|
How would people in credit, commod, fixed income, fx view a person (junior)with expertise in MCMC but not specifically stoch vol models ?
Is this alone a good skill to have,is it strictly "okay", or is it better to have this coupled with other things ?
I would really appreciate some responses.
K. |
|
|
|
 |
FDAXHunter
|
| Founding Member |
| Total Posts: 7526 |
| Joined: Mar 2004 |
| |
|
| Oh my god.... is your question seriously if "MCMC" is enough of a skill to have to get a job in anything? |
On tue un homme, on est un assassin. On en tue des millions, on est un conquérant. On les tue tous, on est Dieu. |
|
 |
|
Alright, fair enough, I will rephrase my question.
I currently have rest of the summer to study fairly intensively, either applications of MCMC to equity pricing models/ filtering models / stoch vol models / any others you guys would be kind enough to suggest. I want to get my hands dirty with some modelling area and since I was getting confused with different applications, I wanted some advice.
Any suggestions appreciated.
K.
|
|
|
|
|
IAmEric
|
| Phorgy PhynanceBanned |
| Total Posts: 2961 |
| Joined: Oct 2004 |
| |
|
Instead of spending the summer trying to do technical stuff, what if you spent the summer trying to learn something about being a quant? Others may disagree, but I thought reading Derman's ("My Life as a Quant") book was helpful. Then you should read all the other classics, e.g. When Genius Failed, Liar's Poker, Inventing Money, etc. That'll give you some idea if you even like finance. It sounds like you might not yet have a good understanding about what you want to do.
Good luck  |
|
|
|
JabairuStork
|
| Beat Box King |
| Total Posts: 970 |
| Joined: May 2004 |
| |
|
You gotta ask yourself, "would I want to hire a guy who spent his summer intensively studying applications of MCMC?"
Maybe you are talking about applying "My Cock, MELT, and Charlie" to art school chicks. |
|
|
|
|
|
JabairuStork,
Thats another way to look at it. Let me ask you this, did you write the above so that other
members will appreciate your humor/senoirity/ something else ?
You can send me private message about my question if you like.
I wouldnt mind that either.
Thanks IAmEric,
I am sure about spending the rest of my useful time in finance. Due to personal reasons i need a job by the end of the year. want to get the best out of my summer. I have a fair idea of what I want to do.
I am asking this question specifically, because I dont have a preference at this point into which area I would like to work in. If i can "get my hands dirty" on a few models it might be useful to target those jobs rather than apply at random and give me an edge over other candidates.
If I am any good, I will be around for a long time. Time will tell.
Cmon guys,
What do you guys see in the industry, does it help for me to focus on these issues ?
(Any) replies appreciated.
K.
|
|
|
|
Cheng
|
|
| Total Posts: 1797 |
| Joined: Feb 2005 |
| |
|
kanukatchit,
to rephrase your question: "I know how to use a screwdriver. Would that be suitable for a position as a plumber or maybe electrical engineer ?"
I think this kinda explains the answers you get, so better go back and think a bit more about what you really want. |
"We walk this earth / with fire in our hands / eye for an eye / we are Nemesis" |
|
|
|
FDAXHunter
|
| Founding Member |
| Total Posts: 7526 |
| Joined: Mar 2004 |
| |
|
Let me put it this way: Nobody will hire you simply because you happen to know MCMC.
There are some people who MIGHT (just might) hire you because you know MC Hammer. |
On tue un homme, on est un assassin. On en tue des millions, on est un conquérant. On les tue tous, on est Dieu. |
|
|
|
Well, thanks for being patient with me.
I understand that no one will hire me for knowing MCMC, its just a tool. I misrepresented the question the first time. Allow me to give a background about where I am coming from...
Generally speaking there are economic theories, mathematical models and statistical inferential techniques. Therefore we can loosely say there are three degrees of freedom while studying a financial model.
For example in a CAPM model there is an equilirbirum theory, there is a regression model and there are underlying assupmtions. Given data we try to estimate beta, alpha and sigma to test our hypothesis about our theory and various assumptions.
Simple linear regression assumes normal distributed errors along with other assumptions . As we all know these assumptions, might not be valid with real data. So lets say we assume now, the errors are t-distributed instead of normal, this dosent change economic theory, the mathematical structure of the model is the same, but the way we would go about estimating our coefficients and variance will change. This is one example where MCMC would be handy.
Similarly, Johannes and Polson give an excellent overview of models regularly employed in finance. These models have esimation problems that involve multidimensional integrals. MCMC is well suited for these problems with reasonable assupmtions.
I understand fairly well the theory behind asset pricing models, equity options pricing models, interest rate models, not so much credit but I did take a class in credit derivatives. Now, I would like to employ inferential techniques (MCMC) to study these models, more closely, from an statistical perspetive. Understanding the details, difficulties, and use my own ideas to better inference in these models, I hope would be useful to prospective employers.
At the same time, I do not want to restrict myself to say asset pricing models which might be useful to only a particular class of employers. Also, I wouldnt like to devote my time to studying option pricing models only to later find out that MCMC is very popular in the industry.
This is why I was wondering if people in the industry with access to these different discplines would recommend me to study certain class of models with MCMC in mind.
Thanks for reading till this point. I hope to get some direction from seniors.
K. |
|
|
|
|
|
Thanks for the reference. It is interesting.
I am not a "senior" but will weigh in anyway. If you have not yet worked in the markets, I think you should spread the net fairly wide. Given that positions come up somewhat randomly, as a newcomer it is suboptimal to target a particular area. Rather, just find somewhere you can work with good people, learn a lot, and do a great job.
If you need to start work by the end of the year, time is short, so start looking now, and good luck. Continuing to study and working on your own projects is fun, but you'll probably remain unfocused until you meet some people, learn what they really do, get some interviews, and crash and burn a few times. It's hard to explain, but somehow things become much clearer after that.
Being an expert on Bayesian inference and MCMC doesn't qualify you for any particular job. But if you can show that you can successfully apply technical skills to practical problems, then you will look good in just about any interview, even if the position is not related to the specific problem you chose to work on. So perhaps it is worth while working on a project - but only if you think you would have fun anyway.
However, here are some questions to ask yourself when you are thinking about what problem to work on:
- Can you explain the problem to an MBA, and why it is relevant to making money? (This is annoying, but very important. To understand why, think about who will be interviewing you.) Perhaps you could try this for each of the topics discussed in section 5.
- Can you get hold of actual market data to work on? Non-quants, sadly, are not very impressed by simulation studies.
- Can you realistically get it done by the end of summer?
You may find out that this rules out all the problems you are interested in working on right now. |
|
|
|
Cheng
|
|
| Total Posts: 1797 |
| Joined: Feb 2005 |
| |
| |
|
tabris
|
|
| Total Posts: 1010 |
| Joined: Feb 2005 |
| |
|
Kanukatchit: I think you are over valuing the relationship between doing a project and the project being relevant enough to get you hired. In all honesty, unless you think your project will make one million dollars for the company, save one million dollars from the company or you can win a nobel prize, the relevance is very small (and even on the last case, you would simply be a marketing tool). Most of the time, you would get hired and do everything that isn't related to what your project is about. So why bother, do something you like and learn about it better than the guy next to you. If you can use it when you get hired, good. If not, it wouldn't hurt as much as doing something simply to get hired... |
Dilbert: Why does it seem as though I am the only honest guy on earth?
Dogbert: Your type tends not to reproduce. |
|
|
granchio
|
|
| Total Posts: 1059 |
| Joined: Apr 2004 |
| |
|
Tabris: 1MUSD wouldnt be enough... hiring somebody in the long run costs more.
Kanukatchit: the project you do over the summer will matter very little. catching up with things you don't know might be helpful. As an example: I met a guy from academia thinking of moving to finance, that consider himself already well versed in quantitative finance issues, who did not know what local volatility is. Now that would be a serious drawback in trying to get a job at a level a bit higher than entry, at least as a quant in equity derivatives sellside. So finding out what is hot and having a few rough ideas about it could be better than knowing details of MCMC (never heard of it before today) |
"Deserve got nothing to do with it" - Clint |
|
|
|
JabairuStork
|
| Beat Box King |
| Total Posts: 970 |
| Joined: May 2004 |
| |
|
Ok my last post was probably a little harsh - it was Fri afternoon and I was probably drunk.
The truth is that issues of personality are going to be more important than a self-directed short-term project in getting someone hired.
However, if you really want to do something that shows you are motivated, business aware (as opposed to overly academic), and that will have some general utility, do the following:
Build valuation models for all the more commonly traded types of derivative contracts. You can find specifications in a reference such as Hull, or in papers that are easily dowloaded from public sites. Implement them using standard techniques, and create an excel front-end.
from an academic POV, this exercise is neither interesting nor difficult. It is, though, very similar to what you would be doing if you work as a quant on a trading desk. It will shift your attention away from a deep focus on a specific technique or method (like MCMC), and toward a wider focus on things like how to move data around, how to calibrate your model parameters, and how to make your code interact with a spreadsheet that the end-user will want to use.
Those things are much closer to the actual skills your employer is likely to want.
good luck. |
|
|
|
Graeme
|
|
| Total Posts: 1491 |
| Joined: Jun 2004 |
| |
|
The advice you have received in the last few posts is in my opinion very useful. If you go to an interview the chances are that, like (at least granchio and myself) they have never heard of MCMC.
By the way, talking in non-mainstream acronyms is in my opinion quite impolite - for example, if I was interviewing somebody and I asked 'what pushes your buttons the most?' and they said 'MCMC' I'd end the interview right there with 'well, FO&D'. (the D stands for die, by the way).
Academia and academics have a tendency to emphasise the academically nice stuff - and that is fine - but at the expense of some important stuff, which is not. For example, in a university honours degree that I was until recently involved with, every year I would ask in my final exam some variation of 'what is a short sale?' and every year I would be amazed at the variation of quality of answers. And these self-same students could derive the optimal stopping time of your uncle's favourite stochastic process with drift.
The point is that a general quant knowledge is much more important than specifics. This is why a resource like Hull, although poo-pooed so easily, is quite worthwhile.
|
Graeme West |
|
|
|
|
Completely agree with the last few posts, especially as it relates to fundamentals you'd find in Hull.
I once interviewed a quant candidate - physics PhD, numerics expert but no common-sense when it came to finance or trading.
I asked him if he'd read Hull - he said he "skimmed it and found the math very basic". I gave him the simplest of simple questions on pricing a forward contract and he came up absolutely empty. I wanted to end the interview right then and there but unfortunately the head guy liked him and hired him (he didn't last long).
MCMC might be the sort of highly-specific skill an outfit like RenTec or Citadel would value. When you hire quants by the dozen it probably pays to diversify (I'm completely speculating here). But even if I'm right, kanukatchit, you'd be hired as a senior quant's personal slave: he creates the Bayesian models, you do all the dirty coding work. |
|
|
|
|
The last few posts are reminiscent of a passage in Derman's book. I'll probably get the details wrong, but it was something along the lines of: Someone asked him a question, he ended up researching it in depth, the person who asked him the question chastised him for wasting so much time on it. Later he got high accolades for improving the GUI of an application.
Learning about what an employer does is the most important thing you can do. If you get a chance to ask them questions, the best thing you can do is to keep asking them more and more about what they do, and why they want to hire you. For long term employment, there is no ideal list of talents, there is no ideal person. You can show interest and personality, and not be correct about things, or forget facts. It's more about convincing them that you are smart and a fast learner. Manage the interview process. Even after you get the job, it is important to understand the difference between getting involved in office politics (which is horrible), and managing people's perception of you. It's one thing to come across as goofy, strange, off-the-wall, too-intense, forgetful, distractable, etc (I am all of these things). The important thing is to give people a reason to believe that when you voice an opinion, it is worth their time to listen.
-t. |
Any advertisement in public space that gives you no choice whether you see it or not is yours. It belongs to you. It's yours to take, re-arrange, and re-use. Asking for permission is like asking to keep a rock someone just threw at your head. -Banksy, street artist (b. 1974) |
|
|
|
|
Wow guys, guess i am getting the drift.
well, when i see people with a Phd in some god damn topic i dont even know exists, it sort of gives a general idea of the people out there in the market. You sort of need a highlighter on your resume somewhere.
Dont get me wrong i really know what i like, I am not merely chasing silver bullets, but i feel like there will be time to get to stuff i like when i have experience and am more in command. I need a foot in the door.From your responses though, i am clearly missing the point somewhere.
The thread pointed out by Cheng is interesting. It probably makes you realize that there is no typical entry point to this profession. although there is probably some bias in category of people that replied there.
I guess i got interested in this (MCMC, Markov Chain Conte Carlo) in the first place because I liked bayesian estimation problems. Estimation problems are sort of cool, cause that is where your statistical assumptions come into play. People dont seem to focus on them, and fail to realize their implications because its a secondary problem. But without error, there is no estimation. I was reading reza's challenge problem on the other forum, and i think most people didnt provide a estimation error with their estimates.
In my short interaction with the industry, I am amazed the way people dont care about properties of estimator and more importantly the error behind it.
Also, its very $@#$ing irritating when people get away with any bullshit.
There was an instance when some senior mgmt people came to my "quant" colleague and asked him to come up with an estimate of some sort of a VaR for the year end. And they began discussing some suggestions like Monte Carlo, volatility in absolutely no context and then it ended when my "quant" colleague said but what is the point of this calculation, volatility keeps changing all the time. Clearly, this industry safely houses a lot of people. I heard somewhere that "A" category employees tend to hire "A" category empolyees and so on and so forth. Enough rambling.
My background is good enough for me to get a interview calls (atleast thats what i think) so, i am just going to concentrate of basics, stuf i like and improve my C++ skills and work along the lines of something like JabairuStork suggested.
Thanks for all your help guys,
K. |
|
|
|
|
Here is a follow up to this thread,
Somewhere along the thread JabairuStork gave the following suggestion,
----------------------------------------------------------------------
However, if you really want to do something that shows you are motivated, business aware (as opposed to overly academic), and that will have some general utility, do the following:
Build valuation models for all the more commonly traded types of derivative contracts. You can find specifications in a reference such as Hull, or in papers that are easily dowloaded from public sites. Implement them using standard techniques, and create an excel front-end.
from an academic POV, this exercise is neither interesting nor difficult. It is, though, very similar to what you would be doing if you work as a quant on a trading desk. It will shift your attention away from a deep focus on a specific technique or method (like MCMC), and toward a wider focus on things like how to move data around, how to calibrate your model parameters, and how to make your code interact with a spreadsheet that the end-user will want to use.
Those things are much closer to the actual skills your employer is likely to want.
good luck.
----------------------------------------------------------------------------------
I wanted to show that I am motivated and more, so I built an Excel Add-In library using C++ from scratch. I took this up a month ago and have succesfully implemented amongst others a Least squares Monte Carlo solution using Longstaff & Schwartz 2001 for the American-Bermuda-Asian call option and the American Put. Here are the results for these options in comparison with LS 2001.
And here is the library if anyone would like to use or compare
Attached File: EOptionsLib_NP_v1.0.zip
It values the following options as excel functions.
- AmericanBermudaAsian Call
- American Put
- Asian Call
- European Call
- European Put
- Digital European Call
- Digital European Put
This c++ library can be easily extended to other options. I would be very interested in any suggestions/comments.
Also, I would be interested if anyone wants to collaborate to build up something. Finally, since I am looking for a job anything in this direction would be very useful too !!
Thanks for reading so far,
K.
|
|
|
|
|
|
This is a good resource for jobs, industry news, etc.
http://www.quantster.com/ |
Capital markets reward you for what you learn that other people have yet to ascertain |
|
|
