According to Wikipedia:
The efficient-market hypothesis (EMH) is a theory in financial economics that states that asset prices fully reflect all available information. A direct implication is that it is impossible to "beat the market" consistently on a risk-adjusted basis since market prices should only react to new information.
On it's own, this definition makes sense and is intuitive. However, I've heard in a couple of informal discussion that the EMH amounts to saying that you can't forecast the market, or that the best model for forecasting the market is a random walk.
I'm having trouble connecting the second statement with the 1st definition: The whole point of univariate time series forecasting is that we don't need to use causal drivers to forecast a time series, so that even if we don't have any new information, we could still model the time series itself to obtain a reasonable forecast. If we look for example at Amazon stock over the last 5 years (see below), an exponential smoothing model with trend fitted on data from 2015~2017 would have done a pretty good job of forecasting the stock price up until mid 2018. How does this tie into the idea that nobody can beat the market?
Am I misunderstanding the EMH? Or is Amazon stock (and similar tech stocks) part of some sort of fluke/tech bubble that represents an outlier statistically speaking?