In other words, when you're buying a stock, you don't actually care about whether the company is in an absolute sense good or bad, you care about whether it's relatively better or worse than what other investors think it is. In a perfectly efficient market, the answer is basically that it's always exactly as good as other investors think it is. Then your returns are based on random unpredictable shocks and the total trend of the market.
In a truly perfect market, it wouldn't matter what you invested in -- everything would have the same return. Your only real choice would be to invest narrowly or broadly, for risk/reward.
I wouldn't attribute the growth or collapse to random shocks, though those would of course play a role. I'd say that new products and distribution methods would still cause rapid and somewhat unpredictable growth or collapse.
Straight value investing would be more difficult with perfect transparency and efficient information, though it might work out as a dividend thing.
All stocks in a perfectly efficient market have the same EXPECTED return -- differential behavior can and will still occur, but it is necessarily unpredictable. Investing in any one stock is as good as investing in any other stock, in terms of expected returns. Now, there's still room for strategy in such a market: it's like, imagine the difference between betting $1 on a fair die, with one of two possible betting options: Either a 1-5 gets you nothing and a 6 gets you $7, or a 1 gets you nothing, a 2-4 gets you your $1 back, and a 5-6 gets you $2. In both cases, your expected return is 16% gain -- over enough iterations, you'll expect to gain money. But they are different in terms of risk/reward.
(And you can get similar risk/reward tradeoffs by investing broadly or narrowly in a perfectly efficient market.)
In theory you can do better with fundamental-weighting (i.e. weighting by revenues, earnings, P/E, PEG, etc.), but there are a lot of subtleties to a stock's future performance that are not captured in the financials and are exploitable against an investor that only naively looks at them. You end up with an investor that's picking up companies that are looting their future prospects for cash because their market is disappearing.
The ultimate problem is that stock market returns are a future-prediction problem. When a stock does better than expected, it goes up. When it does worse than expected, it goes down. Doing this reliably includes both having a complete understanding of the rest of the market's expectations and knowledge of the future, which is pretty challenging.
I think that behavioral-finance approaches, which gauge the emotions of your fellow investors, have actually been shown to work pretty well. "Be fearful when others are greedy, and greedy when others are fearful." Hard to put that in an index, though.
I don't know, isn't that what the junk-bonds guys did in the '80s, and more recent value investors? I could believe that the market still undervalues companies that are about to go bankrupt (but might not).