Mr Bayes may say, given an all-time high, there were 1,565 occurrences when the SP500 (^S5T in GTR1 speak) 5 years later exceeded a 20% total return.

A fun stat, to be sure, but it is pretty irrelevant. A market can hit all time nominal highs at low valuation multiples (as after a long secular bear) or at high valuation multiples (as at the dying end of a bubble). Your figure makes no distinction, so it can and should be ignored.

The past is quite clear on one thing: for time frames longer than the current market move, more than a couple of years, average market returns are terrible starting from high valuation levels, and excellent starting from low valuation levels. It isn't random, so stats starting from random dates in the past, or subsets of dates that don't correspond to the current valuation situation, have no merit. Just "broker economics": )

The logic is simple. The future trajectory of company earnings will be what it will be. That is the source of the value of equities. If you pay more for the same old set of future earnings, you get back less per dollar invested.

Jim