The Trouble with Averages

  • Posted by
  • on August 18th, 2011

“Better to be roughly right than precisely wrong”  John Maynard Keynes

I use that quote a lot…it makes a ton of sense to me in the complex adaptive system we know as the market.  Brian Shannon often uses the term “essence of trend”, another phrase I like to steal.  I thought I’d share a quirk in my analysis yesterday that had me looking the wrong way as the morning rally unfolded.

I’ve grown to rely less and less upon quick reads and more upon a robust collection of evidence in making my trading plans.  The “algorithm” if you will, has become indispensable to me and is becoming the culmination of my years observing markets.

That said, once markets open and start moving, it’s time to act not analyze, and I lean on a few simple reference points to point me towards the right side of the action.  The one I’ll discuss today is “% of Stocks Above 10 Day Moving Average”.  With the help of Vic, I know this number at all times and use both the 1)absolute level and 2)direction over the past 24 hours, to keep me from fighting trend.

Wednesday morning, the number was rocking, above 75% for the 1st time since July 22.  While not a level where I initiate buys, certainly indicative of strength.  As I always do, I studied the data as the morning trade slowed and was shocked to see widespread deterioration.  Tuesday had brought some doubt to my “bouncing from a low” case, but surely the overnight head fake lower and subsequent Wednesday morning rally took care of that, right?

My process is to put my 400 stocks into Excel and scan for “no way” shorts and “no way” longs by a total of 16 factors across time frames ranging from days to months.  I then take the candidates and scan them further by a couple of reward:risk measures.  This gives me some stocks that are probably not entries but were at least able to not fail basic tests of strength or weakness.  More importantly, I get a real sense for market potential by seeing the number of candidates on each side.

The numbers were awful.  Recent days had shown me feeble numbers both ways, better than the overwhelming shorts in late July/early August but not indicative of an imminent bull run.  As I said, Tuesday added some doubt as “short candidates” expanded to 35 from readings around 10 in recent days.  After Wednesday morning, with the market UP, I was getting 78?!  I checked my work 3 times vs. the usual once but still the same outcome…troubling structure.

It’s possible my universe is too heavy in the tech and cyclical names that weakened, but what kind of bull would emerge without those $QQQ types?  The point is, the use of one data point was very misleading due to its construction…my full analysis always contains multiple inputs and, in particular, multiple lookback periods.  In this case, those 10 day averages still included prices from the depths of early August; they’ll soon drop out and we’ll see that number was high by attrition, not power…big difference.

I don’t like averages for this very reason…I use median true range, not average(except when doing position sizing), because I don’t want an outlier dictating my analysis.  I cringe each time I hear someone say “the $SPX is trading at only 13X forward EPS”.  First of all, forward?  Good luck with that.  Second of all, the economy ebbs and flows…who cares about a 4 quarter data point even if it’s trailing?  I like what Shiller does with smoothed EPS; there’s a reason smart guys like Hussman and Easterling do the same.  If I were constructing a valuation model, I’d take the median EPS of the past 13, 21, and 34 quarters and multiply it by 4.  Not my arena, but I’m willing to bet the range of historical P/E multiples contracts and one could actually use an estimate of normalized earnings that isn’t degrees away from the real world.

The point is, any one indicator can be very misleading.  Why not at least measure it 3 ways to prevent overreliance on the starting date?  Smoothing mechanisms can be very helpful if we remember that we’re trying to assess the present to make money in the future; a model built on a faulty starting point is destined to let us down at some point.

The information in this blog post represents my own opinions and does not contain a recommendation for any particular security or investment. I or my affiliates may hold positions or other interests in securities mentioned in the Blog, please see my Disclaimer page for my full disclaimer.

blog comments powered by Disqus
Derek Hernquest Blog