Recently on the bus I’ve been reading Peter Bernstein’s “Against the Gods, the Remarkable Story of Risk”. Its’ basically the history of how modern risk management evolved. One of the most powerful discoveries and concepts was reversion to the mean – the statistical tendency for a measured phenomenon to cluster around an average over time; more so the larger the sample size and the period of time, and the more random and non-correlated the samples. If you haven’t noticed already, nature is highly cyclical, and even the Indian rishi’s understood that matter itself is vibration down at the atomic and subatomic level. So what can we predict from that about the stock market? Because human behaviour also exhibits a high degree of cyclicality, reversion to the mean is a powerful unseen force that brings stock prices back toward that mean when they overshoot one way or another. Just look at a sample of say even 10 stock charts and that cyclicality is immediately apparent.
Because stock prices move in response to people’s reaction to new information constantly entering the markets, generally randomly, and that information comes from many influences, the stock market is a non-linear system. By definition when you have 3 or more variables in a system, you cannot predict a specific outcome for a specific event in that system – for a example a weather temperature or a stock price. However reversion to the mean is definitely at play over many samples – we know that summer and winter come and go in a predictable pattern because of a direct correlation to the earth’s (non-random) orbit. We know that storms come and go; however there are random influences at play here, so since we know with a high degree of certainty that storms will happen, but we cannot predict very far ahead when, where, and how strong those storms might be. Same thing with stock prices.
We know that stock prices oscillate, and that reversion to the mean is much stronger a force the larger the sample size and the longer the time frame. An index such as the S&P500, or a mutual fund or ETF with many component issues represent a large sample, and would exhibit a stronger, more reliable reversion behaviour. Within a given price range over a long period of time, we know that prices will eventually return to a given value in the future, especially to values clustered around the mean of all prices in the time period measured. This does not mean that there are no outliers – you can have a stock that collapses to zero (as many airline stock investors may have unhappily experienced), and you can have extraordinarily successful stocks such as Microsoft and Apple that grow and split and split and make fortunes for those prescient (or oblivious) enough to simply hold. Can you predict these outliers? Not really, the odds are much lower at the extremes. To the extent that you understand and can anticipate certain general economic trends, and how these over time impact stock prices, you can still make intelligent decisions about how and where to deploy capital – but choice of vehicles and timing of when to enter and when to exit can still be problematic, more so the shorter the time frame of the investment or trade.
So let’s say you wish to forego divination, and instead try to simply align with natural forces and try to quasi-mechanically work with that in the stock market. You know you can’t predict when and where a stock, ETF, or index will be in the future, but you know that these fluctuate over time and have a powerful urge to return to the mean. Buy low and sell high stands out as an obvious approach. The question then becomes how high and how low? And how much capital to deploy and extract at the lows and highs when they happen. Also, there are short, intermediate and long-term cycles at play, you have to look at the time frame you are working with and adjust your risks (capital allocation) accordingly. The difference between target low and target high prices will represent a range, and the approximate profitability of a given trade. The narrower the range, the more transactions, the wider the range, the fewer transactions (and lower transaction costs). The ideal scenario is for prices to move up and down more quickly over a wider range – as profitability is influenced not only by the average range of transactions but also the frequency of transactions. The “choppier” the market, the better. Long sweeping trends, especially to the downside are less than ideal.
So taking all of the above into consideration, I am still trading a home-grown system called the Extension Reversal Trading System (ERTS), but after continued exhaustive back-testing have refined it to trade weekly rather than daily setups. The main reason is that the returns improve because the range is increased between buys and sells (price has time to move further distances in the short-term cycles). An added bonus is reduced transaction costs because of the significantly reduced number of trades.
Except for a few blue-chip stocks of large producers of important commodities (Cenovus = crude oil, Potash Corp = potash, Teck Corp = base metals, Silver Wheaton = silver, Cameco = uranium), the rest of the trading vehicles are commodity and sector ETFs. Commodities are cyclical and do not go to zero (because by definition commodities are in demand to greater or lesser degree to support and sustain economies/populations as a whole). Reversion to the mean will most definitely a key force playing on commodity markets (always economically necessary) and diversified sector and index ETFs (large samples) and therefore improve the likelihood that a strategy based on reversion to the mean will ultimately be profitable.
To improve the short-term range and profitability of the system, the ETFs for the most part are either 2 or 3-times leveraged. Though they do erode over time from re-balancing, the benefits from a still relatively fast turnover of trades outweighs the mathematical erosion risk.
Stocks and markets can go into long bear periods – three recent examples are natural gas, Cameco, and Teck Corp as commodities fell out of favour in mid-2011. Therefore it is important to limit the risk of loss on any single vehicle. Each vehicle therefore is allowed a maximum of 4 tranches (4 separate simultaneous trades) during a declining/accumulation period. Another tranche can only be bought again after the next sell. This protects against ruinous averaging. However, to leverage the program a little further and reduce average cash holdings, I allow “oversubscription” of the tranches in the aggregate. For example, if there are 6 ETFs in an account, you would make your trade size Account Equity / (6 x 4 = 24 tranches). However, recognizing that not all markets will draw down to 4 tranches at the same time, I reduce the divisor by n-1 vehicles – in this case 5. So the capital for any given trade will be Account Equity / 19 in this example.
After further testing, I decided to add back 2x crude oil and 2x silver ETFs, and also a new Direxion 3x Agribusiness ETF. This moves provides a little more diversification and adds some more unit volatility to the overall portfolio. Right now there is not much happening because most things have sold out into the current stock market rally. Oppositely correlated to that has been the behaviour of natural gas (in a steep, persistent bear market because of shale-gas induced oversupply) and precious metals – the main long positions at the moment.
Cheers,
Allocator
