• 20DMF and the Long-Term picture

    I have received a few comments stating that the 20DMF has underperformed the markets in 2013 because of its failed short signals and that I should "adapt it." This is indeed correct and I have noticed since many months that the 20DMF have had a high failure rate on its short signals.

    However, instead of adapting the 20DMF to the QE environment, I have myself stressed the need to move to shorter time frames and into stocks selection. This strategy has worked well, but I now understand from the survey that not many people can act on shorter time frames, mostly because of "real-life" activities during market hours.

    I therefore decided to study the use of longer time frames on the 20DMF oscillator.
    The main issue lies on the higher level of short trade failures. Indeed, between Match 2009 and the end of 2011, the win ratio for short signals was 2/3. However, since January 2012, this ratio fell to 50%.

    On the 20DMF, short signals are issued by the upper panel, while long signals are issued by the OB/OS lower panel. Since the issue comes from short signals, the idea is therefore to keep long signals being issued by the current 20D OB/OS pattern, while increasing the upper panel to 50 days (i.e. a 50DMF.)

    We can see below the comparison between the 20DMF and the 50DMF (for short signals only. All long signals issued by the 20D OB/OS.) Note that the 50DMF is computed only from early 2010.
    We can see that both the 20DMF and the 50DMF lead to comparable returns and that both outperformed the S&P500.



    However, if we zoom in and only consider 2013, with January 1rst as the starting point, then the picture is very different: The 50DMF performed much better than the 20DMF, even though both underperformed the S&P500. But, we can see that there was a wrong short signal pointed by the green arrows. To avoid such signal, why not increasing the time frame again and use 75 days.



    Miracle! A longer time frame greatly helps (Long signals are still issues by the 20D OB/OS).



    If we look below at the different shapes of the MF, we can see that higher time-frames flatten the curves and hence issue fewer short signals.





    The major problem is that higher time-frames bring such a lag in the short signals that we short almost always at the bottom. We are saved from a catastrophic situation because the 20D OB/OS quickly cancels the short by a buy signal.


    This means that it is almost even better to stay in cash during a short signal and buy only when a buy signal is issued. On a 75D time frame, the strategy is a buy only strategy The long term results is about at par with the S&P500.



    If we look at the comparative performance table for long and short trades (from early 2010,) we can see that the 50DMF let the long trades run while issuing fewer short signals. However, the 75DMF does not issue enough short signals and hence is responsible for large draw-downs on the long trades.



    Conclusions

    Although I would dismiss the use of the 75DMF, the 50DMF could be interesting for traders who prefer long-term time frames and do not like shorting the market. I am not sure that the 50DMF would have performed nicely during the 2008 crash, but I believe that it would have kept users out of major draw-downs.

    I will continue the work on this 50DMF and if the results are satisfactory, it will become a main indicator for longer time-frame analysis.