The Golden Cross Trading System Lacks Intelligence

One of the main reasons I abandoned trend-following in the mid 1990s was the fact that I was not able to find a system for which the null hypothesis that it possessed no intelligence in pairing raw market returns with its signals could be rejected. Yet, many have claimed success trend-following the markets. But a fair assessement of the situation requires to also count those who got ruined by it and this is not easy. Survivorship bias gives the wrong impression, especially when it is used by the lucky to convince the unlucky that they, meaning the lucky ones, were not just lucky.

Below is a chart of the growth of 100K fully invested in a long/short symmetric golden/death cross (50-200 simple moving average crossover) system since inception of SPY to 05/02/2013:

SPY_1993-20130502

Charts created with AmiBroker – advanced charting and technical analysis software. http://www.amibroker.com/”

It may be seen that the system started running out of luck in 2010. Why then and not before? The short answer is that this was a random outcome. But first, let us take a look at the underwater equity curve:

SPY_1993-20130502_DD

Charts created with AmiBroker – advanced charting and technical analysis software. http://www.amibroker.com/”

The first warning that this system had no intelligence was issued in 1998 with a drawdown of about -26%. Anyone who manages a fund knows that this drawdown level is borderline disaster. Then, by mid 2012 the drawdown was about -38%, signalling that there is something fundamentally wrong with this system. Yet, many still use it in lieu of anything better, just waiting for disaster to strike.

This system since inception of SPY and as of yesterday has a CAR of 8.37%. This statistic looks good at first sight but the results of a Monte Carlo simulation indicate that there is no evidence against the null hypothesis that the golden cross system possesses no intelligence in pairing raw market returns with its trading signals:

SPY_1993-20130502_MC

A Mersene Twister algorithm (implemented in Amibroker) was used to generate random numbers that signal long and short positions with equal probability and are exited according to the golden-cross system exit rules. The mean CAR of 10,000 simulation runs was 8.205%. The p-value for a CAR of 8.37% was calculated equal to 0.3963. Thus, there is no evidence against the null hypothesis that this is a random system with no intelligence. It may have some, but there is no evidence for it in the context of this test.

Actually, I obtained a p-value with the same test of about 0.07%  back in 2000 for the same system. I thought it was too high then. I am looking for p-values less than 0.01 and if possible, less than 0.001 (1 in 10,000 for obtaining the result when the null is true).

Although 2010 and 2011 were disaster years, in 2012 this system  underperformed the market, it is currently gaining about 9% for the year. But this may be a random event. As far as I am concerned, only a gambler would rely on this system. However, opinions may vary on this subject and they are all respected when they come with analysis and numbers. There is no doubt in my mind that this system has no intelligence, it became popular because it survived during the 1990s due to the strong market uptrend. Its random nature emerges when volatility increases beyond what it can handle.

Disclosure: no relevant position at the time of this post and no plans to initiate any positions within the next 72 hours..

Charts created with AmiBroker – advanced charting and technical analysis software. http://www.amibroker.com/”

Disclaimer

This entry was posted in Trading Strategies, Trend following and tagged , , , . Bookmark the permalink.