SMI S&P500 sidestep

In SMI S&P500 out-of-sample tests, we only allowed intervals between 50 and 200 market days. We see that for the three-year testperiods, the performance of the SMI strategy is bad in the last period, due to two very sharp V-shaped corrections. Therefore we want to test whether allowing longer intervals can result in better returns.

In the table and figures below we did the same out-of-sample tests for the one-year testperiodes but we allowed intervals up to 500 market days.

Out-of-sample tests SMI and S&P500 for different 1-year periods, intervals up to 500 market days allowed

We see the results for the last two years are better. With intervals up to 500 market days the strategy delivers returns as good or better than the underlying index.

The average CAGR of the strategy for the last 25 years, is better than the CAGR for the index as well: 9.4 vs 9.0%. On the other hand the minimum CAGR for these one-year periods, is more negative than when only intervals of up to 200 market days were allowed: -17.2% versus -11.8% with the previous approach.

For the three-year periods, we didn’t find this effect yet. The technical exlplanation is the fact that we optimise the interval in the preceding ten years (the so-called training period). For the three-year periods (the last one starting at the beginning of 2017) this optimisation so far doesnot result in intervals longer than 200 market days, even if they are allowed.

Lesson learnt

So far we stick to our first approach, allowing intervals up to 200 market days. Although with this approach results are worse when confronted with sharp V-shaped corrections, it gives us better protection in deeper and longer bear markets.

But we have to stay flexible. We believe basic drivers of market-evolutions will probably not change over time, with human emotions and psychology being some of the most important ones. But technology is changing and this could speed up markettrends. If we see more of these sharp V-shaped corrections, we might change our strategy.

Back to the SMI S&P500 out-of-sample tests.