> But I thought that efficient market theory held that prices were not
> predictable from past prices....
That's another way to say the same thing. No contradiction. The claim is that the *changes* in a stock price (i.e. P_t - P_{t-1}) are a sequence of shocks, where each shock is a random variable independent of all others in the sequence (and also "identically distributed"). The price today is separated from the price yesterday by a memoryless random shock.