MODEL UPDATE AS AT 10 OCT 2013:


INTRODUCTION
This is a quantitative actuarial trading model for the JSE TOP40, using purely seasonal factors to govern if it is invested in the market or not. There are no traditional technical metrics used (moving averages or market breadth) nor are any fundamental or econometric models deployed in the seasonality determinations.  The model is called STM for short.

STM deploys a cyclical multi-annual adaptive seasonality heuristic to determine if market conditions are favorable or not and moves into the SA TOP40 or cash depending on the result. For now, the specifics of the parameters used by the heuristic algorithm remain proprietary. 


STANDARD LONG-ONLY MODEL
The standard model which makes un-leveraged long-trades only, traded 143 times since 1960 (2.7 times per year) with trades ranging from 1 month to 6 months, but averaging 3 months duration. Over 81% of the trades resulted in a profit with the strategy delivering just over 50 times the performance of the buy-and-hold strategy (double the compound annual growth rate), with gains outpacing losses by a factor of 10-to-1.

The distribution of the 143 individual trade returns are shown below. We see a lot more green bars (positive gains) than red bars (losing trades) and the green bars are comfortably higher than the red bars to ensure algorithm positive expectancy of over 10-to-1 (for every losing percentage point, you claw back more than 10 on the winning trades.)

The STM SuperCycle Algorithm makes trading decisions on the last day of each calendar month. That way, you know exactly on what date you need to switch into the JSE or into cash. Just under 40% of the trades only last one month but just under 40% of trades last 2-4 months. Thus 20% of the trades are more than 4 months in duration. The distribution of trade durations is shown below:

Following the model since January 1960 with a starting capital of R1.00, your return versus the JSE buy and hold is depicted below, excluding transaction charges, dividends and interest earned whilst in cash (the model is in cash 40% of the time so the interest component is not insignificant.)

Not only does the model grow capital at more than double the annual compound growth rate of the buy and hold, it does this by only being exposed to stock market risk 60% of the time, meaning it has an even higher risk-adjusted return. This can be visually observed by the blue TRI of the model above which is not affected by large draw-downs of the bear markets in the red TRI. The actual rolling 6-month draw-downs are shown below:

You can see the STM Seasonal timing model is subject to far fewer gut-wrenching draw-downs, averaging 5% draw-downs versus the JSE's 15% The STM model completely avoids the bulk of the devastating bear markets that wipe out more than 30% of capital. This confirms research on a similar model we built for the U.S on RecessionALERT.com, that most bear markets have a seasonal risk element associated with them rather than solely unexpected external financial factors.

To test that out-performance is consistent through time and not isolated to one or two good decades, we have prepared the below yearly returns table for the longs-only STM strategy:

We see that STM out-performs the JSE consistently around 70-80% of the time on an annual basis (% Best column shows what % of months in the decade that STM outperformed the JSE.) On a decade-by-decade basis we see that STM returns are 2.7x, 5.4x, 2.6x, 3.0x and 2.1x that of the JSE respectively. These are outstanding results when one considers we have not taken interest earned when STM is in cash into account.

PREDICTIVE POWER OF SIGNAL
The STM Seasonality Signal can only ever take on 10 different values ranging from -1 (exceptionally bearish)  to +2 (very bullish) in steps of one-thirds. There is a direct correlation between average monthly historical JSE returns and the level of the STM signal since 1960 as shown below:
We can see that STM readings of -1, -0.67 and -0.34 are ideal shorting opportunities, whilst STM readings of 0 and 0.34 are best for retreating into cash. Readings of 0.67 and 1 are very respectable long opportunities whilst readings of 1.34, 1.67 and 2 are "Power Months" with high win rate percentages and excellent gain/loss ratios.

The percentage of time the STM signal spends in the various recommended positions mentioned above is shown below:


The STM signal is displayed daily on our web site for subscribers in a heads-up display similar to that shown below. The signal is the black line with yellow dots.

The long-only strategy that we have been describing thus far goes long the JSE whenever the STM signal (the black line with yellow dots) rises above 0.33. This is shown by green background shading. The STM signal is projected 6 months into the future so that you always know what trades are going to be made six months in advance.

THE SHORTING OPTION
When the STM signal is -0.33 or less then we have seen that you may opt to short the JSE with a high degree of confidence (62% win rate, over 3-to-1 gain/loss ratio.) This obviously boosts returns of the longs-only strategy considerably by raising the TRI from R30,973 to over R254,000, but also adds volatility to the equity returns and lowers the win rates somewhat. This should only be considered by experienced traders who have successfully adopted shorting strategies in the past.


The decade-by-decade performance table of the STM longs + shorts strategy is below:


THE LEVER-UP MODEL
When the STM signal rises above 1, we have seen that you can safely leverage up 2 times in the "Power Months" to make a considerable boost to returns without introducing too much volatility and keeping your win-rates unchanged. The distribution of trade returns for the original long-only strategy and the new leveraged-long strategy are shown below, where we can see that the introduction of leverage (the red line) does not increase loss sizes that much, but considerably increases the size of much of the universe of larger profitable trades:

As we can see from the above chart, the introduction of leverage in the Power Months increases the average win sizes much more than it increases the average losing sizes, which actually enhances the gain loss ratio of the overall strategy. So in fact, leverage reduces risk in the strategy as opposed to increasing it. This is due to the high confidence nature of the Power Months. The new statistics with the Power Months leverage is shown below:

Considering that the "Power Months" only number 152 of the total 393 months we are long (some 38% of the long months), the use of  leverage during these periods has a surprisingly astonishing effect on cumulative returns as shown below. 

You can achieve 2 x leverage in the Power Months two ways. First, you can merely double your trade exposure (double your bet size or risk) or secondly, you can deploy margin deposit on a futures instrument such as a TOP40 Index CFD or a SAFEX ALSI futures instrument. For example, if you normally position-size your JSE trades with a 2.5% risk (max loss of your capital) you could use 5% during the Power Months to "double-up".

The decade-by-decade performance of STM that goes long when the STM signal is >0.34 and deploys 2x leverage whenever the STM signal is above 1, is shown below:


THE SUPERCHARGED MODEL
The next question on your lips is undoubtedly what happens when we use longs and 2x leverage in Power Months and also short the market. The chart below shows the inclusion of the TRI of the strategy that goes long, leverages 2x in the Power Months and also shorts the market:


The decade-by-decade statistics of this hyperactive, aggressive seasonal strategy is below:


CLOSING REMARKS
Note that the suggested direction (long, cash or short) and leverage (1x or 2x) for any one month is always shown on the live Heads-Up display by the shaded green and red areas. Red shading of -1 means short, green shading of +1 means normal un-levered long positioning and green shading of +2 means two-times leverage suggested. No shading means cash is recommended. These are shown 6 months into the future at all times on the heads-up display. 

Whilst this strategy produces excellent risk-adjusted returns, you must bear in mind that with the execution of seasonal strategies, you need trading skills to be able to accept the imperfections  that come with seasonal models (which are essentially blunt instruments), and the many 1-2 month trades that pepper the strategy, and you also need the fortitude of a buy-and hold investor to sit tight in the market for up to 6 months at a stretch no matter what you hear and see about you. But as you can see from our research note these models have very robust  multi-decade performance returns provided you stick with them long enough for the positive expectancy inherent in them to come to the fore.

It is not wise to take readings from the Seasonal model and apply them literally to any other strategies you are running. For example a future 3 bearish months shown by STM should not mean you go and execute a wholesale liquidation of all your investments and other models you are currently following. This is just another arrow in your quiver and another model to follow in parallel with others in your strategy. For example, you may use 40% of your account for Quattro, 20% for Sincerity, 20% for LazyBoy and 20% for STM as an example, to get a nicely balanced mix of short, medium and long term trade-horizons across a diversified set of technical, econometric and seasonal algorithms. 
 
Make a Free Website with Yola.