The efficient market hypothesis (EMH) was once a widely believed theory in financial economics (see, for example, “Efficient Capital Markets” [Fama 1970]). The EMH states, informally, that financial market prices fully reflect publicly available information concerning the underlying value of the traded security. As expressed, the EMH gives rise to a joint-hypothesis problem: when directly determining whether a given financial market is pricing securities efficiently, one must suppose a particular model of underlying value. Should the EMH, coupled with such a model, fail the test, one is (logically) free to revise the pricing model without rejecting EMH.
Consequently, the EMH itself is usually tested in one of three ways: (1) ‘weak form’ tests explore whether one can consistently outperform the market as a whole by studying patterns in past price movements (i.e., whether technical analysis is a loser’s game in the long run); (2) ‘semi-strong form’ tests explore whether one can consistently outperform the market as a whole by studying publicly available information (i.e., whether fundamental analysis, too, is a loser’s game in the long run); (3) ‘strong form’ tests explore whether one can consistently outperform the market as a whole by having access to information that is only privately available (i.e., whether even insider trading is a loser’s game in the long run). Consistently outperforming the market is a challenge for the EMH, because the EMH implies that, going forward, price movements constitute a random walk, meaning that the limiting probability of consistently earning excess returns is precisely zero.
Why does the EMH imply that future price movements conform to a random walk? If the price of a security fully reflects publicly available information, then the only cause of future price movements is genuine news, which is by its very nature unpredictable. Hence, movements in future prices cannot be predicted, meaning that trading is purely a game of chance. It is certainly possible to win games of chance a large number of times (consider Warren Buffett’s career, for example), but the larger the number of games, the less likely consistent winning becomes.
Most proponents of EMH concede the observation of Grossman & Stiglitz  that, if markets were efficient, it would not make sense for traders to sort through the news, thereby depriving markets of information needed to price securities efficiently. Thus, markets may well be efficient enough to deny traders much in the way of excess returns, but must be inefficient enough to continue to encourage informed trading.
The EMH near-consensus was supported by a large number of empirical studies showing that weak form, and many semi-strong form, tests of EMH conformed to the theory’s predictions (the track record of strong form tests is considerably more mixed). The consensus began to unravel when a number of criticisms from the behavioral finance school emerged, coupled with the increasing conviction of some econometricians that price movements may indeed be predictable. More recently, the EMH has faced a great deal of public criticism pertaining to its purported role in the financial crisis of 2007-2009. Many of these criticisms seem to me to be based upon some key confusions:
First, efficiency is a separate issue from stability. The fact that prices may run up, then suddenly fall off a cliff, does not necessarily condemn efficiency. Suppose, for example, that based upon the best information, housing seems to be a really good investment (perhaps because of a wave of immigration). Later, unexpectedly, compelling evidence to the contrary emerges (immigration seems to be slowing for whatever reason). Prices would rise sharply, then fall sharply, but there is nothing inefficient about this. It is just that, for a time, it was reasonable to believe housing to be a great investment, but this is no longer a reasonable belief. Instability is not necessarily evidence of inefficiency.
Second, efficiency does not imply perfect foresight. The claim of the EMH is that markets efficiently price securities based upon publicly available information, not that it prices securities based upon presently unknown considerations. The future is, to a very large extent, unpredictable. Expecting markets to get it right every time is folly.
Third, efficiency is a separate issue from rationality. Suppose a deeply irrational trader enters the market. His transactions cause a number of securities to become temporarily mispriced (with respect to efficiency). The consequence of this, however, is that rational traders now enjoy opportunities for arbitrage. Their efforts to profit off of this irrational trader will, in very short order, correct the mispricing. Thus, it is possible for many traders to be irrational, but for the market to still be efficient, so long as there are enough rational traders.
As for price trends, I will believe it when I see it. If you stare hard enough at any large data series, you will observe patterns therein. This does not mean, however, that you can reliably predict future price movements on the basis of past movements. If you can, then you can also profit off of those predictions. Where is the evidence of consistent excess returns derived from price trends? Show me the profit!
In my view, the EMH, like any good social scientific model, is a useful approximation for the purposes to which it is put, even if it is not precisely true. Assuming the EMH, we may infer from financial market prices what sorts of (rational) expectations traders have about future developments of interest. We may also explain why sticking with low-cost index funds makes more sense for casual investing purposes than turning one’s money over to expensive, actively managed funds. EMH may have this or that hole, but no alternative theory of financial market pricing gives us a useful interpretation of the relevant prices. Most importantly, no competitor theory can explain the most striking fact in finance: consistent excess returns are extremely, extremely hard to come by, in spite of some of the best and brightest working their very hardest using the most sophisticated tools to find them.