Random Walk Theory
Random walk theory holds that the future path of stock prices cannot be predicted from past price movements, because prices already reflect all available information and react only to new information, which by definition arrives randomly and unpredictably. Under this view, a stock's next move up or down is no more predictable than the next flip of a coin, and any apparent pattern in a price chart is just noise that looks meaningful in hindsight.
The idea was popularized for a general audience by economist Burton Malkiel's 1973 book "A Random Walk Down Wall Street," and it underpins the efficient market hypothesis as well as the practical case for passive index investing: if prices really do move randomly around their fair value, then technical analysis patterns and most attempts at active stock-picking add no real predictive edge after costs, and a low-cost index fund is expected to do about as well as most professional stock pickers over time.
Random walk theory is genuinely contested rather than settled fact - critics point to documented anomalies like momentum and value premiums, and to the existence of investors like Warren Buffett who have beaten the market over very long periods - but it remains one of the most influential frameworks for explaining why consistently beating the market is so difficult, and a big part of the intellectual case for diversified, low-cost passive portfolios.
Two ideas get conflated here that are logically separate, and keeping them apart clears up most of the confusion around the subject. A random walk means price changes are independent and identically distributed, so nothing in the history of prices helps predict the next move. Market efficiency means prices reflect available information. Neither implies the other. A market can be efficient while prices follow a predictable path, if the predictability reflects a time-varying risk premium that investors rationally demand, and prices can wander unpredictably in a market that is not processing information well at all, simply because the noise dominates. The famous joint hypothesis problem follows from this: any test of efficiency is simultaneously a test of whatever model of expected returns it assumes, so a rejection never says which of the two failed.
The strict random walk has been rejected empirically. Andrew Lo and Craig MacKinlay's variance ratio test, published in 1988, used a simple property of a random walk: the variance of returns measured over a long interval should equal the variance over a short interval multiplied by the number of short intervals it contains. For weekly US stock index returns it did not, and the deviation was large enough to reject the hypothesis decisively, particularly for smaller companies. Later work found related patterns in momentum and in long-horizon reversal. None of this made the market easy to beat, which is the point most often missed: the deviations are small relative to transaction costs, they are unstable over time, and several weakened noticeably after being published. The honest summary is that prices are not a pure random walk and are close enough to one that the difference is difficult to convert into money.