Sorry, no, I'm saying that historical returns have not demonstrated an ability by anyone to consistently generate alpha net of fees. Even the longest running market-beaters thus far do not fall far outside the expected range. Some have demonstrated statistically significant alpha, just not enough to account for their fees. In one sense it is certainly possible to imagine a string of returns that would do so, but it hasn't happened yet. (Also if you believe in (relatively) efficient markets, it seems unlikely to happen in the future either, at least in a fashion that could be identified ex ante, and thus profited from. But that's tangential to the point I was making above.)

The paper demonstrates this by comparing the actual distribution of returns to many sets of simulated returns over the same period. When the simulations are built assuming that fund managers have just enough skill to generate zero net alpha, the actual distribution has consistently lower returns across the board than the majority of the simulated distributions. It gets closer in the very extreme right tail (the 98th to 100th percentiles), but still not enough to generate positive four-factor alpha net of fees.