Skip to content

De-Trending Indicators

January 11, 2011

I would like to apologize to readers for my prolonged hiatus from writing educational material. I recently accepted the role of Chief Investment Officer at an asset management firm that I will formally introduce in the near future. The firm is based in Los Angeles, California. We have spent  several months conducting internal research and development to support the firm strategy.

One of the common concepts in testing trading systems is to observe how systems perform on de-trended market data. That is, we would like to see how the performance of a system has benefited from the prevailing trend. If performance is unduly influenced by the long-term trend then it is possible that it does not in fact provide a significant trading edge. What CSS prefers to do to achieve this goal  is to either: 1) normalize the data for the current/historical environment or 2) create de-trended indicators. The latter method is the subject of this article.

The logic underlying de-trending the indicators versus de-trending the data from a performance measurement standpoint is straightforward: why not remove the influence of the trend on an oscillator to increase the profitability of its entries/exits versus merely using it as a tool for evaluation? Unbounded oscillators such as the DVU or the RSI are especially susceptible to being influenced by the current trend. Most traders vividly remember many examples in 2009 where RSI readings stayed above 90 for more than 5- 10 days at a time. This did not imply that the market was due for a correction, but in fact was just an indication of how strong momentum was during this period. In the examples below we have removed the 1-year trend from both indicators by subtracting out the net difference between the average values from center point (either zero in the case of DVU or 50 in the case of RSI). We used the 1-year average of the indicator values in this case, but you can use different lag periods depending upon which one best captures the dominant trend. As you can see, the de-trended indicators outperform the standard indicators in terms of both absolute and risk-adjusted returns. Small changes often make big differences– and big differences add up!

17 Comments leave one →
  1. January 11, 2011 3:39 am

    Congratulations for accepting the CIO role! All the best,


  2. January 11, 2011 9:53 am

    Congrats! Somewhat selfish question – will you continue to support DVindicators and update the CSS blog? Your work has been a great spark to furthering my own research efforts. Even if you don’t I’ve got a lot to digest…

    • david varadi permalink*
      January 12, 2011 3:00 am

      thanks JR, the answer is YES! I will continue to support both efforts–in fact more announcements on that to follow. thanks for the kind words.

  3. January 11, 2011 3:15 pm

    I’ve worked on detrending data for quite some time for my systems. But, I haven’t found the results from such manipulations entirely promising thus it is good to see some small performance improvements from such transformations.

    It might be interesting if you did a series on time-series analysis (for laymen) with a focus on dealing with the special statistical properties of markets such as auto-correlation, heteroskedasticy, detrending, and stationarity.

    • david varadi permalink*
      January 12, 2011 3:02 am

      hi curtis, the results on de-trending data are certainly disappointing relative to their promised benefits– i agree. that is why i turned to the indicators themselves which seems to perform better. I also agree that those subjects are worthwhile covering in a future post. thanks for the good comments.

  4. Larry P. permalink
    January 11, 2011 5:35 pm

    Congratulations on the new position! They are fortunate to have you.

    I too am wondering if you will continue with the blog as I have really enjoyed reading your posts. Many ideas to explore and thoughtful suggestions. Thanks for sharing your work and best of luck.

    • david varadi permalink*
      January 12, 2011 3:03 am

      Hi Larry, thanks very much–and appreciate that. As I mentioned above to JR I will continue to do so for the forseeable future.

  5. January 11, 2011 9:28 pm

    Agreed, they are very lucky to have you DV. My best, Jeff

    • david varadi permalink*
      January 12, 2011 3:05 am

      Hi Jeff, thanks very much!

  6. Zack Sullivan permalink
    January 12, 2011 5:00 pm

    I’ve been mulling over this post, trying to determine the difference between the non-parametric ranking of, e.g. DV2-bounded vs. removing the median or average of the DV2 unbounded. It seems the only real difference is the bounded version compresses extreme events into the edges of the scale (near 0.0 or 1.0), but otherwise should give a fairly similar answer.
    Do you see this type of detrending as adding more information than ranking over the same period, or do you think it just helps to keep a linear perspective some times?

    In either case, congratulations. Your posts have always been thought provoking. I hope they continue.

  7. January 13, 2011 5:53 am

    Congrats on the new gig David. Look fwd to hearing more! michael

    • david varadi permalink*
      January 16, 2011 2:42 am

      hi, many thanks michael–indeed I will share more at the appropriate time. I do intend to continue blogging and being more involved in the coming months now that we are set up.

  8. January 13, 2011 7:39 am

    great news, David…congrats! glad to hear you’ll still be provoking thought, I enjoy your insights immensely!

  9. January 17, 2011 12:25 pm

    Congratulations on the new venture David! Also glad to hear that you will keep publishing the quality work for us to enjoy!


  1. Tweets that mention De-Trending Indicators « CSS Analytics --
  2. Gradient Adaptive Dual-RSI Method

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: