There's a very long multi-year thread at a website called Nuclear Phynance about this. Nobody seemed to have the faintest clue although amusingly some of the more outlandish suggestions were later (laughingly) cited by an insider in an interview.
One decent general idea was that RenTec was willing to entertain purely statistical signals with no intuition to back them up (or, alternatively, unintuitive signals which you can explain in plain language but sound stupid).
There was an intesting quote from Simons about how astronomers are the best suited hires because they are used to making measurements/hypotheses purely from observation, without the chance to do controlled experiments. It was also noted that they seemed to employ a lot of NLP specialists, although consensus was that this was down to pattern recognition rather than anything to do with language.
The best suggestion (or at least the best sounding one) was that they were doing hidden-markov models which took into accound second and third order effects. And then there's always just the 'best execution' argument (doing basic market making but engineering better systems than anyone else).
However information on them is so scarce I sometimes wonder if most of what we hear is pure myth.
Haha, aggressive indeed. What are people's opinions on regulating this kind of sidestepping?
Seems to me the status quo of regulation means a constant game of catch-up. No governing body has the ability to say 'You can't do that' until it's too late.
One decent general idea was that RenTec was willing to entertain purely statistical signals with no intuition to back them up (or, alternatively, unintuitive signals which you can explain in plain language but sound stupid).
There was an intesting quote from Simons about how astronomers are the best suited hires because they are used to making measurements/hypotheses purely from observation, without the chance to do controlled experiments. It was also noted that they seemed to employ a lot of NLP specialists, although consensus was that this was down to pattern recognition rather than anything to do with language.
The best suggestion (or at least the best sounding one) was that they were doing hidden-markov models which took into accound second and third order effects. And then there's always just the 'best execution' argument (doing basic market making but engineering better systems than anyone else).
However information on them is so scarce I sometimes wonder if most of what we hear is pure myth.