
#1
Jul610, 07:46 PM

P: 101

Ok. So I passed my defense, submitted, and graduated this past may with a PhD in physics. I am now looking for a job. I want to stay in technology and engineering, but I am studying finance (Hull, Shreve, Oksendal, etc....) just in case nothing pans out.
I was trained as an experimentalist, so my math background is minimal, but solid. The Calcs, up to PDE, with Complex and Linear Alg. And some numerical/computational methods, and probability picked up in quantum and stat mech. However, no analysis, modern alg, or group stuff. Would this explain why I find the Shreve book so confounding? I have to spend ages to make even a little progress. Is such difficulty normal? I've only glanced at the Oksendal book, but it looks a hell of a lot clearer. Heck, I've used +"noise" in my own notes a billion times. I was considering a postdoc, but after some career research, I've come to the conclusion it doesn't pay. I can only get a postdoc in my field using my experience, and my field is saturated and dying. Getting up to speed in another field will take a at least three months. So it's either industry or finance. Unless some one knows of finance oriented postdocs? 



#2
Jul710, 07:15 AM

P: 6,863

Personally, I'd spend more time reading up on statistics and econometrics than pumping a huge amount of time into Shreve. 



#3
Jul710, 09:28 AM

P: 478

C++ Patterns and Derivative pricing Teaches some of the finer points of solving financial problems with sophisticated C++ solutions. You should know about classes, inheritance and memory management (new, this, >) before you read this book. Concepts and Practice of Mathematical Finance A good intro but should be supplemented. In particular, chapters 6 and 8 were hard (for me) to learn from. Financial Calculus A good supplement to the above. Not as much content, but very nice explanations. Financial instrument pricing using C++ This is more of a reference. It's long, but goes into detail about a lot of the numerical methods that quants (presumably) use. Quantum Finance This is a great book for taking your intuition about quantum mechanics and path integrals and applying it to finance. As physicists, we've spent a lot of time thinking about quantum mechanics. At the end of the day, though, math is just math. If you can parse finance in terms of something you already (presumably) understand, it will give you some intuition/frame of reference regarding the new types of problems you will be solving. High Frequency Trading Seems like a broad introduction, but I'm not very far into it just yet. Check phds.orgthe bulk of the jobs are in high frequency/algorithmic trading. It might help to have some idea about the methods these guys use before going into a job interview. 



#4
Jul710, 09:37 AM

P: 101

Ok. So I passed my defense, submitted, and graduated this past may
Ben!!!
Your are the man!!! I've actually be looking for a good book on C++ and Finance. I find that nothing really cements my understanding of some phenomena better than writing some sort of analysis or simulation code!!!! Do any of these books go over hashing? I was thinking that's something I should also pick up too...! (I bombed a phone interview when it came to hashing) 



#5
Jul710, 10:00 AM

P: 478

Hmmm...the only hashing I know about probably isn't discussed in the presence of employers :) Browsing the indices of the books I listed, hashing doesn't appear to be discussed.
It sounds like you'd benefit from both of Joshi's books (Patterns and Concepts). He also has a discussion forum where he (personally, I think) answers questions regarding the texts. The longer book (Concepts) has about a dozen programming exercizes in the back, and Patterns comes with a bunch of libraries that Joshi wrote for derivative pricing. Let me also add that Hull seems to be kind of a waste of time to study. I think there's a lot of info in it, and it is probably a good read. But it seems to be written at more of an undergrad level. If you spend a month working through Hull, you will probably gain a good, juniorsenior level understanding of finance. Joshi's books are intended for a more advanced audience, like intro grad level. That being said, I have a copy of Hull on my desk that I turn to when I don't understand what Joshi is talking about. 



#6
Jul710, 10:08 AM

P: 478

Let me also say that, at the end of the day, we probably won't end up pricing derivatives. But you had to pass your candidacy exam by studying Jackson, even though you (most likely) never used any thing in it. For me, I want to have a good foundation and I want to learn the tools that the finance people use. I've learned, in my N years in physics, that the biggest barrier for entry into any field is just learning the language. The math is easy, and it's always the same, except there may be some new symbols. This is why I like the Quantum Finance book.
Finally, I want to be perfectly clear, that I am in the same position as you are, and this might be a bit like the blind leading the blind. I picked up Joshi's book when I couldn't understand a book by Carol Alexander (Market Models), a book to which I hope to return in the next month. I also have a stack of books that have turned out to be worthlesse. twofish is actually employed, so I defer to his advice. 



#7
Jul710, 10:58 PM

P: 6,863

There are two types of books that you want to read. One is about general C++ programming, and one is about general algorithms. What you want is to be a good *general* C++ programmer. Lakos and the O'Reilly books are good for this. Also google for quantlib. That gives you a good idea of the type of code that you will be expected to deal with. The other thing is that there tends to be too much emphasis on *writing* code, and not enough emphasis on *reading* code. A lot of your work will be on dealing with horribly written code that someone wrote five years ago and forgot. 



#8
Jul710, 11:08 PM

P: 6,863

If you are looking at something that you have no experience in, then the most useful think is to read the books enough so that you can know the major terms and the jargon. It's not that useful to try to develop deep knowledge of something you know absolutely nothing in, because if the job calls for deep knowledge in statistics and you don't have it already, you probably aren't going to get it, and you are better off interviewing for jobs that do focus on areas you have deep knowledge in. On the other hand, you can very quickly learn the buzzwords and the terms, so that you can understand the question and communicate sensibly, and answer very basic questions. You aren't going to be an expert in statistics if you aren't one already, but you can do a quick google and understand what Granger causality and cointegration is. It's very superficial knowledge, but superficial knowledge is important because it allows you to communicate with an expert with deep knowledge. 



#9
Jul810, 08:16 AM

P: 478





#10
Jul1210, 06:19 AM

P: 101

Thank you all for the advice. I think the Joshi book will still be good, because it's been 5+ years since I coded in C++. As for editing crappily written legacy code, my full name is Iforgot "I read and edit crappily legacy code for breakfast" Iforgot.




#11
Jul1310, 07:18 AM

Emeritus
Sci Advisor
PF Gold
P: 6,238

I would like to ask something. Apparently, coding (in C++) seems to be very important for quantitative analysts. Now, what I wonder is, why do these people code in something like C++ and why don't they use mathematical software like MATLAB or Mathematica or others in which  a priori  there's more power under the fingers, models can be implemented quicker, results analyzed with less effort and the like ?
Probably very naive question from my part but I wondered about this already some time... 



#12
Jul2410, 07:43 PM

P: 101

Hi Vanesch,
I was hoping some one else would answer this, because my answer is going to be pretty crappy. Here's my guess. As far as I know, Mathematica isn't good for large data sets. It's good getting relatively small models done quickly. I'm guessing the same applies to Matlab C++ can compile (Mathematica can too, but I don't know how good the compiler is). Object oriented programing allows for clever algorithms and clearer coding. More control over which algorithms are used and fine tuning them. Mathematica's library is very good, but I don't know how optimized it is for enormous data set processing 



#13
Jul2610, 09:45 AM

P: 478

Also, mathematica is pretty slow. With something like C++, you have a lot more control over designing specific solutions to specific problems.
Judging by job listings, though, many people in finance DO use MATLAB and R, for dealing with large data sets, and prototyping C++ algorithms. 



#14
Jul2610, 11:31 PM

P: 6,863

The problem with mathematica for hardcore derivatives pricing and algo trading is that it simply can't handle the volume and have the speed that you need for that sort of work. Also for that sort of work, you want to be able to tweak the internal models and implementations, which means C++. One other thing is that standard math engines often do not have the features and functions you need, and so most investment banks have an internal math engines that in some situations do what mathematica can do. Wolfram research is a very small company, and the major banks have more programmers and money than is available to Wolfram is often if you need something similar to, but not identical to Mathematica, it's often much cheaper and faster to write your own version of Mathematica, and if you rewrite Mathematica, large parts will likely be in C++. 


Register to reply 
Related Discussions  
Physics to Finance?  Career Guidance  11  
Going into finance after physics PhD  Academic Guidance  13  
studying maths and finance  Academic Guidance  4  
Finance Problem  Calculus & Beyond Homework  1  
Finance  Precalculus Mathematics Homework  4 