Let's take a look at seasonal investing.  By this, we mean breaking the year into different periods and trying to discern trends that might repeat in the future.

Our interest in the subject of seasonality, predictably, started with curiosity regarding the well-publicized January effect.  We decided to scour historical data, and see for ourselves if the phenomenon of December's losing stocks gaining in January is bunk or not.  The answer:  it's not.  Since Jan 1994, you could have substantially outperformed the market in every year except 1999 and 2002 by going with stocks that were beaten-up in December, and you wouldn't have lost your butt in 1999 or 2002 either.

You'd measure the "beaten-uppedness" of a stock by taking a moving average.  The question arises, exactly what sort of moving average would work best in winnowing out stocks that might perform well in January?  Five day?  Twenty day?  100 days?  As long as you've got a mound of indicators to play with, you might as well test volume data, volatility data, the actual stock prices...whatever you fancy...as predictors of a big gain in January.

We've done that.  While beaten-up stocks perform nicely in January, there's another simple indicator that does quite a bit better than a moving average in predicting January gains.  You'd have beaten the market every year since 1994 (the beginning of our test) without exception by following the indicator, assuming you purchased a large enough basket of stocks.

Given that you've hashed through January data and found some statistically significant predictors of future gains, you might then begin to wonder if there are other periods of the year where a statistical analysis would show that there are indicators that perform well above chance in predicting future gains or losses.

Again, we've done that.  And, of course, we've found some very nice indicators of future gains and losses, depending on the time of year.  The January effect is dramatic, but there are significant trends in almost every month and quarter of the year.  We've even come up with some strong trends that manifest over half-month periods.

So you've identified the best indicators of gains and losses in various periods during the year from a historical perspective.  So what?  If we say that December losses are a great predictor of January gains from the standpoint of history, mightn't it actually be the case that this trend is more, not less, likely to reverse in the coming January?  After all, the market has a way of obliterating these sorts of inefficiencies.

What's needed is some backtesting of the data.  Of course, if you find that high yearlong gains is a good predictor of further gains in June based on a compilation of data from 1993 to 2002, it would be "cheating" to apply this standard to data prior to 2003...you'd be using results from the future to predict the past.  What you could do, though, is find the best indicators of gains and losses in 1993 only, and then apply those indicators to 1994.  Or you could "compound" data from 1993 to 1995, and use that data to make predictions for 1996.

Of course, we've done that.  We've run a number of different tests on different seasonal strategies and they've all beaten the market in a big way.  In one test, we beat the performance of the Russell 3000 Index by a factor of 4.7 (that's 370%, or a 700+% gain overall) from May of 1994 to March of 2003  by simply mirroring the best monthly indicators derived from the previous year.  In all our tests of long strategies (as opposed to short strategies), we've beaten the indexes handily by mirroring past seasonal trends.

Bear in mind that we're looking at groups of 100 or more stocks out of several thousand...it's essentially impossible to "get lucky" and accidentally choose a couple stocks that make an otherwise mediocre portfolio look wonderful.

The nice thing about screening the market via these seasonal indicators is that it doesn't necessarily preclude further screening via complementary indicators that you might be using already...fundamental, technical, chart-based, whatever.  There's really no telling what sort of percentage gains might be possible by combining a long list of seasonally "viable" stocks with your own investment methods.

The Site...

...has three functions.  If, perchance, you feel that this whole philosophy of looking at historical seasonal trends with future gains in mind is bunk (something we'd strongly dispute, given the backtesting we've conducted), or something akin to a search for the holy grail, you still might find a good deal of value in the site via two other offerings on the site:

1)  A page where the last market session is analyzed in terms of terms of the indicators that best predicted yesterday's gains.  Let's assume that today is Wednesday and the market just closed.  We ask which of Tuesday's indicators best predicted Wednesday's gains and losses and present the data in a relatively easy-to-understand format.  There are many explanatory notes in the site, but the main thing to remember is that the gains and losses that you see on the page are "annualized"...since there are about 250 days in a year, a 1000% annualized gain is the same as a 4% daily gain.

The analysis here is purely mechanical.  The computer does the work.  Only rarely will there be subjective commentary.  Thus, instead of "fears of terrorism dampened the spirit on Wall Street today", you get raw data that tells you that the best indicator of gains in Wednesday's market was a low 20 day moving average in Tuesday's session, or increasing short term volume, or a high "close-low" differential, or whatever.  You even get the statistical significance of the best indicators...are they flukes, or should I pay extra attention to them?  Industry data is also considered.

2)  A big, ugly text file that contains data on about 2300 stocks from the last market session.  You can download it, import it into Excel (or another spreadsheet program) and sort and screen the data to your heart's desire.  It's NOT intended as a stock quote service, however, so we will not offer actual high, low, or volume data...only percentage differences versus prior sessions.  If you think highly volatile stocks will gain tomorrow, or throughout the next quarter, sort the stocks by volatility.  The only prerequisite here is a tad of skill with a spreadsheet.

3)  And, of course, there's the historical data.   

Seasonal Investments Home

Copyright © 2010 MarketSynopsis.com. All rights reserved.