## SMI S&P500 part 3

In this section we will illustrate some important pitfalls of investment strategies. And we’ll propose a possible solution.

When we used interval 175 for the analysis of the Single Momentum strategy in the previous sections, we didn’t choose this interval by accident. On the contrary, we did extensive backtesting and found out the optimal interval for the complete 92-year period is 175. Of course, this doesn’t mean the interval is optimal for subperiods. More-over, so far we have no idea whether the strategy with this interval will have good chances to be successful in future shorter periods.

Same strategy, different interval

So, let’s first go back and find out what the results can be with a different interval. We’ll use the same SMI strategy, the S&P500 index and do the test for the complete 92-year period.

Oops! Changing our one parameter ‘interval’ a little bit from 175 market days to 250 market days, makes the results much worse. The yearly return (CAGR) falls from 6,42% to 5,22% and even gets worse than the return of the buy-and-hold strategy.

Same strategy, same interval, shorter period

And what would happen if we use the same SMI strategy and the same 175 interval but apply them to a shorter subperiod? For example: the last ten years.

This is even worse. The last ten years we saw excellent returns with S&P500 (CAGR 11.2%), but the return of the SMI strategy lags very much. It is 6.57%, which in itself is not bad, but an invester would not be happy with it, realizing that the underlying index did so much better.

Lessons learnt and the way ahead

Both examples illustrate nicely that we have to be very careful even when we use a very simple strategy with just one parameter.

No doubt the best way to find the optimal value of the single parameter of this strategy, is relying on out-of-sample tests. In out-of-sample tests we determine the optimal value of a parameter in a subset of the available data. Next we use this optimal value in a test with the same strategy in a different subset of data and determine the results. Actually in real-life investing we see a completely similar situation. The results of our investments will be determined by the price evolution in the future. Since we don’t know these future prices, we cannot determine the optimal value of the parameters for the future period. Therefor we use the next best: we determine the optimal value for a period in the recent past and hope this will give us good results. With out-of-sample tests we simulate this situation. This is why these tests have the best possible predictive value for future results.

If this sounds complex, we hope an example will clarify this a little more.

In the next part we’ll perform out-of-sample tests to determine when to use and what to expect of the Single Momentum investing strategy: