Sunday, October 26, 2014

10/26/2014 The Purpose of the Full Model

                The “Mousetrap” is a method for merging technical and fundamental indicators into a single model.  The concept for each aspect of the model is mean reversion.  On the technical side I measure long term volume-breadth against price-strength in a sector or industry.  On the fundamental side I measure long term growth in cash flow, book value, and earnings against total return and total debt.

                If the debt is low and the price return is cheap compared to long term growth, we can find a relatively safe stock within an industry that is technically primed for a short term reversion toward the mean.

                At a point in which price begins to balance against both technical and fundamental indicators, the stock will become attractive to momentum traders, and I will sell just as demand begins to catch up to supply.

                In this manner I buy stocks when no one wants them, and sell them when everyone wants them.

                The returns for this method are better than 20% annualized, and I have been enjoying this return rate in live trading since 05/31/2011, when I launched the model.

                In April of 2012 I began publishing the trades at – advertising each trade the night before it was to occur.  There is no ambiguity or hindsight bias in these trades.  I post them before they happen, and publish the running returns, both good and bad.

                During this period I have experimented with some of the fundamental features, and tightened up the technical parameters, but the general trading concept has been the same: buy stocks in low demand sectors and sell when demand begins to pick up.  I sell them well before the top, and in most cases only enjoy the first third of a price move.  But the first third of a price move allows me to unload them with ease (and to hold for longer, if scalability ever required it).

                Measurements are long term in order to avoid the effect of High Frequency Trading (HFT) algorithms on the model.  With a long enough gauge, the HFT effect becomes background noise, and the model will not degrade in the future.  Humans tend to think in terms of weeks and months, logarithmically discounting longer time frames.  HFTs are designed to take advantage of human behavior, and in order to compete with each other they work in ever shortening time frames.  Their goal is a 60% return.

                My goal is a consistent 20% return, leaving me out of their time frames and target return rates.  I don’t compete with them, but I’m happy to pick up the debris they shake loose in their activity.

                Live trading has been consistent with back-tested results:

By the numbers:

Sector Model


                That is, during the 12/23/1998 to present period of combined back-test and live trading, the S&P has given an annualized rate of return of 3.14%, while the Sector Model has returned 21.03%.

                No back-test is possible for the Full Model, but the annualized return rate for the live trade period of 5/31/2011 to the present has produced 21.89% annualized returns, which is consistent with the Sector Model.

                Note that the primary goal of the Full Model is not out-performance per se, but rather scalability.  Containing only 10 stocks with an average holding period of 70 days allows me as a small investor to perform one trade a week.  As the portfolio increases in size, I can increase to as many as 10 trades a day, with 500 logical positions, without even needing to resort to intraday measures. 

                In other words, the Sector Model is fine for a small portfolio, but the Full Model can function with a nearly unlimited portfolio base.

No comments:

Post a Comment