Interesting article. But rolling period analyses completely ignore and miss the most important driver of long term investor returns - the power of compounding. Would like to see the cumulative Dollar return from beginning of the data set. It is possible, indeed mathematically certain that one can compound at a lower return than the benchmark index for several years and still be climbing away from the index on a cumulative since inception basis, IF the base at start of the underperformance period is higher enough than index due to previous outlerformance periods. Rolling periods completely expinge the returns that came before, which is not the investor’s experience, assuming they have been invested for a longer period than the rolling term being analysed. Previous outperforming returns are still in the base, they do not disappear. Investor behaviour is the bigger problem - very few stay invested through a full cycle, being driven all to often by backward looking three year rolling period surveys.