The most interesting exchange I had was with some derivatives traders, and I asked broadly what kind of math was being used at their bank. Was it graduate level, was any of it novel, and was it within the reach of programmers. The answer was "no, no, yes" and they wouldn't say anything else. I suspect the secrecy is creating an intellectual mean reversion, where you do a thing that works and run it until it is traded out, and then never say anything about it so you can cost your competitors the time thinking about it.
However if none of the math is novel, then current strategies are not information problems that need new science, they are work problems that piece together solved ones, and whose efficiencies come from competency in related domains. I could see how one could become seized by markets like that.