It's important to note that the EMH doesn't suppose that all actors agree on the price. It just means that the price represents the summed outlook of participants.
Those who think a stock is underpriced will bid it up to get more; those who think it is overpriced will sell out at progressively lowering prices.
Because the movement in prices creates an opportunity to profit for those who have information that is not yet revealed, those people will enter the market and create pressure on the price.
In some mathematical forms, this is assumed to happen instantaneously and universally. Thus: all available information is factored into the price. Essentially, you can't arbitrage because you have instantaneously driven up the price already by arbitraging. (It's weird, yay calculus).
The looser forms basically say that this is what happens in the long run, on the average, even without instantaneous adjustment of prices. And when you compare market time series to pure randomness, they have similar characteristics. So in a sufficiently large group of agents, some will profit and some will lose merely by chance. Then you're back to taking an average and being unable to beat it in the long run, because in a game of chance outcomes converge to the long-run probabilities of the game.
Okay, so would I be correct to assume that you claim it is in the long run impossible to beat the market except by dumb luck?
I don't see how this follows from your reasoning. You've allowed the possibility of beating the market if an individual trader has some knowledge or insight which the average of the rest of the market does not. So presumably an investor which only acts when he has such knowledge would beat the market also in the long run.
The EMH grew out of the empirical observation that prices wiggle around randomly (and the different forms look at different ways available information could be responsible for that).
Buffet himself, in the letter, has already said that most investors start with comparable amounts of information and capability. So speculating by buying and selling according to predictions of future values of stocks will eventually cause a regression to the mean.
Buffet disagrees with the EMH but he agrees with the conclusion that you can't beat the market by speculating on the movement of prices. He basically lets the random wiggle happen, and when a price dips to what he thinks is a bargain, he buys.
The problem is that we struggle to test the null hypothesis. We can't create 1,000 parallel 20th centuries with 1,000 Warren Buffets to see if he comes out significantly ahead a statistically meaningful number of times. We can only compare him to chance. In a sufficiently large sample, long streaks of perfect performance can emerge purely by chance.
So what you're saying is that it may (or may not) be possible for an individual inevstor to consistently beat the market through skill, but that it is impossible to verify whether an investor who beat the market is in fact skilled or just lucky.
If you believe Buffett's thesis that skilled value investors who do their homework consistently beat the market, it stands to reason that there should also be other strategies that consistently beat the market (let's say over 20 years or so). It also stands to reason that value investing, since it is such a heavily publicized strategy, probably does not beat the market today, since it is such a widely published and acknowledged strategy.
Those who think a stock is underpriced will bid it up to get more; those who think it is overpriced will sell out at progressively lowering prices.
Because the movement in prices creates an opportunity to profit for those who have information that is not yet revealed, those people will enter the market and create pressure on the price.
In some mathematical forms, this is assumed to happen instantaneously and universally. Thus: all available information is factored into the price. Essentially, you can't arbitrage because you have instantaneously driven up the price already by arbitraging. (It's weird, yay calculus).
The looser forms basically say that this is what happens in the long run, on the average, even without instantaneous adjustment of prices. And when you compare market time series to pure randomness, they have similar characteristics. So in a sufficiently large group of agents, some will profit and some will lose merely by chance. Then you're back to taking an average and being unable to beat it in the long run, because in a game of chance outcomes converge to the long-run probabilities of the game.