• C++ Programming for Financial Engineering
    Highly recommended by thousands of MFE students. Covers essential C++ topics with applications to financial engineering. Learn more Join!
    Python for Finance with Intro to Data Science
    Gain practical understanding of Python to read, understand, and write professional Python code for your first day on the job. Learn more Join!
    An Intuition-Based Options Primer for FE
    Ideal for entry level positions interviews and graduate studies, specializing in options trading arbitrage and options valuation models. Learn more Join!

No Longer Seeking Quant Jobs - Market is Changing

Joined
5/29/15
Messages
30
Points
28
I've been concentrating on developing my deep learning resume for the past couple of years, and have recently started to get promising interviews for deep learning R&D roles, specifically for self-driving cars.

It seems every company these days is scrambling to staff a new self-driving car division. The labor market demand is very stiff. Because of this, I almost feel embarrassed that I even considered getting a machine learning job at a hedge fund. The whole idea of using machine learning to try to predict the stock market seems asinine, in hindsight--especially considering the fundamental statistical in-feasibility of doing this.

Folks, we PhDs suffered greatly during the whole 2008-2015 social media bubble. The market had become a cesspool of apps, websites and trivialities. However, things are changing, and companies are starting to staff real R&D efforts again, and downsizing their "on demand petrol delivery app" efforts.

I feel we have reached an inflection point where many companies are now aggressively pursuing deep learning knowledge. Meanwhile, the hedge fund industry openly said at SALT that things are in a long term decline. There are simply too many hedge funds who can all download Torch or TensorFlow. I don't see how anybody can beat the market doing this.
 
youre a typical PhD guy. you make general comments about things that you think you know about but never had skin in the game to actually talk about. give things a go before making big bold statements about beating the market
 
youre a typical PhD guy. you make general comments about things that you think you know about but never had skin in the game to actually talk about. give things a go before making big bold statements about beating the market

Ok, so by your logic, in order to be a good FBI agent, you first have to be a career criminal?
 
lulz, so much logical fallacy in this thread.
 
Anyway, this discussion is really besides the point. I'm not saying there aren't some funds that use clever data sources and/or insider information to make good predictions. All I am saying is that now that I am able to get a great research job in the tech industry, the idea of doing machine learning to skim a few micropennies off of the market for some abusive billionaire, while wearing a suit and commuting in a cramped, dilapidated subway car seems ridiculous in hindsight. Of course, I didn't know a year ago that I would have such good fortune.

But since I can wear a T-shirt and shorts to work, work 40-50 hours a week, and make 75% of what a top hedge fund pays quants, and do interesting work that helps society, it's kind of a no brainer.
 
Anyway, this discussion is really besides the point. I'm not saying there aren't some funds that use clever data sources and/or insider information to make good predictions. All I am saying is that now that I am able to get a great research job in the tech industry, the idea of doing machine learning to skim a few micropennies off of the market for some abusive billionaire, while wearing a suit and commuting in a cramped, dilapidated subway car seems ridiculous in hindsight. Of course, I didn't know a year ago that I would have such good fortune.

But since I can wear a T-shirt and shorts to work, work 40-50 hours a week, and make 75% of what a top hedge fund pays quants, and do interesting work that helps society, it's kind of a no brainer.
I don't normally say this when I read posts like this, but you will have to come back to us once you have had a few years experience in the real world.

Vertigo sums it up well, I just hope you wise up after a few years - at least you're not like some IT server admin clowns I know that think because data science is 'tech' they can give me advice as if I am 3 year old, nothwithstanding that I had a career before using maths in finance so know more about data science than them.

Wind people up all you like and kid yourself but it will come back to bite you in the ass in the future if you do it in your own industry and get a reputation for it and are looking to move. Much as you think you've hit a goldmine, politics will matter as you progress. Let's see how you fare then.
 
Ok, so by your logic, in order to be a good FBI agent, you first have to be a career criminal?
Nope, perhaps I should let Vertigo comment, but I will add my own comment first.

My own experience with PhD guys is that until they do something like actually invest in the business they work in they never truly understand it.
 
Nope, perhaps I should let Vertigo comment, but I will add my own comment first.

My own experience with PhD guys is that until they do something like actually invest in the business they work in they never truly understand it.
could not have said it better myself.
 
Trying to grade grub by buying term papers is a little like drug cheating in sport where it is obvious who does it by looking at who suddenly improves. But while drug cheats can use masking agents and not get tested, a lecturer can simply interview someone about their project and see if they are full of shit. And when I was in undergrad about 15 years ago some lecturers already had plagiarism software in place.

I just can't see how such a website would have any impact on the quant market at all. Even if it was just used as an additional research resource, it is a drop in the ocean compared to what else was already out there when I was in college.

Also don't forget that to become a quant you will have to answer a battery of technical questions in interview. Some firms just give a chat interview and you can simply say "yep, did Monte Carlo" but even then they will ask questions like why certain distributions are used and what you think the biggest challenge is in pricing.

I think what has had much more of an impact is that people are more aware of the role. When I was doing my masters I had math and actuary undergrads asking me if my masters was "an accountancy thing". Nowadays every Tom, Dick and Harry is trying to be a quant, so even if you had the availability of quant jobs we had in 2003 chances are it would still be hard to get in.
 
Last edited:
Cheating in academia is as rampant as doping in sports. Know enough LSE finance jocks who had outsourced courseworks & dissertations and still mananged to land up snippy job offers at top funds. When the teaching deviates from good ole' socratic method to Power Point, Grade/exam gate keeping this is bound to happen.

Also you are overstating the integrity of lecturers, some of them are just unfit to teach and this is the last bastion they can find a refuge. The 80/20% rule apply there too.

Trying to grade grub by buying term papers is a little like drug cheating in sport where it is obvious who does it by looking at who suddenly improves. But while drug cheats can use masking agents and not get tested, a lecturer can simply interview someone about their project and see if they are full of shit. And when I was in undergrad about 15 years ago some lecturers already had plagiarism software in place.

I just can't see how such a website would have any impact on the quant market at all. Even if it was just used as an additional research resource, it is a drop in the ocean compared to what else was already out there when I was in college.

Also don't forget that to become a quant you will have to answer a battery of technical questions in interview. Some firms just give a chat interview and you can simply say "yep, did Monte Carlo" but even then they will ask questions like why certain distributions are used and what you think the biggest challenge is in pricing.

I think what has had much more of an impact is that people are more aware of the role. When I was doing my masters I had math and actuary undergrads asking me if my masters was "an accountancy thing". Nowadays every Tom, Dick and Harry is trying to be a quant, so even if you had the availability of quant jobs we had in 2003 chances are it would still be hard to get in.
 
Cheating in academia is as rampant as doping in sports. Know enough LSE finance jocks who had outsourced courseworks & dissertations and still mananged to land up snippy job offers at top funds. When the teaching deviates from good ole' socratic method to Power Point, Grade/exam gate keeping this is bound to happen.

Also you are overstating the integrity of lecturers, some of them are just unfit to teach and this is the last bastion they can find a refuge. The 80/20% rule apply there too.

There's more. Even without this overt cheating, the structure of formal education militates against a mastery of course material. The lectures have the nature of a forced march -- there's little time to admire the scenery or to grok the material in fullness. Often the instructors themselves are oblivious to the finer and subtler points, and these can often only be found in older texts. Spengler was right about the decline of the West.
 
present to me the socratic method to derive the BS PDE, implement a finite-difference method, calibrate from market data, or simulate anything.
 
There's more. Even without this overt cheating, the structure of formal education militates against a mastery of course material. The lectures have the nature of a forced march -- there's little time to admire the scenery or to grok the material in fullness. Often the instructors themselves are oblivious to the finer and subtler points, and these can often only be found in older texts. Spengler was right about the decline of the West.

Methinks that thou art overly pessimistic.
Maybe if you would deign to teach a course as an adjunct, then you could merge the philosophies of Hull and Shakespeare.
Explain to the students that if Monte Carlo does not work, Macbeth will.
 
Let's see whether I can put some real examples of the problems one will face even when they are armed with the most powerful open source machine learning library out there, one that makes them think all problems are solved.

Assumption - You have a library that performs every ML algorithm out there.

1. You pursue StatArb. Out of the 5,000 tradable assets out there, which will you choose?
2. You attempt to backtest them ALL. One pair at all timeframes and parameters take 4 hours to complete. That's 4,999*5,000/2*4 hours = 5,785 years to complete. You realize brute force optimization doesn't work.
3. Your boss asks you for a technical strategy that uses at most 2 parameters. You use ML optimizer without regard to qualitative meaning. Oddly, one of the best indicators is something like sin n*exp m. You're thinking of how you're gonna explain that.
4. With US i/r decision around the corner, you can't seem to find a definitive observation to feed into your markov model to predict the outcome. It can't be EURUSD, maybe inflation, possibly SPX. You resigned to the fact that markets are ultimately driven by humans.
5. You finally need to make markets and you're boggled down by the combinations of buy the ask first, then sell the bid, or was it the other way? Then you're not sure whether you should quote above the lowest ask, or above the second lowest ask?
6. Lastly, there are multiple academic sources that say in-sample / out-sample testing does not work. You see a ML optimized backtested sharpe of 6. You execute the strategy and in 3 months, your PnL is filled with losses. Probing the data too deeply may not be the best idea.

My conclusion is there still needs to be human discretion on which and how ML algorithms are used. And this human discretion is nutured through experience, not by having the best ML library out there.
 
Nowadays every Tom, Dick and Harry is trying to be a quant, so even if you had the availability of quant jobs we had in 2003 chances are it would still be hard to get in.

No, nowadays every Qing, Peng, and Chidamberabad is trying to be a quant... if you're strong at math/programming, speak English, and are able to display a genuine interest in the financial markets, you'd be surprised how many good jobs are available at smaller shops that don't want to have to deal with visa sponsorship and language barriers.
 
The criticism was directed against all declarative form of teaching. MFE is a collection of cook book techniques forced in a hurried manner and border a form of apprenticeship despite relying on computation and inference as a teaching methodology. The general criticism is the tunnel view of the teachers where brute force mastery over a set of preset tools without any appreciation for the nuances and subtleties that foster a sense of evolutionary intelligence about the subject matter itself.

present to me the socratic method to derive the BS PDE, implement a finite-difference method, calibrate from market data, or simulate anything.
 
Back
Top