The failure of efficient market hypothesis is bluntly empirical. A wide range of observations contradict it. Two examples suffice. (1) Most sharp movements in stock prices are unaccompanied by any new information which can adequately explain the change. (2) The Shell stock has traded at different prices for years at a time on different bourses. This latter observation contradicts not only the efficient market hypothesis, but an even more fundamental assumption of modern finance, the "no arbitrage" assumption (which is an implication of "The Law of One Price" in economics).
It is not a problem for the efficient market hypothesis that market participants rely on past information. Analysis of past information may be essential to adequately interpret new information. In any case, the efficient market hypothesis is maintained even if there are market participants who act irrationally so long as arbitrageurs who invest rationally have sufficient liquidity.
It is also not a problem for the efficient market hypothesis that some market participants make money. A problem only arises if it is possible to make money by acting on old information. This raises the question of when new information becomes old. For example, the weak form of the hypothesis holds that no publicly available information is new. The strong form requires that even insiders cannot consistently make money through their access to insider knowledge. The examples provided above violate the weak form of the hypothesis.
The first chapter of Andrei Shleifer's "Inefficient Markets, an introduction to behavioural finance" provides an outstanding overview of these issues.