Skip to content

Markets within the Market

April 5, 2010

In the last post, I was trying to highlight that markets and stocks can have their own idiosyncratic behavior and that it is possible to isolate them using different variables. The ability to do this permits a wide range of opportunities: 1) finding stocks/markets best suited to your trading style 2) allowing for more focused analysis by controlling for these effects 3) alllowing us to create internal breadth measures that are far more accurate.

One variable that stands out in our research is the propensity of a stock or market to trend. This variable is a core component of the Livermore Active Issues Index and is called the “LTR” which stands for Livermore Trend Rank. Note that this is not a relative strength based score, but rather a historical measure of relative “trendiness” over a longer period of time. We then combine relative strength with the LTR and volume metrics to create the final index score.  While the technique used to create the LTR is proprietary, it does not involve anything that is statistically complex. The LTR is accurate and highly robust and has show to work across 35 different markets and on the stocks from all major stock indices. A high LTR score indicates that a market or stock is expected to trend, a low LTR score indicates that a market or stock is expected to mean-revert. LTR scores are calculated monthly and do not turnover frequently. This means that behavior is not a temporary volatility effect but rather a more intermediate or long-term behavioural tendency for a stock or market to trend/mean-revert.

Below we show the performance of a standard mean-reversion/reverse  daily follow through strategy (long down days, short up days) using the   LTR factor to separate the top and bottom 50% of the Nasdaq 100 index.

As you can clearly see, there is a market within the market–half of the stocks have a tendency to trend while the other half have a very strong tendency to mean-revert. This has enormous implications—especially given that the index itself the QQQQ has a tendency to mean-revert. It means that we can more accurately predict the direction of the QQQQ or SPY or any other ETF/index by controlling for the effects of this variable. It also means that you should focus different strategies such as trend or mean-reversion on different stocks of the index rather than trying to create a universal system.

10 Comments leave one →
  1. Stefan permalink
    April 5, 2010 3:30 am

    This was super interesting. I have also developed a similar indicator with your LTR, for mutual funds. It has to do with measuring trend persistency, more than propensity for mean reversion. I was inspired by the Gil Blake Market Wizards interview. Anyway, I have a stable of measurements and indicators, like relative strength, trend persistency, alpha, etc. The problem is, ho wcan you combine them to get one unique score? I could normalize them all, between 0-100 and add them with equal weight. But, really, any hint of idea how to make one indicator from a bunch indicating different characteristics would be a great, great help.

    • david varadi permalink*
      April 6, 2010 2:28 am

      hi stefan…..big fan of that book and the gil blake interview was fascinating. combining disparate scores is
      a very complicated question to answer, multiple regression is one method and should be used with only
      3 scores that have the highest power and the lowest correlation. other nonlinear and statistical methods are idea for
      more than 3 scores and very complicated relationships.


  2. jim permalink
    April 5, 2010 11:09 am

    There are indicators like ADX which measure the level of trendiness. The problem is always the same though. The future is always unknown so something that was trendy last month can chop this month and vice versa.

    • david varadi permalink*
      April 6, 2010 2:24 am

      hi jim, i didn’t find the ADX very useful either. this is a long-term indicator and the ranks are rebalanced monthly. the test was out of sample.


      • jim permalink
        April 6, 2010 2:29 pm

        Yes I see. But the ADX is a measure of trendiness is it not? The time portion can easily be adjusted to a very long time frame, and be rebalanced monthly.
        I guess it just depends on your definition of trendiness.

  3. April 5, 2010 1:19 pm

    just caught up with your last 3 posts David – they were very interesting – thanks!
    One thing I am looking into (primarily for my Trend Following system) is very similar: some “trendiness ranking” to use as a filter to select instruments to trade (ie similar to a momentum filter portfolio selection). One thing I have started looking at is calculating the rolling Hurst coefficient (ie fractal dimension) but this has not proved too useful so far.
    Any hints or bits of info you published on what the LTR is based on (I understand it’s proprietary, so no probs if you prefer to keep it private..)

    • david varadi permalink*
      April 6, 2010 2:21 am

      hi jez, thanks very much. the LTR is proprietary so I wish I could share more but I can’t at this point. I didn’t find the hurst to be valuable either, so I basically created my own measurement using common sense—strangely enough this worked the best. I find that many of the serious mathematical measurements have too many built in assumptions from the phyiscal science world to be directly applicable to financial markets. As such, I often build my own (with help :o))


  4. April 5, 2010 1:22 pm

    @jim, the problem with the ADX is that it only looks at too recent history. I think what David refers to with his LTR indicator is how “trendy” a market has been over the previous n-months/years periods (ie something a bit more structural, than just the strength of the trend right now – which is more what the ADX does)

  5. April 6, 2010 6:47 am

    Fascinating study, David…thanks for sharing. I have found great improvement from simply forcing myself to consider the makeup of the players within the stock I’m considering. For now, still anecdotal but helps me avoid applying trend following techniques to obvious reverters and vice versa. I look forward to reading your code, and applying it 🙂


  1. Getting to Know Your Stocks « Derek Hernquist

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: