What each holding period returned

The table is most useful when read as a distribution rather than a ladder. Longer holding periods have historically produced larger total returns, which is the expected result for an asset that has appreciated over its life. But the annualised column tells a different story: the very long windows show lower annualised rates than the medium ones, because they include the early years when the asset was small and the later years when it was large. Compounding at a spectacular rate from a tiny base is not the same as compounding at a good rate from a large one.

The number of negative windows is small but not zero. No window in this table finished below its starting price, which is a fact about the particular windows chosen rather than a guarantee about holding periods in general. The windows are anchored on year ends, and a different anchor would produce a different set of outcomes from the same price series.

The honest summary is that holding period matters, but it is not the only variable. The start date matters at least as much, and the two interact: a long window that begins at a peak looks very different from a long window that begins at a trough. The rolling four-year returns page shows what happens when every possible window of a given length is measured rather than a single anchored one, and the twelve-month extremes page shows the same idea at the shortest horizon.