- You get to know something that isn't published. You find a strategy, you don't tell anyone. Sure, you can talk about it in broad strokes (we're trend following...) but it's pretty unlikely you'll ever tell anyone enough for them to be able to replicate it. Except people you trust, of course.
- You're always thinking about philosophy of science. How do I know that this series of steps I'm doing is not just a monkey and a dartboard? Again, there isn't going to be a specific written paper that tells you the answer.
- At one end of it, you are working with cutting edge technology. Anything that shaves a microsecond off the speed is useful, and you can spend a lot of time optimising such things. I've been staring at a C++ solution that is 100% in house. No STL.
- At the other end, you are working with data in very modern ways. All that ML stuff in the news is interesting to quants, because if there's one thing ML papers are about, it's how to avoid overfitting. Just a few years ago people in the field had not come across ML, or at least the point of view that ML brings. Plenty of other things had already come up, though, like information theory, time series econometrics, signal theory, and so on.
- About derivs quants, it's not my cup of tea, but I can see how others like it. You're basically pricing things hoping that you've noticed something the counterpart hasn't, or hoping you've found a cheaper way to hedge than your competitors, or simply hoping your salesman is a better salesman than the other guy's.
- You get to know something that isn't published. You find a strategy, you don't tell anyone. Sure, you can talk about it in broad strokes (we're trend following...) but it's pretty unlikely you'll ever tell anyone enough for them to be able to replicate it. Except people you trust, of course.
- You're always thinking about philosophy of science. How do I know that this series of steps I'm doing is not just a monkey and a dartboard? Again, there isn't going to be a specific written paper that tells you the answer.
- At one end of it, you are working with cutting edge technology. Anything that shaves a microsecond off the speed is useful, and you can spend a lot of time optimising such things. I've been staring at a C++ solution that is 100% in house. No STL.
- At the other end, you are working with data in very modern ways. All that ML stuff in the news is interesting to quants, because if there's one thing ML papers are about, it's how to avoid overfitting. Just a few years ago people in the field had not come across ML, or at least the point of view that ML brings. Plenty of other things had already come up, though, like information theory, time series econometrics, signal theory, and so on.
- About derivs quants, it's not my cup of tea, but I can see how others like it. You're basically pricing things hoping that you've noticed something the counterpart hasn't, or hoping you've found a cheaper way to hedge than your competitors, or simply hoping your salesman is a better salesman than the other guy's.