Thread by @tdgraff

When I first got into the bond biz, I spent a lot of time picking apart various kinds of securities. I made my own OAS models, rebuilt ABS structures w/ triggers from scratch, deconstructed derivs/cash basis, etc.

All good learning, but the best lesson was something intangible.

It is important to learn stuff like how much does various assumptions change the outcome, like lumpy losses in ABS or unstable vol assumptions in OAS. This exercise helped me see all that. But for all the months I spent building these complex models, I never use any of it now.

There were two huge big picture lessons I learned. First is that you should treat things like OAS, yield, duration, KRDs, etc. like concepts not facts. A bond w/ a duration of 8 is more volatile than one w/ a duration of 2. But a portfolio with a dur of 4.5 vs. 5.0 is less clear.

All that modeling from scratch taught me that quant is really important, but you can't take it literally. If you sat on my desk, you'd hear me use terms like around, about, approx, etc., in regards to every portfolio/bond stat I quote.

The other huge lesson is that there is no magical insight hidden deep within a bond calculation. E.g., OAS has lots of flaws. Say you come up with some conceptual improvement on OAS. Great, but would that actually help you outperform? Probably not.

This idea really stems from my first observation about the quant being more conceptual. If you try to narrow your quant down, you wind up putting more weight on something that just has a different set of assumptions and brings a different set of uncertainties.

I'd rather just embrace the uncertainty in the first place and not distract myself.

Now that isn't to say you always gloss over details. We stress the heck out of structures using all sorts of odd ball scenarios. But we are aiming to guess at a level of uncertainty, not seeking to find the "answer" for where the structure breaks. There's a distinction there.

When I am counseling young analysts on these ideas, I know it feels unsatisfying. It would have when I was 25 too. It might even sound less sophisticated. I assure you it is not. I argue this is the difference between wanting to be the most clever analyst or trying to make money.