#stationarity

Question 1: Single choice

An investor ranks securities by their past returns and buys the ones that did best, expecting the trend to continue. Seen as a sequence-modeling problem, what is the main flaw?

It assumes stationary dynamics, but the process behind returns can change, so the past need not predict the future1
It ignores the noise in daily prices, which averaging the returns over a much longer history would remove completely2
It looks too far back; returns obey a first-order Markov condition, so only yesterday matters3
It predicts only one step ahead, and one-step-ahead predictions are always unreliable4

Timed test