The basic premise of hedge funds is that smart people amply compensated can make lots of money. The empirical evidence for this is mixed. On average, including failed firms and management fees, it's doubtful that hedge funds beat a diversified market average (this is the basis for Buffett's bet). But a substantial number of firms have piled up impressive records over the years with high "alphas," or above market performance.
I just came from a talk by Markus Brunnermeier, who circulated the ideas I've discussed before. Much of the talk was focused on the amplification methods that turn a drop in asset prices into a financial meltdown, with some commentary on potential regulatory solutions. The idea is that the old agent-based macroeconomic models really fail to get at the connectedness of modern finance; but he suggested some tools for understanding those links. One was a measure of spillover risk across institutions, measuring by looking at past performance during crises. These turns out to be significant, and there were all sorts of interesting comments.
But Brunnermeier also examined the impact of taking these "tail risks" into account on manager alphas. It makes them disappear. That is, paper gains achieved by hedge funds--which have presumably hedged against shifts in the overall market--come through neglecting the counterparty risk associated with the occasional collapse of financial institutions. That's not to say hedge funds are useless--you would like to have the wealthiest people bear crisis risk, and it might be profitable for them--but it implies that hedge funds as a class rely on undisclosed crisis arbitrage for their profits.
Of course, hedge funds have not peformed as badly as markets in general this crisis, so perhaps this is a phenomenon particularly concentrated among certain strategies (or an artifact of old data, or not yet apparent). But it's more support of the claim that modern profit machines rely more on frauding investors than ouwitting markets.