SMI S&P500 Out-of-sample tests

In this section we’ll perform the out-of-sample tests, we introduced in the previous section. These will show what to expect from the Single Momentum Investing strategy.

Out-of-sample tests for one-year investment periods

In our first series of tests, we simulate 25 one-year investment periods (testperiods). For each of these testperiods, we determined the optimal interval for the previous ten-year-period. We will call this preceding ten-year period the training period (in analogy with the training set, used for AI-algorithms). Next, we apply this optimal interval to the subsequent testperiod. In the table and the figures below we show the results for these tests.

CAGR SMI and S&P500 for different one-year testperiods last 10 years

For these first tests we looked back 10 years, and performed the out-of-sample test for 25 one-year-periods, with 100 market days between them. It ‘s clear that in the last ten years, someone who invested for just one year, was most of the time better of to buy and hold an S&P ETF in comparison to using the Single Momentum Investing strategy. The range of the returns (CAGR), as well as the average return, is worse using the SMI strategy.

But, the question is whether these last ten years are really a good guide. As illustrated in the second figure of the previous section, the last ten years we have seen a wonderful, almost uninterrupted bull market for the S&P500.

So let’s go back 25 years instead of 10 years. In the last 25 years we had two real bear markets: the burst of the dotcom-bubble in 2000 and the financial crisis in 2008. For the sake of visibility we ‘ll have 250 market days (one year) between each test.

CAGR SMI and S&P500 for different one-year testperiods last 25 years

Now the picture is different. We see that even for investment periods of only one year, the Single Momentum strategie protects portfolio’s against excessive losses in bear-market-years 2000, 2001, 2002 and 2008.

Although the average return (CAGR) of the SMI strategy is less than the average return of the S&P500 index, the minimum return is -11.8% compared to -40.2% for the index.

Out-of-sample tests for three-year investment periods

For investors with a three-year horizon, the advantages of the SMI become more obvious. The biggest loss in any three year-period in the last 25 years, is -6.1% vs -13.2% for the S&P. The average return is 7.3% vs 8.1%. More-over, in the bear market at the beginning of the century (2000-2003) the losses are limited. in the second bear market (2008-2009) the CAGR is positive for each three year-period whereas the S&P has negative returns for three three-year periods.

This is also very well illustrated in the first figure below, which shows the maximum breakdown for each three-year period. This maximum breakdown is never meaningfully higher with the strategy, and multiple times much higher in the S&P.

The three-year period with the worst performance for the strategy, is the last one. This is illustrated in the second figure below. The reasons for this underperformance are two very sharp V-shaped corrections, where the strategy pushes us out of the ETF for too long. And so twice, we miss an important part of the race upward.

Now that we have seen the effect of these two sharp V-shaped corrections in the three-year periods, we can show you something interesting about the one-year periods. Since it’s a sidestep, you can find it on a different page: sidestep.

Out-of-sample tests for five-year investment periods

We see for the five-year-periods, the strategy delivers only once a minimal negative return. The average return is only slightly lower than the average return of the index. Again the last periods show underperformance due to the sharp V-shaped corrections, mentioned above.

Out-of-sample tests for ten-year periods

For the ten-year periods, the results are, as expected, better again. No more periods with negative returns for the strategy, while for the index we still see three periods with negative returns. The average return is only slightly lower than the average return for the S&P500.

And next?

Now that we see where this is heading to, in our last section we will formulate general conclusions for the Single Momentum Investing strategy.