Madoff's returns were clearly too good to reflect the 'split strike' strategy he intimated generated such returns. Further, they could not scale to the size of funds he managed. I suspect many investors thought his purported strategy was clearly a ruse, and that he was doing something illegal, such as front running, or using inside information from his various contacts.
This kind of thing happens quite a bit, though to work, you can't do it with $15B. In James Cramer's Confessions of a Street Addict, he covers his career as a broker for Goldman Sachs, and then a hedge fund manager from 1988-2000. He claims to have generated a 12 year track record with 24% annual return after fees. With an average hedge fund taking out 20% of the pnl, and traders taking out 10%, that would be an even higher 34%. Let's assume he and his traders didn't take the 10% trader bonus because he was not only the head trader, but the general partner. That implies a 30% average annual return for 12 years.
Now 30% is plausible, but he also states in the book he had only one down quarter. If we assume the probability of a down quarter is p, and he traded for 48 quarters (12x4), that means a maximum likelihood estimate for p of a mere 2.08%. If his 30% returns were normally distributed, one down quarter out of 48 is consistent with an annualized volatility of only 4.25%, for a Sharpe of 7.05 (using a simple Sharpe that ignores the risk free rate).
Warren Buffet's Berkshire Hathaway generated a measly Sharpe of about 1.1 over that period. The average of all Long/Short equity hedge fund returns, with survivorship bias, has only a 1.1 Sharpe from 1994 to 2005 (this is an upward bias of the average hedge fund sharpe, since the mean is the same but the volatility diversifies). Given he had some years up 65% and 45% (he mentions these anecdotally, not a complete list), and a maximum drawdown of -38%, it just does not square with a 4% annualized volatility.
I suspect there were at least three things going on. First, these are unaudited results, and he's exagerrating somewhat. Secondly, his strategy of paying lots of money to brokers and getting and giving them lots of information, with only $300 million in capital, might generate some abnormal alpha (though Cramer suggests he mainly trades on fundamental analysis, though his trading horizon seems to last weeks, not years). It's a unique strategy that probably worked well in the 1990's when broker upgrades and downgrades had more patent insider information--brokers would leak their recommendation changes to accounts generating lots of commissions. Lastly, he admits receiving lots of IPOs, and again, back in the 1990s that was pure arbitrage for anyone smart enough to understand the game. The quid quo pro is commissions to brokers for underpriced IPOs to the trader, all paid for by the issuer (because they only issue new stock once, and are afraid of challenging conventions). Again, with only $300 million in capital, a well executed IPO quid pro quo strategy might have been a viable and highly profitable strategy.
But he's clearly not telling near the whole truth when he says in interviews he mainly outperforms by using fundamental analysis, though his investors didn't care. A similar case is Michael Steinhardt, who's reportedly generated a 24% average annual return over 29 years, who like Cramer mentions he pays full commissions for the information, so perhaps there was some dealing going on their (like Cramer, he suggests fundamental analysis is the genesis of his alpha).