Trend following and global stock market health.

The last 9-12 month would have been a good environment for trend following. I missed it as my trading strategies are heavily focused on mean reversion.

This post is about traditional trend following strategies and how to outperform those by looking at the global stock market health for timing entries and exits.

.

Background

So let’s start by looking at traditional trend following strategies first

• Buy and hold: the baseline to compare initial results
• Golden Cross: MA50 is crossing above MA200 and exit when MA50 < MA200
• Tripple MA: enter when MA5 and MA10 above MA20, exit when situation changes
• ROC200: Enter when ROC200>0 and exit when ROC200<0

My Dax data goes back as much as 1990, so my first day of trading is 01.01.91. All trades are done close/close without commission and slippage. Be aware that DAX is a performance index, meaning all dividends are reinvested.

Trend following in its purest form

.

Buy&hold performs almost as good as any of the strategies. However one would need to take some severe punches, e.g. up to a 72.7% draw down. Among the strategies chosen there isn’t a clear winner as you can see. In average all return about 8% in annual return in a risk adjusted base it’s about around 13.5 The main difference is in the amount of trades a given strategies needs to make the returns. So far there aren’t any big news.

.

.

Let’s take a different twist to trend following

Let me first  explain my thoughts: I’m a firm believer that the global stock markets have become increasingly correlated:

” It’s all ONE trade “

If so, can we measure the healthiness of the world’s stock markets in order to time a trend following strategy? For simplicity reasons I’m not looking at every single stock market in the world: I’m looking at the most important ones and the ones I’ve got the data for. So I decided to look at the following indices

Asia

• Japan: Nikkei
• Australia: ASX

Europe

• Euro Stoxx 50

US

• SP500
• NASDAQ 100

For the entire test I did use the official cash index data (no ETFs). Unfortunately I don’t have historical data for emerging markets going back as much as 1990. Please let me know if you have! The simplest ( and commonly accepted) way to measure if a given index is in a bull market or not is the 200 days moving average. For simplicity we will measure this with the Rate of change (200) . A ROC(200) greater is positive and bellow 0 is negative. ROC(200) of the vehicle we trade (DAX) isn’t even considered, just the ROC(200) of the “other” indices.

• Entry: Enter when 4 (or more) out of the 5 indices a rate of change (200) > 0 have
• Exit: When 3 (or less) have a rate of change (200) < 0.

The result looks pretty good and there is very little to none curve fitting in it. All metrics are better compared to buy + hold as well as traditional trend following strategies. Of course many of the typical mean reversion strategies perform way better, but’s not the point. The point is to diversify into a set of VERY different strategies and this looks like a promising idea. Actually i wanted to finish the post at this point in time, but i couldn’t resist to continue my research and sharing this with you here and now.

..

.

.

Version 2: Improving risk adjusted returns

Looking at the metrics of version 1 (v1) you might have noticed the severe draw down of 30.7%. This is still better than buy&hold or any of the other traditional trend following methods, but severe enough to continue my research. In order to address the draw-down i decided to add a second shorter term filter: rate of change (50) for a faster entry / exit.  Now the system is entering the trade when the majority of indices have a ROC(200)>0 and/or ROC(50)>0. The exit takes ROC(200) as well as ROC(50) into consideration. If you want to have the exact details you can check out the AFL – code bellow.

Adding ROC(50) helps to improve ALL risk adjusted returns, of course absolute returns (CAR%) take a hit. However i would prefer to trade V2.

.

.

.

.

Final thoughts

With this post i wanted to present a very simple idea to improve trend following results on a risk adjusted as well as absolute base. Here are some suggestions how to further improve the strategy:

• Add a short term mean reversion filter to delay entry in order to buy into short term weakness and sell into strength aka RSI(2).
• Less binary view on when to go long.
• More indices: Data is a big issue testing trend following strategies for non US markets. I wish i could have added more indices, e.g. Hang-Seng and Bovespa or EEM as a proxy for all emerging markets.

.

Amibroker AFL Code

#include <Cbt_statistic.afl>;
SetPositionSize( 100, spsPercentOfEquity );
idx1  = IIf(ROC(Foreign("$NK","Close"),200)>0,2,0);  //Nikkei
idx2  = IIf(ROC(Foreign("$SX","Close"),200)>0,2,0);  //Euro Stoxx 50
idx3  = IIf(ROC(Foreign("$NQ","Close"),200)>0,2,0);  //Nasdaq 100
idx4  = IIf(ROC(Foreign("$XI","Close"),200)>0,2,0);  //Australia
idx5  = IIf(ROC(Foreign("$SP","Close"),200)>0,2,0);  //SP500


idx1b  = IIf(ROC(Foreign("$NK","Close"),50)>0,1,0);  //Nikkei
idx2b  = IIf(ROC(Foreign("$SX","Close"),50)>0,1,0);  //Euro Stoxx 50
idx3b  = IIf(ROC(Foreign("$NQ","Close"),50)>0,1,0);  //Nasdaq 100
idx4b  = IIf(ROC(Foreign("$XI","Close"),50)>0,1,0);  //Australia
idx5b  = IIf(ROC(Foreign("$SP","Close"),50)>0,1,0);  //SP500


IdxSum= idx1  + idx2 + idx3 + idx4 + idx5 + idx1b  + idx2b + idx3b + idx4b + idx5b;
Buy   = idxsum >= 13 AND Year()>1990;
Sell  = idxsum < 11;