Quote: "However, a financial crisis in the Western hemisphere coupled with other macro/socioeconomic factors has lead to a reduction in scientific and engineering budgets and later-stage investment in such technologies."
Translation: The recent worldwide economic meltdown was caused in part by quants' inability to accurately assess investment risks, or to influence decision-making, or both. The fallout was public distrust (justifed or not) of quants and what they do for a living.
The bottom line is that reliable quantitative analysis hinges on the degree to which economics is a science. Economics isn't a science. Any questions?
Blaming quants for the economic crisis is like blaming programmers for the 2001 tech bubble. In both cases, there's a decision maker to whom the quant/programmer reports, an executive who decides which boat the company wants to float. The engineer's role is just to inform his/her manager that, "Uh, boss, this boat will probably sink. We shouldn't be claiming otherwise". The manager's response in both cases was to say, "Shut up and do your job, we'll go to IPO with just a business plan / we'll sell so many CDOs just the commission is going to make us millionaires. Fuck consequences". If you want to blame someone, I believe you need to look at the executives who took the decisions. This meme of "the quants did it" is a bit ridiculous, and it irritates me because it successfully removes attention from the real culprits.
Be honest : you have never worked on Wall Street. This is just some opinion that you picked up from your favourite newspaper or conspiracy web site.
> Blaming quants for the economic crisis is like blaming programmers for the 2001 tech bubble.
I should have blamed trust in quantitative analysis. That would have been more fair, and it was the point I was trying to make. My shorthand way of saying it may have given the impression that I was speaking of the analysts, not the idea of analysis.
> Be honest : you have never worked on Wall Street.
Be honest -- where I have worked isn't the topic. And if I had worked on Wall Street, that would represent a much better justification for doubting the merit of my views.
A science is a method, not a domain or result or anything else. I.e. one can approach economics or medicine or traveling salesman problem using the scientific method, or one can approach it using woodoo.
> A science is a method, not a domain or result or anything else.
No, this is false. Science is defined more clearly than this. Science is not a method, it is a set of rules for creating a method. Primary among those rules are a requirement for empirical evidence in support of scientific theories, and falsifiability -- meaning scientific ideas must be testable against reality, and those that fail the test must be discarded.
TL;DR: Science is more principle than method.
> ... one can approach economics or medicine or traveling salesman problem using the scientific method ...
By that reasoning, astrology is a science, because it can be studied using the scientific method. But this is not how science works. For X to become a science, scientific study must support the theories that define X. Explanation here:
>By that reasoning, astrology is a science, because it can be studied using the scientific method.
exactly, one can do pure scientific study of correlation between that star position and people's lives and most probably would find zero correlation - it would be a valid (though practically useless given current state of science in related areas) scientific study in the domain of astrology. That domain happens to be full of non-scientific results and statements while has no useful non-trivial scientific results known. Thus people label the domain as 'non-science'. Economics has a lot of both.
The problem with rejecting economics' claim to be a science on the basis its underlying assumptions survive even as models based on them are usually out by a few percentage points (and under certain circumstances give counterintuitive results) is that you'd probably have to throw a lot of natural science with the bathwater. Physics is the exception rather than the rule for the accuracy and robustness of falsifiable predictionswhilst medics - like economists - are left with statistical tests, ad hoc explanations of confounding factors and a raft of exceptions that arguably falsify claims as uncontroversial as "drinking in excess is bad for your health". On the other hand, unlike astrology or homeopathy neither discipline is afraid of testing its assertions against a null hypothesis.
As I understand it there are several 'schools' of economics which use different models and which council different responses to the recent financial crisis - sometimes completely opposite responses.
I'd be more inclined to consider economics a science if, of the multiple contradictory models, all but one had been falsified.
Personally I don't consider social sciences to be real sciences for the same reason. If you want to say economics is a social science I would agree with that.
> The problem with rejecting economics' claim to be a science on the basis its underlying assumptions survive even as models based on them are usually out by a few percentage points (and under certain circumstances give counterintuitive results) is that you'd probably have to throw a lot of natural science with the bathwater.
Actually, it's a bit simpler than that. Sciences, and scientific theories, can be easily identified by the property of falsifiability. If a theory makes coherent, testable predictions, and if the predictions can be tested empirically, then there's a basis for falsification. Economics doesn't have this property. Many natural sciences do. For example, anything having to do with evolution -- very specific, very testable, and excellent at resisting falsification. Not that surviving falsification is required to call something science -- failed theories and fields are scientific in a sense, certainly more so than those fields that never test their theories.
> ... whilst medics - like economists - are left with statistical tests, ad hoc explanations of confounding factors and a raft of exceptions that arguably falsify claims as uncontroversial as "drinking in excess is bad for your health".
Yes, but those aren't really falsifiable, are they? One can argue that a failed test like that is equivocal, not a basis for casting out the underlying theory. In a formal sense, that prevents their being called sciences.
I have to say about modern medicine that it's rapidly moving toward being evidence-based. Many clinical practices from years past have failed more rigorous efficacy trials and have been cast out as a result. So I would call it science on the basis that its theories and practices are open to falsification.
> If a theory makes coherent, testable predictions, and if the predictions can be tested empirically, then there's a basis for falsification. Economics doesn't have this property.
The subject matter of economics certainly involves the kind of observable quantifiable results that admit of falsifiable hypotheses.
Its true that many of the interesting hypotheses offered in the field are falsfiable in principle but faced with significant pragmatic difficulties in conducting unambiguous tests which leads to the practical tests frequently providing ambiguous results, and that many of the practical-and-unambiguous falsification tests are unsatisfying (in that, while they could unambiguously falsify the hypotheses they are offered as tests for, they aren't differential tests wherein the result that fails to falsify one interesting hypothesis necessary does falsify a competing interesting hypothesis.)
This doesn't make economics "not a science". It makes it a difficult science.
(The fact that the questions in the domain are often important to public policy and other decisions in which lots of people with lots of resources have vested interest also results in a lot of things in the field being oversold to decision-makers, but that's equally true in similarly salient physical science fields that don't face the same kinds of testing difficulties, and is irrelevant to whether the field is a scientific one.)
Translation: The recent worldwide economic meltdown was caused in part by quants' inability to accurately assess investment risks, or to influence decision-making, or both.
Part of the problem is the science has a feedback effect; theories based on observations change actors' behaviours, which in turn breaks the theories. So macroeconomics is not on very solid ground as a science. It would work better if it were performed by aliens.
The bottom line is that reliable quantitative analysis hinges on the degree to which economics is a science. Economics isn't a science.
I think you're confusing macroeconomics with economics.
"A great way to get into such a fund is to apply as a software developer, with aspirations of becoming a portfolio manager"
That's not a path well traveled, especially at banks. At good hedge funds roles are well delineated, and chances are that you'll be working hard enough at your day job that you'll have little chance to explore on your own. Policy may prevent you access to the interesting data sets (execution data and historical P&L is sensitive to reverse engineering, and market data is often subject to expensive licensing). For banks the issue is far more bureaucratic - different teams with different hiring budgets and lines of business yada yada. There are relatively few managers high enough to wall cross between back office (technology) and front office (trading) so there's a stochastic component to your success at moving internally based on what group you end up in and how well connected your MD is. It doesn't help that coders are usually the limiting reagent (and paid less than traders) so the firm's interests aren't necessarily aligned with your own.
If you're passionate about trading and coding, HFT is a great way to go. Money makers come from all backgrounds (grad/undergrad/super technical/creative coder) and you'll likely have better control over your coding/research ratio.
I think I may have suffered somewhat from selection bias when writing that, in that a lot of my friends/colleagues work at smaller shops, where they initially started as technology infrastructure developers and worked very hard to move over to the research side.
For instance, these were the guys -running- the data infrastructure so they were looking at it all day, every day. After a while it was probably straightforward to test out intuition on patterns they may have seen.
Thanks for pointing out that the difficulties in doing so at a larger firm.
Two Sigma (where I work) is hiring for a variety of positions, not just Math/Physics types but security, systems, UI, networking, etc. anyone with a serious love of and talent for technology should consider it. i left google for two sigma and am loving it, especially the people and culture. feel free to ask me about it either in this forum or by email.
I did a year at DESCO in Systems (didn't go so well, I made some mistakes) and everything that I hear about Two Sigma is that the place is awesome. The number of postings you've had recently has _exploded_ since I last looked at you guys around 2009-10.
I would definitely recommend people look into this place and might consider it myself if I thought they'd ever hire me (not likely).
I was surprised to see an investor relations page. Most prop groups aren't actively looking for more money. And they talk about clients, which is unusual. Can you explain?
I understand if you can't answer this, but how does the pay for software engineers at Two Sigma compare to pay at top tech companies, like Google? Glassdoor is ambiguous.
I got a raise, but salary is only part of compensation. In the end it will depend on how the stock does (at a tech company) or what your bonus is (in finance).
I've read elsewhere that unless you're in a role that directly generates revenue (traders, etc.), you shouldn't rely on bonuses in finance, and should instead maximize the cash portion of your compensation. I assume this would be true for software engineers at Two Sigma too.
Two Sigma, more than most other companies (financial or otherwise), understands that innovation and success is largely driven by technology. The content of twosigma.com is designed to communicate this, and it's absolutely true. Anything you've heard about how other companies are structured is unlikely to apply here.
I may have sent my resume to Two Sigma a couple years ago. If I did, I never got a response. I was really excited when I found out you had a Houston office, but somewhat deflated when your careers page put all the interesting jobs in NYC. I get the impression Houston is basically just an IT/infrastructure office. Is that the case?
The Houston office is mostly infrastructure and data analysis, but we're always looking for good people more than specific roles. Most of the company is here at the HQ in SoHo.
The article makes a good point about MFE degrees. They're a dime a dozen these days and most of them can't answer simple interview questions about linear regression. Its pretty shameful, not so much on the student's part, but on the universities that graduate these kids fully unprepared for the jobs that they think they can get...
The article really buried the lead on this one. The bottom line is that it's flat-out crazy for anyone to get an MFE these days. They cost as much as an MBA and won't get you a job as a quant.
Indeed, I am considering actually writing an entire article on that whole topic.
I haven't done an MFE personally, so I don't feel I can comment too much on what they're like, although I have a few friends who have. A lot of them simply went into research afterwards.
Another solution is to write the algorithm but avoid the hedge fund - lots of suits, and they take most of the money. You're better off if you trade it yourself.
People work for hedge funds because hedge funds provide mentorship and really powerful tools and lots of data to sift through. We're trying to provide all of those things, for free, at Quantopian. https://www.quantopian.com/ Check out our community (for mentorship), our backtester (very powerful, and open source), and our 11-years of minute-bar data - all for free.
Sometimes I think of the HFT and quant trading industries are like the Formula 1 of tech. A way for extremely talented people to apply their skills and compete with each other, with lots of money involved, but ultimately the technologies developed for the 'race track' make their way onto 'main street' in some shape or form.
I'm sure all this research into ultra low latency infrastructure and live data mining is bound to come in useful somewhere else. Maybe this is off base but why be cynical.
Quantitative trading seems interesting because I would hazard that you need strong math background along with solid hacking skills.
There are other problem domains with similar requirements, but I'm quite curious to hear stories from other HN'ers who have tackled quantitative trading either for themselves and in the day job. Specifically, if you can share a bit on your technical work in such a role, that would be cool to hear about.
I was a quantitative developer working alongside a quantitative trader in a small 'proprietary trading fund'. This is roughly how all quant funds are separated. There is the "technology" side, which builds the data/trading infrastructure and then there is the "research" side that generates the trading strategies to run on the infrastructure.
My job involved anything from hooking up to brokerage APIs to optimising MySQL replication topologies. Quite varied!
In a way, it wasn't too different from your average startup, with the possible exception that you deal with a non-trivial amount of data from day #1 (hundreds of millions of rows are not uncommon).
Open source has gained significant ground in funds these days. Python/R are now the "default" go to languages for quant trading research, with some MatLab too. Libraries such as NumPy, SciPy and pandas have really brought 'algo trading' to the 'retail' (algo) sector as well.
.NET is still generally used quite a lot in investment banking, particularly C# for front-office GUI code, and C++ for any legacy number crunching libraries.
I worked at an HFT for several years as a developer on the quant lib team. I'd say the algorithm side is the hard and interesting part, while the math is relatively straightforward and solved. For example, it's well known how to price an option - but doing that efficiently and at large scale is the challenging aspect (think kd-trees, etc. for sophisticated caching, messaging protocols, dbs for organizing data). I also built a sophisticated graph algo for relating securities, which built the core of one of our most successful strategies.
So technically, I found it very appealing. However, my experience wasn't so great otherwise. You really get the sense that people are only in it for the money - everything revolves around the year end bonus. And at least where I worked, this kind of huge money drove a lot of really nasty politics.
> Quantitative trading seems interesting because I would hazard that you need strong math background along with solid hacking skills.
Strong maths knowledge can be good for general developer roles anyway. It's one (small) reason why I did a maths degree part time (Open University in the UK) to complement my existing Comp Sci degree.
I'm sure you could get a better bang per buck picking and choosing specific areas rather than an entire degree but that wasn't my motivation (I'm equally interested in both in general).
A good number theory course (even just plowing through Crandall and Pomerance, or Apostol) will definitely help with understanding/analysing asymmetric encryption.
While you may not be solving partial differential equations in your average tech startup, there are plenty of instances where a maths degree can be directly applicable. Statistics is one instance, for A/B testing. Another example is the use of vector calculus in machine learning and "data science".
It was good. I took it almost as slowly as you can (typically one module per year) so it was 8 years from start to finish. In that time I got married, moved house twice and became a father so I wanted to avoid it taking over my life. Tutorials were local to home or work (I was lucky in that respect) but they did seem infrequent.
Tricky to recommend now the fees have quadrupled though (I paid about £4k in total for my degree, it'd be closer to £15k now); unless it's your first degree and you're considering the OU over a traditional university; then £5k a year is quite cheap as you can be much more flexible with life and (part/full-time) work.
I was considering either continuing with the OU with Maths on some of the Master's courses (those fees aren't subsidised in the same way that UG courses are), or switching to languages (French, German, Spanish) but the prices put me off those. For now I'll have a year or two off as a break.
I agree, the OU Maths degree is excellent, but it's been ruined by the new pricing arrangements. None of the really interesting people I met on the course would have been able to afford the new fees.
I've been toying with the idea of doing an OU maths degree. Not for the credentials but because I'd like to be good at maths... I hate not being able to understand it well. However I'm dismayed to hear the cost is now £15k.
Sorry, somewhat OT but wanted to chip in on this as I'm actually into my 3rd year of a (hopefully) six year BSc in Mathematics. I find the course material really well presented. I've been taking it far more seriously and enjoying it far more than my very first degree (which I had a hard time of). The mathematics department is considered somewhat 'backward' compared to the others as they're one of the few that mandate the postal submission of 'TMAs' (tutor marked assessments).
You are compelled to do certain computer parts in Mathcad, which isn't too bad if you use Latex. I tend to develop and study using Scipy however and then do final bits in whatever software I need to. I've put up some code for my own use here https://github.com/jaymzcd/oucode/.
Totally agree with the other comment around fees - I'm "locked in" to transitional arrangements meaning my total degree cost will be approx ~£5k. Had that stayed for all I'd have no qualms recommending it; now it's quite a bit more expensive I think you need to weigh up what you expect from it. Still, I think it's been a great decision to go for it; it's hard work combined with a demanding full time job but the glow of the mental accomplishment is great.
I don't exactly need to know how to solve differential equations to build a Facebook tab but it keeps me constantly thinking of new ways to do even simple things. I did use some of the knowledge of t-tests and sampling from my stats course to quantitatively report whether a stats rise after we launched a competition was the cause or if it could be considered random. Not exactly required but it did read and sound impressive on the slides we sent the client.
As an added bonus I also save ~ 30% on my TFL London travel card as you qualify for student support even with the OU; which actually, over a year, accounts for around 60% of my actual module cost!
A good number theory course (even just plowing through Crandall and Pomerance, or Apostol) will definitely help with understanding/analysing asymmetric encryption
I was hoping you would also recommend G.H Hardy's Theory of Numbers
tom_b, basically every shop cooks their own solution for just about everything. if you do mid- to low-frequency a Bloomberg will suffice. if you do high-frequency you are in the infrastructure game. you co-locate to the exchanges and battle for the milli- to microseconds. there are a few specialized databases in this area for dealing with TBs or rows.
everything else in the area of finance is a complete waste of time IMO. basically the financial system is toast. I hope that bitcoin will improve some of the absurd structures we have set up.
anyway, if you want to know some specifically you can go to nuclearphynance. There are some cracks around there. good traders would consider an MFE as the opposite of helpful, I'm sure. standard stochastic calculus is pretty much useless.
There are fully managed solutions to receiving market data etc, that a lot of companies do use, both in the enterprise, and HFT realms (ie tickerplants).
Yeah, for 100k$ or more a year. So you need a lot of capital to get started on your won. Getting started with a private account is hard to impossible in my experience. Lime Brokerage offers good stuff for HFT. You need about 12 months to get really going.
What is the difference between strategies used in mid/low frequency vs high-frequency?
I understand high-frequency is mostly stat arb but I haven't heard of too many quant funds that do mid/low frequency. I've seen one present before that used small changes in portfolio optimization to get an alpha of ~0.2% over the benchmark but that didn't seem too useful.
Hi, I worked in both roles. I did an undergrad in theoretical Math and MFE. Its particularly easy to get a risk management job. My first job was for 100k + bonus out of school for BNPP. It is a highly quantitative risk department and was one of the best (if not THE best) institutional risk management gigs on the street. I did not know this at them time but I started in 2007 and after 2008-2009 I came to this conclusion.
Anyway back to main subject. Risk jobs are easy to land and you can learn TONS or nothing interesting. Every shop has Risk integrated at different levels. Some shops its a backoffice type job where the risk guys are in the back seat (and this is the dominant case actually 8/10 I'd say). And rarely in places like BNP will you sit on the trading floor and have more power than traders. Yes traders will actually fear you... Rather than vice versa where Traders that are making money are big swinging dicks doing whatever they want (this is still the case at most shops).
I started quant trading by chance when some guys running an arb blackbox were started constantly loosing money and were working late hours. I gave them a hand fixed a few bugs and word quickly spread. Next opening they offered me a position and I did an internal transfer...
After I went to managing a portfolio (2 years after). And successfully did that for 4 more years.
The summary is this. Quant Trading roles actually come in tones of flavors.
Banks have option desks and in the recruiters will call these "quant trading roles". Hell, for all they know there is complex math involved and 'quant' sounds sophisticated. Truth is all option traders are proficient and calculus and derivatives to the point good traders are able to calculate pretty accurate 2nd orders in their head. And the tools available are also vast.
But this is not a 'quant trading role'.
A true 'quant' role is what some institutions have in house 'quant' teams that manage all the pricing models and libs.
A 'true' quant trading role is stat-arb or some other arb strategy desk or similar type of strategy. Most banks do not have them anymore so its hedgefunds only. But even these vary vastly in quality. I know desks that would make 2-3 million a day net with very shitty hardware and patchwork system with lowlevel to mediocore guys working on it. And brilliant guys whom struggled (talking tier 1 guys, PHd MIT, princeton).
So its a strange game and I would recommend going into a trading role into a medium level shop. Whats important is a good team and having a good vibe with the people you will work with.
Trading is high stress and high emotion business so for your sanity in the long term its more important to work with good people rather than a top tier shop in the beginning.
Tech types are getting a little arrogant, out of touch, self-absorbed and judgmental these days too. Getting people to click on ads or privatizing all their personal info in an insecure social network ain't God's work any more than taking companies public, developing the infrastructure for pension funds etc. to invest. And if you'd rather subvert the Constitution than say, help people allocate their savings, or reform the financial system, then your priorities need re-examining.
If working in porn and for the NSA are in the same category of morality for you, you've got problems. Same if you think destroying consitutional/human rights is preferable to making money in a questionable way.
Despite the intense competition, it is still possible to
shine as a candidate, but you must be very well aware of
what the funds are after, and then make sure you are
providing it to them.
I think this is a truism for any role in any capacity in any organization. Just replace "funds" with the organization/manager/client/whatever name.
RenTech assembled some of the brightest technical minds in the world to trade purely quantitively. They did pretty well for themselves.
Jim Simons, famous mathematician, former MIT professor and DIA code cracker, earned the following amounts according to Forbes:
2011: $2.1 Billion
2010: $2.5 billion
2009: $2.8 billion
2008: $1.3 billion
2007: $1.5 billion
I applied for a large quant hedge fund out of university (Man Group AHL) with a masters in civil engineering and got to the final round. The other 5 candidates all had MFE degrees and seemed to struggle with the technical interviews a lot more than I did. It seemed like the fund was more interested in my understanding of statistics and general mathematical aptitude than specific knowledge of derivatives pricing etc, the same was true of my interviews at Jane Street.
I would say it's easier to get a job in risk management (likely in investment banking) than quant trading coming out of an MFE degree. This is primarily a consequence of how the MFE programs are set up, what they teach and the network of the professors etc.
So quant funds don't use that much calculus? Primarily statistical techniques?
I would guess that might be the case since pure arbitrage no longer works and thus quant funds have to delve into more probabilistic strategies like stat arb.
The quant derivatives pricing teams at banks are where the stochastic calculus folk tend to head to. Their teams are generally highly respected in this area. Also, banks are doing a different job to funds. Banks are generally interested in assessing the risk or trading risk of these products, either on prop (i.e. with their own funds) or to clients.
Funds tend to concentrate more on statistical/machine learning/econometrics research approaches. The culture is generally more like a research institute thank a bank. They tend to hire more PhDs from Comp Sci, whereas banks will hire directly after MFE or straight out of undergrad.
Is that why physicists (vs computer scientists) are no longer recruited as actively for quant trading roles?
It seems like all the low hanging fruit has been arbitraged away. I've worked with a Math PhD from Princeton before in prop trading and he always lost money.
Machine learning techniques are becoming more common. Hence a shifting trend towards CompSci away from Physicists. The latter were often hired due to their modelling/probability capabilities in PDEs for derivatives pricing.
Also CompSci comes with a (perceived) "built in" ability to carry out good software development practices.
Translation: The recent worldwide economic meltdown was caused in part by quants' inability to accurately assess investment risks, or to influence decision-making, or both. The fallout was public distrust (justifed or not) of quants and what they do for a living.
The bottom line is that reliable quantitative analysis hinges on the degree to which economics is a science. Economics isn't a science. Any questions?