MUST READS --> | JSE Seasonality | SUPERModel Ver 1.0 | Use Cyclical Stocks to treble returns |
How to use the PowerStocks Seasonality Actuarial Table & Chart (SATC) |


You will recall from one of our very first research pieces  "JSE Seasonality"  that we identified very definite periods during the year when the JSE traditionally performed better than usual.

Whilst over time, it was possible to make good gains by capitalising on this seasonality, you will recall that we noted some issues such as sitting on the sidelines when the market was running hard, or getting into the market when it was on its way down and more importantly, an uncomfortable level of volatility (up one trade down the next). This is to be expected since this strategy merely seeks out periods that statistically have shown superior returns and takes no note of any environmental or market conditions. In the backtests on seasonality you saw how the strategy continued trading through the great 2008 crash, since it was assuming "August is a great month, let me get in the market!" and "Oct is historically a reasonably OK month - let me get into the market!". Now over extended periods of time this strategy is designed to prevail (as it has the odds stacked with it) but sometimes things happen in the markets and economy and probabilities go out the window!

After the completion of our SUPERModel Composite Market timing system, we attempted to configure seasonality into the model by investing in the "strong months" in addition to those periods the SUPERModel was flashing a BUY signal.

But we had mixed results - the gains were certainly up versus the standard SUPERModel strategy over 31 years but at the expense of considerable additional vested time in the JSE (hence raising our market risk) and tripling the transactions from 10 to 40 over the 31 year period. As noted before, volatility was also high. And months where only 2% to 4% gains were to be had were consumed by brokerage charges. It just didn't feel right.

Then we tried another approach - what about incorporating the seasonality into the actual SUPERModel timing signal? What if we boosted the SuperModel Signal Line (SSL) by +1 whenever we entered a "power month"? Now since any positive SSL was already interpreted as a BUY, this would only really change the SUPERModel  strategy for those "neutral" periods that were previously tagged -1 and -2.

You will recall from SUPERModel theory that periods having none of Trendex, Repo Rate nor Econometric models flashing a BUY were tagged -2 and considered "very bearish". In these periods, a "power-month" would raise the signal to -1 which would still not create a BUY. And quite frankly, that's the way we would want it since these periods are characterised by so much risk, and such lacklustre returns (see the below table of stock market earnings power for various SSL signals) that we want absolutely nothing to do with the markets for the 22% of the time it is showing a -2.

Now for the 29.8% of the time that the JSE is showing a SSL of -1 we don't really have "bearish" conditions, we merely have "neutral" conditions. Remember that during these periods ONE of Trendex, Repo Rate or Econometric models are flashing a BUY signal. As we have shown in 31 year backtests on all these models, you very rarely lose money when you are in the market and they are flashing BUY signals. We are scared of being in the market when SSL is -2, but when SSL is -1 the only reason we would rather have our money in the bank is that we can get more than 5.3% return for much less market risk. However, investing in the JSE during "power-months" during -1 SSL periods might not be such a bad thing. Maybe this is how we could harness the power of seasonality without its volatility problems - if external conditions are not negative for the JSE (ie we have a SSL of -1) then maybe seasonality will work really well.

When we add the +1 from the "power-month"  to the SSL when it is showing -1, it pushes the SSL into 0 which is SUPERModel "BUY" territory. When the SLL is showing 0,1,2 or 3 and we add the "power-month" bonus point it merely reflects a higher bullish signal (which it should) but it does not change any existing BUY or SELL signals.

Another advantage of this strategy is that it should not generate a ton of transactions since the requirement of these periods that at least one of Repo Rate, Trendex or Econometric models are flashing a BUY will severely limit the amount of transactions.

Our mechanical rules for the incorporation of seasonality have been dubbed "Adaptive Seasonality Module" or ASM for short. The Adaptive Seasonality Module (ASM) is not a straightforward mapping to historically “good” months on the JSE, but an algorithm we have developed that takes the monthly performance of the last 6 months on the  JSE and matches it to historical norms and deviations to determine likelihood of recurrence of seasonality. It also has logic embedded in it to ensure you are not “whipsawed” in and out of the JSE for only 1 month or midway through a month whereby a 2-4% gain will be eroded by brokerage charges. As a result the ASM is optimised to either deliver minimum trades of 2 months duration for the seasonality component or ensure trades for less duration than this are profitable enough to not be swamped by brokerage charges.  

When ASM determines we have a "power month" coming up it adds a +1 bonus point to the standard SUPERModel-I signal line.

With the above refinement to the SUPERModel, we re-ran the 31-year backtest and achieved an astonishing 200,721% return (versus the vanilla strategies' 113,000%). This amazing result required only 15% extra time in the JSE (37 trades in total), taking the vanilla strategies' vested period from 48% to 63%. Just think about that - 200,721% return by only being vested in the JSE for 60% of the time, versus the ALSH buy-and-holds strategies' 7,663% return!

Since we have just completed extensive 11 year strategy analyses on our "Triple your returns with Sector Rotation" research piece, we decided to show the SUPERModel with embedded Seasonality performance over this period to give a comparative view against other strategies and of course what the effects of cyclical stocks, other ETF's and portfolios  would be on performance.

The chart below shows yearly growth of an ALSH buy-and-hold strategy (ALSH HOLD), a SUPERModel strategy executed against the ALSH index (ALSH SMODEL), a SuperModel strategy executed against the JSE Resources Index (RESI SMODEL), a SUPERModel strategy executed against our BLUECHIP portfolio (BLUCHP SMODEL) and then the same SUPERModel strategies but with the new embedded seasonality criteria.

The first thing we need to note, that is not in the above chart, is that all the seasonality strategies were vested in the JSE for 65.4% of the time and made 11 transactions in total (1 transaction per year on average versus the 1 per 2-3 years for the unseasonal strategies)

The second thing we note of course is the astonishing percentage returns delivered by the seasonality strategies versus the same vanilla (non seasonal) strategies - across the board. For a strategy executed against the ALSH index (say via a SATRIX ETF), the introduction of seasonality literally doubles returns from 756% growth to 1,527% growth. The minute we move into executing the timing strategy against an index of cyclical stocks, such as the JSE Resources Index, returns are taken from an already impressive 1,284% to an eye-popping 3,147%.

And then the icing on the cake is BLUECHIP. We thought we had reached the pinnacle of timing strategies with 1,897% growth with a SUPERModel Ver 1.0 strategy - but just look what happens when you implement SuperModel Ver 2.0 with the embedded seasonality - a mighty 6,264% growth which is an incredible 45% compound per annum!

The third most important thing we note is how embedded seasonality boosted the win-rates against the ALSH buy and hold (%>=ALSH) across the board, taking them from 63.6% to 81.8%. This is important as very few private investors can stomach a year when then strategy has under-performed the ALSH regardless of how well ahead they are from inception of the strategy. They succumb to loss of faith, thinking the strategy "no longer works" and bail out prematurely only to kick themselves after a year when the strategy roars back to life again.

The final point worth noting is the 2.1 Reward-to-Risk ratio of the SUPERModel Ver 2.0 strategy on the ALSH index. This is simply outstanding and a testament to the consistency, year in and year out of this strategy. Now do not get us wrong - the SuperModel 2.0 on BLUECHIP may "only" have a ratio of 1.4, but for 6,264% return this is simply unbelievable. There are very few private investors with even the best individual share picking ability that achieve Reward-to-Risk ratios of 1.4 - and for far less return!

The TRI return curves of the various strategies are shown below, with the SuperModel embedded with seasonality dubbed "SM2". SM2-BLU is SuperModel with seasonality on BLUECHIP. RESI is JSE Resources Index. SM-BLU is SuperModel without seasonality on BLUECHIP. ALSH is JSE buy-and-hold.


The SUPERModel strategy with embedded seasonality was back-tested 31 years against the ALSH index (assuming you had invested in the ALSH through a hypothetical ETF or index tracker such as SATRIX40) and over that period it delivered a stunning 200,721% return or 28% compounded per annum, versus the timed strategy without seasonality that recorded 113,000% return (25% compound). The transaction log for the individual trades executed appears below:

The first thing we note is how few "wrong way" trades we had - only 6 - versus the periods on the JSE (GRW OUT) when we were safely in the bank, that experienced a whopping 22 negative periods! This is the miracle of market timing at work. Sure, there were periods we were sitting on the sidelines when the market was doing well, but they were few and far between and the negative periods we side-stepped more than made up for the growth periods we missed. An yes, we did encounter 6 small negative trades while we were in the JSE but they were the minority.

Average CAGR when we were in the market was 47.9% with a standard deviation of 61.82%. But average JSE period growth when we were on the sidelines was only 11% with a massive 67.83% deviation meaning they were far riskier periods.

The 2nd thing we note is that average trade length was 6 months but ranged from 1 month to 37 months. Time "sitting on the sidelines" in a money-market account ranged from 1 month to 17 months.  Some more interesting data about the 31 year period with the seasonality based SUPERModel strategy appears below:

The most important thing that concerns us is the "positivity rate" - what percentage of trades resulted in positive growth. Here we see that the SUPERModel with seasonality achieved a positive trade 83.8% of the time! (% POSITIVE). This was achieved with 31 up trades and 6 down trades. We also note that we were in the JSE only 60% of the time - the other 40% of the time we were in some fixed interest vehicle such as a fixed deposit or call account or money market funds (your choice). This leads to a far better risk profile than a buy-and-hold strategy as we were only exposed to stock market risk 60% of the time.

Some other interesting stats are in the right hand part of the table. Over 75% of the trades resulted in a compound growth of more than 10% (ie we just beat inflation during the trade). Over two thirds of the trades delivered more than 20% compound growth! And just over half the trades delivered greater than 40% compound annual growth. This is vindication that if a SUPERModel BUY signal is present you chances of outperforming fixed interest investments are very great indeed.

Finally we look at the last two pieces of information - over 59% of the periods when we were "sitting on the sidelines" delivered negative JSE growth! Thank goodness the system gets us out the JSE and into some nice high yielding fixed-income instruments during these periods (typically these periods have a nice fat crash in them.) Also, 64.9% of these "sitting on the sideline" periods delivered growth less than 10% - meaning they could not even keep up with inflation!

The Total Return Index (TRI) of SUPERModel-I (no seasonality), SUPERMODEL-II (with ASM seasonality module) and an ALSH buy-and-hold strategy for the 32 year test-period are shown below

Incorporating seasonality characteristics of the JSE into market timing can deliver substantial increased returns as well as raise your win-rates against the JSE buy and hold strategy.

Seasonality works with Cyclicals and sector rotation just as well as market timing as is visible from the SM2-RESI and SM2-BLUE charts above. Over the 31 year backtest, the number of trades required only went up from 10 to 37. That's basically 1 trade per year.

Market timing is not perfect, but it certainly stacks the odds in your favour and lowers your exposure to stock market risk. The long and the short of it can be summed up as such : "Very little goes wrong on the JSE when the SUPERModel with Seasonality is firing a BUY signal. But quite a lot seems to go wrong when it is firing a SELL signal."

Basic Subscription Plan subscribers have access to SUPERModel 1.0 and Enhanced/PRO Subscription Plan subscribers have additional access to the SUPERModel 2.0 timing signal. The respective signals are shown below as an example:

MUST READS --> | JSE Seasonality | SUPERModel Ver 1.0 | Use Cyclical Stocks to treble returns |
How to use the PowerStocks Seasonality Actuarial Table & Chart (SATC) |
Make a Free Website with Yola.