I am a pricing quant - what is the most helpful thing to do increase my employability? Machine Learning?!

[YZhao]

I studied maths undergrad followed by a PhD in applied probability (optimal stopping). I currently have about 7 years experience as a quant, 1.5y in my first job (gas & power), 5y in my second job (equities) and less than a year in to my current role (gas). In my two stints as commodities quants, I got to code a lot of models in Python. Specially, in my current role, I have done some interesting least square Monte Carlo work which used ideas from existing literatures as well as my own (obviously developed from studying) However, the qualities of the team are / were average. I am currently one of the more experienced (if not the most experienced) quant in my team in term understanding derivatives pricing. Many other quants work on my quant dev-y stuff, ad hoc analysis / pricing. In my current shop, which is an energy firm, the set up is strange and outdated. There is usually not an agreed model used for pricing a particular type of products for deal structuring. This is partly due to the physical commodities products (especially long term and/or illiquid stuff) are not as straightforward as equities, but partly due to the way things worked in a lot of energy company.

My stint in the bank was not so good as banks have developed most models a long time ago. The job was in C++, but the job had so much support / documentation that I do not think my C++ skill is up to scratch for someone with this amount of experience. I also hated debugging C++ due to the bad set up in my previous work place. However, I did have exposure a wide range of equities products and is a reasonable quant technical wise (I read a Bergomi's Stochastic Volatility Models almost cover to cover among other things), though I am sure I still some gaps in term of understanding (e.g. volatility fitting / parameterisation).

I now have an optimisation problem - I have limited time I am willing to spend to improve my skills. The objective function I want to maximize is expected financial gain and the employability. I want to use my time most efficiently. I see a few paths going forward

1. Improve my programming skills / pricing quants skills - this means C++ but this is difficult because learning from a book is not the same as using the work to do practical work. I am more interested in modelling than coding. On the quant side, the issue I currently have with my current shop is that there are no senior people than me to learn from. I expect models at Goldmans to be much better, but it is unclear how they are better. The traders at this shop are also not technical.

2. Broaden my skill set - the one I have in mind is Machine Learning. Even with a maths background and trained in statistics, I found Elements of Statistical Learning too dense. I have switched to an easier book "Bayesian reasoning and machine learning". However, what I do not know is that how many roles are there out there that requires you to do know both derivatives / optionality pricing AND machine learning.
The other thing I have in mind is to learn more about optimisation (for the purpose of LNG portfolio optimisation) but this is a niche field, but I do not know where to start with this.

3. Look to switch role to be a structurer. I currently work closely with the structuring team (due to the organisation set up). A role change is possible. In fact, internal mobility is encouraged and many people who stay many years in the company have worked in multiple different teams before being given managerial positions. This would certainly have another option open for me. Though I do not expect the way to improve significantly (if at all) in the short term. This also allows me to get better understanding of the business and work in do more commercial, (but less technical) works. It is unclear here what is the best way to spend my study time if I choose to go down this tract.

Any inputs and comments would be deeply appreciated.

 

[MikeLawrence]

If you are comfortable sharing, where did you get your Ph.D at?

 

[YZhao]

If you are comfortable sharing, where did you get your Ph.D at?

Warwick. Why do you ask?

 

[MikeLawrence]

Just curious. I don't have any basis for this, but I just assumed a Ph.D in probability with a math undergrad background would be able to read through ESL and get most of it with some work.

2. Wasserman's 'All of Statistics' comes recommended as material you should really have down before you launch into deeper ML stuff. You can always start there. It's where I'll start whenever I finish the million other things on my list of things to study in my free time.

1 -- 3. I'm not sure how to comment on your role hopes, but the C++ courses here are incredible and are instrumental in getting people jobs and into top MFE programs. I highly recommend you check them out.

 

[YZhao]

I can probably read through ESL, but I would not be able to fully grasp all of the concepts. By contrast, when I read through Glassman’s Monte Carlo, I can grasp most of the concept quite easily because I did a Monte Carlo course before at university.

My c++ level is let’s say average. My question is more about whether that should be prioritised over other things like learning Machine Learning. Even though most banks use c++, many shops do not. I’d rather work for a place like my current shop than a bank even if I am to change roles.

My current role allowed me enough time to read a lot of papers and do some very interesting stuff.

The last question about structuring is more to see if anyone made this switch or vice versa and see how they found it.

 

[MikeLawrence]

I can probably read through ESL, but I would not be able to fully grasp all of the concepts. By contrast, when I read through Glassman’s Monte Carlo, I can grasp most of the concept quite easily because I did a Monte Carlo course before at university.

From what I understand if you try and go to a bank in a research type role you'll end up doing mostly data science type stuff, I think they'd expect you to have the stuff in ESL covered pretty well.

A better route might be to connect with recruiters and try and find a job at a more prestigious firm that fits your skill set. You've definitely got a marketable, niche skillset from the Ph.D, and while it is beyond what I'm able to work with at the moment I'd bet applied probability is applicable. There are others who know more who can help you better than I can. It seems like you know what you're doing, but are just in a sub-optimal place. I recommend you find a recruiter who can shop your profile around and see what opportunities present themselves or are waiting to be filled.

 

[YZhao]

I am happy with my current job. It is derivative pricing. I don’t want to work at more prestigious places. I’m making quite a bit more than I would make in a bank. Hedge funds might pay better but pricing quants are rare in hedge funds and the competition is brutal.

The question is more about whether it is broadening my skill set into ML after working in derivative pricing for this long. My preference is not moving into a role into doing only data science stuff. More thinking if there exists roles that need both where I would be good for. I don’t think I can ever do the data science stuff better than people who spent their entire career doing this type of stuff. I’m only think about acquiring it as a supplementary skill.

 

[MikeLawrence]

I am happy with my current job. It is derivative pricing. I don’t want to work at more prestigious places. I’m making quite a bit more than I would make in a bank. Hedge funds might pay better but pricing quants are rare in hedge funds and the competition is brutal.

The question is more about whether it is broadening my skill set into ML after working in derivative pricing for this long. My preference is not moving into a role into doing only data science stuff. More thinking if there exists roles that need both where I would be good for. I don’t think I can ever do the data science stuff better than people who spent their entire career doing this type of stuff. I’m only think about acquiring it as a supplementary skill.

I'll leave you to the experienced users then. I'm a lowly undergraduate whose main function on this site is to point new users to more knowledgable older users. Also, occasionally telling mid-career HR reps who don't know calculus that they should probably not quit their day job.

If no more experienced user shows up I may edit this message and then @ them. You've got a Ph.D so they'll be somewhat interested.

 

[YZhao]

I'll leave you to the experienced users then. I'm a lowly undergraduate whose main function on this site is to point new users to more knowledgable older users. Also, occasionally telling mid-career HR reps who don't know calculus that they should probably not quit their day job.

If no more experienced user shows up I may edit this message and then @ them. You've got a Ph.D so they'll be somewhat interested.

Your reply is much appreciated. When I got my first job and second job, they came easy, they were the first roles I applied for. I did do some preparation but I did not think it was an issue. This can make the people who are working in this field forget how hard this stuff is for the people without the same background.

What people don’t tend to realise that every decision we make, we are opening up possibilities but at the same time, kills other optionality. Once those optionality are killed, it might be very expensive to buy the equivalent optionality back.

 

[redoctober]

A professor of mine still dabbles in the industry and says the trend of machine learning is clear to him, its almost a required skillset now. He's also a friend of John C. Hull (from the very famous options textbook) and strongly recommends Hull's other book on ML, Machine Learning in Business. Apparently Hull's true passion has always been ML and derivatives pricing is just something he worked on on the side.

The book is relatively introductory and not directly specialized for quant topics, however almost all contributors to the book hold senior positions in the quant industry or academia and I strongly recommend it.

 

[weab]

ML is heavy trend on buyside

 

[YZhao]

A professor of mine still dabbles in the industry and says the trend of machine learning is clear to him, its almost a required skillset now. He's also a friend of John C. Hull (from the very famous options textbook) and strongly recommends Hull's other book on ML, Machine Learning in Business. Apparently Hull's true passion has always been ML and derivatives pricing is just something he worked on on the side.

The book is relatively introductory and not directly specialized for quant topics, however almost all contributors to the book hold senior positions in the quant industry or academia and I strongly recommend it.

Thanks for this. I work in commodities Physical commodities is not quite the same as derivative pricing because not hedging instruments are either non-existent or not particular. While in equities where I use to work, there has not been many new products on the exotics side and the people are turning to more vanilla products, pricing physical commodity optionality will be needed for a long time in this sector.

The people working in ML in my firm are usually in different team though my team might do this type of work too. I will bear in mind your comment and most likely spend some time learning this.

 

[MikeLawrence]

Hull's other book on ML, Machine Learning in Business. Apparently Hull's true passion has always been ML and derivatives pricing is just something he worked on on the side.

The book is relatively introductory and not directly specialized for quant topics, however almost all contributors to the book hold senior positions in the quant industry or academia and I strongly recommend it.

From the back cover and Amazon back cover:

"Provides the knowledge managers need to work productively with data science professionals."

"This book is for business executives and students who want to learn about the tools used in machine learning. It explains the most popular algorithms clearly and succinctly without using calculus or matrix/vector algebra."

If there's only hand-wavy math (I'm assuming there isn't NO math), and the book is clearly geared towards non-technical managers to get talking points, I'm not sure it's worth our time. If you're going to learn ML, learn it right and learn it in-depth. Anything else is just for cocktail party talk.

Edit: maybe @redoctober can send us a screenshot from the book's explanation of a topic to show the book is deeper than its marketing is making it out to be. I don't plan to spend any time on it though.

 

[YZhao]

From the back cover and Amazon back cover:

"Provides the knowledge managers need to work productively with data science professionals."

"This book is for business executives and students who want to learn about the tools used in machine learning. It explains the most popular algorithms clearly and succinctly without using calculus or matrix/vector algebra."

If there's only hand-wavy math (I'm assuming there isn't NO math), and the book is clearly geared towards non-technical managers to get talking points, I'm not sure it's worth our time. If you're going to learn ML, learn it right and learn it in-depth. Anything else is just for cocktail party talk.

Edit: maybe @redoctober can send us a screenshot from the book's explanation of a topic to show the book is deeper than its marketing is making it out to be. I don't plan to spend any time on it though.

The one I’m reading is good btw. Bayesian reasoning and Machine learning and it is freely available on the author’s website.

I’m sure what John Hull has to say is interesting but thanks for pointing this out.
I’d say Elements of statistical learning is pretty hardcore mathematical statistics I’ve not seen since my Cambridge days. I like it but the book I am looking at has more layman’s examples. I think practical stuff should always be before theory. We learn how to compute the probability of getting a pair in a 5 card draw before learning what is a probability space is.