On a completely unrelated note, if I were interested in creating a fund that appeared to have market beating returns for decades and I wasn't concerned about the legal consequences, here's one way I might do it:
I would create fund A and B and seed them with some initial capital. For fund A, I would create a machine learning model that took in lots of opaque parameters and hire a team of very smart mathematicians and computer scientists to optimize this model to perform profitable trades. In order to easily generate alpha, I would subtlety feed in some parameters that were correlated to the future market moving actions of fund B. Since I also run fund B, I uniquely have this information.
I would use the performance of fund A to woo outside investors into investing their money into fund B. In order to prevent fund A from overtaking fund B in asset value and thus diminishing the value of the subtle signals, I would close the fund to the public when it got sufficiently large and periodically distribute its assets to its investors.
In order to avoid any of my employees from eventually figuring out what's going on and reporting me, I'd incentivize them to avoid looking too closely by requiring that they invest a large part of their income into the fund with a long vesting date. I'd also require them to sign a long term non-compete so that they cannot work anywhere else in the financial industry if they leave.