  ‘true’ parameter added at the end. Setting this last value to true means that the stop loss won’t be fixed and it can vary throughout the duration of the trade.

ApplyStop(stopTypeTrailing, stopModePercent, trailpercent, True, True );

So now we can run this variable stop loss idea and see the results.

Variable stop 1
CAR: 12.77%
MDD: 26%

Unfortunately, making the stop loss tighter when the S&P 500 drops below its 90 day MA didn’t worked out too well.

The drawdown has dropped but our compounded annual return has fallen to just 12.77%.

Next, I’m going to run a closer version to Nick Radge’s idea.

So in this example, I’ll use the 50 day moving average in place of the 10 week moving average. So when the 50 day MA is trending higher, the stop loss will be 40% and when it’s trending lower the stop loss will be 10%.

So as you can see, once again the results are no better. Annual return has dropped to 7.37% and the drawdown is now at 30%.

Variable stop 2
CAR: 7.37%
MDD: 30%

So does this mean this is a useless idea?

Well no not at all, and I still highly recommend Nick’s strategies.

It’s just that this type of stop clearly works better with Nick’s rules and not the rules that I’ve developed. I don’t know what the full rules are to Nick’s system or the settings so I can’t say.

Chandelier Exit

So now I’m going to go back to the original rules and I’m going to run the system with a Chandelier stop, which is based on the ATR (Average True Range) of a stock.

The basic idea is that when a stock is volatile the average true range is wider so the stop loss should be looser, and when the stock is not so volatile the stop loss can be tighter.

Typically, you take the ATR value (often 14) and times that by a multiplier to get the number of points with which to place your stop away.

So we can set this up in Amibroker with the following line:

ApplyStop(stopTypeTrailing, stopModePoint, 3*ATR(14), True, True );

We are now saying that our trailing stop will be held 3*ATR(14) points away from the recent high.

Next, I optimised the multiplier to see which chandelier stop works best during the in-sample period.

From the optimisation, the best multiplier was 8 x ATR 14 which produced an annual return of 21.09% so you can say that it’s actually worked quite well.

But even though we have optimised this result, when you look at CAR/MDD this still isn’t as good as the original system.

Here are the full results for the different types of stops tested:

Obviously, we could test some more variables but I’m going to stop there. I am not a big fan of the chandelier stop when trading stocks, mainly because I don’t think you can predict stock volatility very easily. A fixed trailing stop is also easier to manage and maintain.

For now, the original system rules work just fine and with fewer moving parts.