I liked this article because the roll of the programmer on wall street has been going down recently from my perspective.
Programmers used to work along side quants and were well respected. This was generally because practicing quants didn't used to be able to program. This was especially true in the 80s and 90's where programmers and quants where almost valued identically as one required the other to make things happen.
Somewhere around mid 2000's most quants started to be able to program pretty well. Not necessarily with C++, though lots (especially former physics) could use C++, but certainly with mat-lab, etc. This really shrunk the required programmer head count,
This, coupled with the fact that few programmers were trained in fields like stochastic calc, and the advent of offshoring meant that programming started to be come a lower status job around this time.
Some relatively small fields like HFT kept good programmers in need, but these jobs are relatively few and far between compared to the sell side bank jobs.
Now, programmers are much more valuable/respected once again, due to the requirement for top notch software, not even as a competitive advantage, but just table stakes to compete in the markets.
And like most fields programming in finance has really started to specialize:
- lockless programming, low level Linux, user mode network drivers, etc for HFT
- distributed systems, low latency messaging for quote dissemination in large systems like what most sell side banks run.
- real time risk limits, web UIs for helping to visualize and alert portfolio risk
- "production" modeling of quantitative models, for those that can span the programming/math boundary
> This, coupled with the fact that few programmers were trained in fields like stochastic calc,
Anecdotally though, I've met plenty of coders who did this to themselves by refusing to learn more about the domain or the business.
They basically silo'd themselves, then wondered why their colleagues didn't treat them as an integral part of the team when they couldn't speak the language or understand the concerns.
Taught myself enough stochastic calc to change and validate pricing models while on the job.
Coders who understand their domain well are immensely valuable, and are easily worth at least 3x, because they don't over- or under-engineer a solution, and because they can conjure up with technological solutions for problems that nobody else in the company can even conceive of.
> Coders who understand their domain well are immensely valuable, and are easily worth at least 3x, because they don't over- or under-engineer a solution, and because they can conjure up with technological solutions for problems that nobody else in the company can even conceive of.
This is a crucial point to me. Being able to code is good, but being able to code well while also having an intermediate or expert understanding of the domain that you are working within is what is going to set you apart in the long run.
In the industry, developing software can usually be seen as the process of creating a tool that will help to better solve a problem. It seems strange to me that you would want someone who doesn't have a solid understanding of the nature of the problem (or the solution) to be designing and building the tool, just because they claim to be good tool-makers.
> It seems strange to me that you would want someone who doesn't have a solid understanding of the nature of the problem (or the solution) to be designing and building the tool, just because they claim to be good tool-makers.
I couldn't agree more and this is my main objection to "AI will replace [insert any job here]".
A lot of problems are inaccessible to programmers just because not enough of them seem to be willing to understand them. Complex rules about accounting in large companies seems to be one of them for example. This is why I think AI will certainly overtake many professions but mostly professions that are required on a large scale and that are easy to understand: driving, cleaning a house, building a house, ironing, cooking, assembling objects, etc. For complex accounting or legal issues, I am a lot more skeptical. You don't need complex AI to run accounting systems, they can easily be done with regular software. However in large companies you still have huge accounting departments booking and maintaining these complex logics by hand (excel), a lot could be automated now. And the same reason why they haven't been automated in C# now will apply to AI in the future.
Re your AI thesis, I think it’s the exact opposite. Deterministic, rule-based processing applied to lots of similar data are perfect for AI-fication. It’s all in the computer, and you can kick out to a human-driven exception process when required.
All that physical, real-world stuff you mentioned (cleaning, cooking etc) is vastly more complicated, riddled with experience-driven subjective judgements (is this paper on the floor trash or should I put it back on the desk?) and highly situational. Extremely difficult to generalize and handle with an AI. Not to forget it’s messy stuff in the real world, with engineering and hardware and grease and failures.
Most of the complex accounting and legal questions that these departments are doing manually aren't completely trivial either (we are not talking about just adding assets). But the determination of whether the paper on the floor is trash or not is something that pretty much any AI professional can apply his mind on (and will likely do given the size of the market), whereas very few will apply their mind on how IFRS should treat a tax asset in consolidation when the corresponding tax treatment under USGAAP is a still in discussion.
If you look at self driving cars, they are already doing the sort of wizardry of your trash example when analysing the other objects in sight.
And again it's not so much a matter of technical complexity, it's a matter of the domain being inaccessible to those who can best automate it.
>Most of the complex accounting and legal questions that these departments are doing manually aren't completely trivial either (we are not talking about just adding assets).
These are exactly the types of problems that machine learning is perfect for. Problems which involve chains of nebulous decisions that would be difficult to program traditionally.
Couldn't agree more. I have met so many programmers in banking who seem to actively avoid having to learn anything about a bank, not even complex math.
In theory, project managers are supposed to bridge the gap between the business and programmers, however my experience is that they usually understand neither the business nor the IT side, and are only adding more confusion.
People with both knowledge of both worlds are incredibly rare and valuable.
All but a few programmers I have met have tried to understand the underlying reason for their employment and the businesses issues they are employed to solve.
This approach have caused programmers to loose a lot of influence and status.
10 years ago companies had developers develop their own customized CRMs, which can really benefit a company if their domain-specific needs and processes can be integrated to such core-business software.
Nowadays, Adobe Campaign have a cake walk selling their bloated, general and non-tailored “CRM” solution to business.
How can they Adobe etc do this? They approach the marketing and sales team and offer a nice slide show.
Why does not IT or programmers even get involved before the purchase? Because no one likes talking to the idiots in IT that does not understand the businesses needs. They only nag about some “technical debt” (that no matter how much money is poured on the teams never disappears) and the need for re-writes of fully functioning parts of the system that introduces new bugs.
>Anecdotally though, I've met plenty of coders who did this to themselves by refusing to learn more about the domain or the business.
Anything outside of learning enough to do your job should be at the business's expense.
1) some fields are not good for self-teaching. It leaves the possibility open to misunderstanding and writing code to that misunderstanding
2) the business will probably have a good idea of conferences and workshops that impart the knowledge the business is looking for
3) this is more fuel for "unpaid studying outside of work hours"
Based on posts like this, it seems really dumb to enter software first -- better to get a degree in some other technical field and just teach yourself to code.
> Anything outside of learning enough to do your job should be at the business's expense.
There's no "should" here — it's an open market with a lot of agents. You're free to pursue certain working conditions and then look for employer that will agree with them. And another developer is also free to go an extra mile and ask something extra in return.
This is why I'm against any kind of unionizing for developers, by the way — too many different situations and opinions about what our working conditions should be.
It'd add even more value if companies spent more on proper training. Maybe even training geared for software devs learning the field from that perspective if that's available.
But if they think I can become a domain expert by reading junk on the internet with no one really verifying my skills, I'm happy to take their paycheck.
I apologize for the off the wall comment. Every time I read about wall street programming, I've got this subliminal voice in the back of my head wondering how anyone could find the motivation to do this work. I'm obviously revealing my biases against the financial industry here. Also, I'm unable to make myself do work if I don't see it contributing to some greater vision or cause. Are there people out there who love finance and learn these types of specializations on the job, or are they hired away from other industries lured by big comp packages?
I guess it's just a case of lack of empathy on my part. I just can't understand why anyone would want to put so much effort and skill into making bots for farming money for a bunch of questionable people.
I don't feel too guilty for saying this. Every time a story about bitcoin is posted to HN, people go off topic and start complaining about everything that is wrong with bitcoin. I don't see the same when we talk about finance in general, even though similar sentiments apply, maybe more so.
I think it helps to understand that the financial sector provides capital that serves to energize virtually every other sector in every industry.
Without finance, there is no biomedical research (private especially). Without finance, there is no aerospace advancement. Without finance, there is no alternative energy. Being able to facilitate a more liquid, more transparent financial sector is, in my opinion, a calling worthy of any programmer who seeks to make the world a better place.
Your assignment of people as 'questionable' is likely an artifact of the 2007 recession, and that kind of thing is a personal choice. I can guarantee you that there are 'questionable' people in every sector and in virtually every company. Sometimes people are great, and sometimes they aren't. Their field of employment has very little to do with what makes them that way, and I think that it's a little ignorant to be so casual (and I say this because you are certainly not alone in your judgment) with how you view people.
Remember how, as children, we are taught to not judge books by their covers? Why do the same with a programmer who works in finance? Sure, some of the work may be a little pedantic, like squeezing an extra millisecond out of an algorithm, but hey, that's computer science in general, and I liken that to an F1 racing team spending hour upon hour sculpting the perfect frame for their car.
Sorry for the rant, I've spent a decent amount of time working on financial algorithms, and what you said touched a nerve.
I stumbled into finance a few years ago and I’ve had overall the best bosses and work env ofanyone I know. I’m also doing a more audacious program of research engineering/ computer science than I could likely do anywhere else.
I like to sometimes describe my work as some combination of “making the systems that fund aerospace at least as reliable as commercial aircraft”, as transparent as intergalactic vacuum, and easy to use to boot! Or at least that’s the idea :)
On the hft front I recently started at poking at how to get accurate time on a dev computer, which gets fiddly the moment you want really interesting accuracy on commodity hardware :)
Thanks for the praise, and for the insight per your experience. Sounds like a great gig!
I looked you up, and see that you are likely working exclusively in Haskell, but thought you might derive some value from Carl Cook's presentation on optimizing HFT code at the most recent cpp convention. I recognize the languages are distinctly different, but he provides some interesting theoretical points along the way that might be advantageous to you.
Also, I can sympathize with your plight to run scalable sims. I've done some work in bioinformatics, and am building a small cluster at home so that I can learn to write code that'll scale to more massively parallel systems that are de rigueur in that domain.
I’ll have a gander! Thx for the link. I do code in other stuff when it makes sense ;)
Yeah building stuff that works easily in the small and sanely in the large is a fun challenge. Also hard.
Possibly because of your original comment, or perhaps unrelatedly, I’ve been lately saying “finance done right is the lock free wait free scheduling algorithm for moving society’s resources around” —— all the other stuff folks think of as finance is really just icing and fancy wrapping on top of that core truth
As someone who works (hard) in the part of finance you're describing, it's a mix of a few things. From a purely intellectual standpoint, it's a very hard, very interesting quantitative/modeling problem to solve (ie. predict the price N seconds in the future is non-trivial). It's an extremely competitive field (you vs. other teams like you in the world competing for a slice of the pie), and you get results back from your experiments very quickly, so you know exactly where you stand. And finally, to your following point --
>Also, I'm unable to make myself do work if I don't see it contributing to some greater vision or cause
This simply is not something I experience. I've never felt any particular satisfaction from that particular aspect of a job, it's just something that never crosses my mind and incentive me in any way.
There's actually quite a lot of puzzle solving involved that makes it interesting. I think there are a lot of misconceptions of the industry at large and it suffers from bad PR after the global financial crisis. It's an interesting software problem to try and write algorithms that can earn alpha regardless of what is happening in the market. If anything computers have made our markets more efficient and it is great for price discovery.
It may not be the same as creating a product or service that people love or solves a need, but it is an interesting problem to tackle.
See that's the thing. I don't know what "earning alpha" means and why I should care, and working on "price discovery", which is also something I am unaware of, sounds more like a punishment than a reward.
alpha is a return in excess of what is expected. If you invest in a low cost index fund you expect alpha of 0 as compared to the index the fund is mirroring. If instead you put money in a hedge fund, alpha is the difference (positive or negative) between the returns the fund gets vs some benchmark like the S&P 500. Negative alpha is bad, positive alpha is good. Earning alpha means you are beating "the market".
There's also beta, which despite being the next greek letter, isn't the same at all! Beta is a measure of the price correlation between an asset and some benchmark in terms of volatility. Beta of 1 means that the asset moves in lockstep with the benchmark. Low beta means the asset doesn't move as much as the benchmark and high beta means the asset moves a lot compared to the benchmark. Tech IPOs are often high beta compared to the S&P 500.
Anyway, finance has a lot of intellectually challenging problems that happen to overlap well with people trained to build mathematical models that attempt to describe stochastic processes. CS, Physics, Stats, some branches of pure Mathematics -- all have direct applications in the work they do to some aspect of being able to understand and predict markets.
Price discovery means that anyone can find a fair price for some asset. For example, algorithms that provide liquidity in agriculture products ensure that farmers can easily liquidate their inventory, as well as hedging their risk through derivatives.
I wouldn’t close your mind to it. Finance attracts “questionable people” but finance is also very interesting on many levels and is a central part of life. Work through some finance math. Calculate an IRR. See how it works. If anything there’s a lot of great human history in it.
Thanks. I'm trying to be diplomatic here with my criticism and appreciate the positive and informative responses despite my negativity. You are right, finance is critical to social function and perhaps I've only been focusing on the negative aspects.
Finance is a vast world and unfortunately too often reduced to pure speculation.
First you have basic lending and borrowing activities. The bank can afford to pay you interest (well not now that interest rates are at zero) only because your money is invested, usually in loans.
Then you have payments, cash management for companies.
Then you have more complex services offered to larger companies, raising capital (equity and debt) on the market, helping them manage their risk with derivatives (interest rate, currency, commodities (oil, metals, livestock, etc prices), inflation, etc) plus some advisory on corporate actions (mergers, IPOs, etc).
You have insurance, reinsurance.
And then you have asset management, which you can call speculation if you want, but which is essentially people managing investment portfolios on behalf of savers, pension funds, insurance premium for long term risks, etc. With a wide variety of approaches from very conservative money market funds which invest in highly safe and liquid assets, to trackers who try to replicate a benchmark by over performing it by a little, to investments in illiquid assets (property, non listed companies) or very aggressive/leveraged strategies (hedge funds). Fundamentally what they are doing is to ensure your money gets invested wisely (at least they try).
And of course all sorts of intermediaries: consultants, brokers, market data providers, research (sort of financial journalists), prime brokers (who provide financing when you leverage liquid financial assets), system providers, auditors, rating agencies, etc.
Now you could call everything speculation: when a bank lends you money, it bets on your financial soundness. When an investor takes the opposite side of a commodity (say oil) transaction, it bets on the price of oil. But that allowed you to get money to run a business or to offset an oil price risk that could have made you lose money if oil increased (say your are an airline company). There is no real dinstinction between risk taking / risk transfer and speculation. But in most instances it does have a social benefit.
My feeling is that most criticisms of finance are based on a combination of a lack of understanding of what finance is and does, and indignation about how much money some people in the industry get paid. It's any easy target. Absolutely there are 'questionable people' in it, but are programers focused on getting users to click on more ads or spend time on a given app any better? Is tech really saving the world? There is so much more to tech than that, and most people in tech don't feel that's a fair representative of what they do. I think finance is similar, but I don't work in it so maybe I'm wrong...
Cash motivates plenty of people, as does the joy in solving a problem. Whether the value of that solution to society's benefit is positive or negative does not always factor into a solution architect's moral calculus. This is how I explain your observed point to myself when I see folks stoked to work for $largetechco or $adtechco.
This is it for me. Futures trading is a zero sum game (negative sum if you factor in fees), so writing a program to automatically extract value from an ever changing market is a nontrivial and ongoing project. But unlike a game, getting good at it can pay the bills in a real way. I understand that in the big picture long term, I'm not making much of a mark on society, but nevertheless it satisfies my personal and financial needs.
That's why I specifically called out futures trading. From [1]:
> In the financial markets, options and futures are examples of zero-sum games, excluding transaction costs. For every person who gains on a contract, there is a counter-party who loses.
That quote is misleading. He's talking about perceived price. The actual price of an equity is the price at which the execution occurs. And for every execution there are two sides, hence the zero sum game is the only game in town.
You are completely ignoring the enormous value that financial activities provide for society. Efficient capital markets, risk pricing and exchange are fundamental ingredients to an effective economy. Do you think these businesses get paid for nothing?
You’re comfortable enough to be ethical about your capital, for me it would be a big constraint. I’m sure I’ll get to be picky someday but that’s years from now. And at least there’s cool maths :)
Some make a whole lot less than you and still feel ill at ease working for the finance industry when it seems to soak up so much capital and contribute so little to society
Most of the quants and quant devs I work with are out-of-work physicists who want to provide for their families. Most of them would rather be tenured. I work in this industry because I cannot find employment in my field of engineering near my friends.
You wouldn’t be able to type this comment on your computer (and neither would I) were it not for financial markets to organize funding for the company that produces your device. Or you might not have gotten a mortgage to buy your home.
Financial markets serve a hugely important social function of allocating capital (intermediating between people who have money and those who need money).
>farming money for a bunch of questionable people.
The "questionable people" include grandparents, firemen, teachers, etc with pensions and retirement funds.
>I don't see it contributing to some greater vision or cause. [...] I guess it's just a case of lack of empathy on my part.
Your sentiment is common and I think it's caused by people not connecting the dots from grandparents' retirement pension to the hedge fund trader.
As one example, at the micro level, grandparents are depending on a pension.[1] They also don't want their _real_ purchasing power to diminish. The "real" not "nominal" purchasing power is an important distinction because if the price of bread is $2 in 2017 but becomes $4 in 2037, they want to have adequate future money to buy food.
In other words, they want their money to grow and keep up with inflation.
Your grandparents stashing money in a savings account paying less than 1% will not protect their purchasing power from inflation. Likewise, the pension manager that oversees a $1 billion retirement fund to pay retired teachers will declare it bankrupt if it only gets a 1% return by buying Treasury Bills. The pension's assumptions to avoid insolvency might require 8% or better annual returns.
Therefore, all these participants in the economy are looking for better returns. You can't get 8% from savings accounts and t-bills. That's where money managers like private equity, hedge funds, and VCs in Silicon Valley come in. They sell their ability to generate higher returns.
This expertise attracts capital from entities like pension funds.
So basically, a bunch of programmers are fine tuning algorithms because [...connect a bunch of dots...] your grandmother doesn't want to be a starving destitute and live under a bridge in her retirement years. If you multiply millions of everyday people not wanting to be destitute (macro level), you end up with hedge fund managers with billions to play with.
That said, can we "reconfigure society" so that all this competitive activity of shuffling money from one pile to another isn't done? Maybe. It would involve fundamental changes that we all can't agree on. For example, if you stop inflation you can stop the incentive to generate returns that try to beat inflation; but then some would say that "deflation" would lead to people hoarding money and the economy would crash. Inflation is the current winning ideology for the modern economy. It's hard to see how we could alter society such that Wall Street traders go away.
> That said, can we "reconfigure society" so that all this competitive activity of shuffling money from one pile to another isn't done? Maybe.
Your proposals boil down to ending deficit spending. Aside from the authority issue (no single entity has authority to do this, not even congress, definitely not the president), that means the US Federal government has to cut $650 billion dollars in spending.
Or to put it more directly: We currently have pensions, social security, an army (meaning having an army at all). Pick 2. The other one gets destroyed.
I do HFT and think about this a lot. I am essentially a middleman that facilitates trading across:
1. Time - I quote buy and sell prices continuously in Microsoft stock trying to earn the spread between them. You want to buy stock and my offer to sell is the cheapest in the market. We trade. Later, someone else comes along to sell and I have the highest bid, so I buy the stock back from them.
In between I'm exposed to price fluctuations, so sometimes I make money and sometimes I lose, but I make a profit on average by charging a spread between my buy/sell prices and predicting small price movements. The two of you could have met had you waited, but I let you lock in a sure thing by bearing risks you didn't want. Think of this like insurance. Odds are you spend more on car insurance than you're expected to receive in payments, otherwise the insurers would go out of business. But you probably prefer spending $1000 a year vs. spending $0 most years and $30000 once in your life.
2. Place - You're in the US and want to buy Nokia stock. Someone in Finland wants to sell for a cheaper price. You don't have access to the European exchanges and she doesn't have access to US exchanges, but I do. I trade against both your orders and earn a couple cents per share in profit. I helped you meet halfway across the world and also keep prices efficient to reflect global supply and demand.
3. Product - You want to invest in an index ETF that holds 40 different stocks. I quote an offer to sell based on where they're trading in the market plus a small profit margin. You buy my offer. I turn around and buy the stocks to hedge my risk.
I'm better at trading than you are and have very low costs. Even considering my profit margin, it's cheaper and less risky for you to buy the ETF from me than 40 individual stocks. I keep the ETF and stock prices aligned and also helped people who wanted to sell stocks get their orders filled. Win win win.
I think the work I do clearly benefits the market and natural investors. All of these trades are competitive too, so I can only earn razor thin margins per trade, or someone else will undercut me.
I do sometimes question how the market picks winners and losers, typically based on speed. Is someone whose system is hundreds of nanoseconds faster more deserving of arbitrage profits? Do high technology costs naturally lead to consolidation and reduced competition? https://faculty.chicagobooth.edu/eric.budish/research/HFT-Fr... this paper has some interesting thoughts on the topic.
FWIW I do think continuous price-time priority trading is flawed in that it creates an arms race in speed, but it's the least bad option. Other mechanisms like batch auctions are even worse. As a thought experiment, consider the #3 ETF trade I described. If I can't immediately hedge my risk and instead go into a batch auction for each stock, how will that affect the price I offer in the ETF?
PS: I do think it's a fun job. You get to play with cool technology and it's like a game that gets harder every day against very skilled opponents. Even the arms race isn't all bad. HFT firms have effectively bankrolled modern innovations in high speed ethernet, FPGAs, low latency kernel bypass networking, etc. that have knock-on benefits for other applications.
> As a thought experiment, consider the #3 ETF trade I described. If I can't immediately hedge my risk and instead go into a batch auction for each stock, how will that affect the price I offer in the ETF?
My idea for this is that when you rest an order, you should be able to attach a set of hedge orders to it that will initially be inactive, but will fire when the owning order is filled. This would let everyone get hedges done at better-than-HFT speed, and make things like speed bumps or batches a lot more tolerable for market makers.
You'd want some kind of pro-rata, or multiple orders at fill thresholds, so you could show big size and respond sensibly to partial fills. You might want alternative hedges to cope with price moves in the hedge target. You might want some mechanism whereby the exchange will pull the owning quote if the hedge disappears. There's all sorts of fun you could have.
For this to work really well, you'd want the hedges to be processed synchronously with the fill of the owning order, so there's no HFTable window of vulnerability between fill and hedge. That couldn't be done across different matching engines in the typical architecture current exchanges use. For futures, though, it might be workable - i think all the different expiries of a given underlying, and their spreads, are already handled synchronously, so that the exchange can do implied matching. That might not include some of the more exotic constructs which aren't in the implied chain, though.
This sounds similar to OneChronos, though I don't know if they've launched anything: https://www.onechronos.com/
You're right that there's precedent for this:
-Implied engines as you mentioned where traders can atomically express a view on the term structure without legging risk. For the benefit of non-finance readers: Say you believe the fair price difference between Dec and March futures is 2 index points, you could quote in the spread between the contracts (long one, short the other) buying 1.75 selling 2.25 without being fastest to respond to every shift in the index itself. Your order in the spread will create implied orders in the underlying contracts themselves and you can only be filled if both your long and short trade.
-IEX's D-Peg effectively runs a high frequency pricing model inside the matching engine to predict when an adverse price tick is likely.
-I think D.E. Shaw has a patent on a domain-specific language for complex orders where they can be priced off any other instrument(s), ratios between instruments, features of the order book, etc.
The real problem with batch auctions is coordination. So long as you have other exchanges trading correlated assets in continuous time, there will still be a race to adjust orders in the batch auction market after a pricing signal occurs. Even with a random delay or crossing time, it's still beneficial in statistical expectation to be faster.
You'd need to trade most similar products worldwide on a single platform. Since exchanges are a network effect business, this would likely lead to monopoly profits for the exchange operator.
I'm not sure "no high-speed arbitrage/monopoly exchange" is net better than "some high speed arbitrage/competing exchanges." I suppose you could have the government run the exchange as a utility, but that would stifle innovation. A lot of things we now take for granted started in experimental ATS and ECN startup markets.
I don't work in finance, but I see the modern global financial system as one of the greatest achievements of humankind and one of the main reason behind humanity's prosperity. I would love to work there if I had the skills or patience to gain them (instead, I work in gamedev, where I acquired immense domain knowledge over the years).
To add an angle from someone who shares your concerns about the financial industry, the high pay makes it so you can donate large amounts of money to charitable organizations of your choice.
It's actually a recommended route in the effective altruism community for those with the aptitude.
We are not all some hippies who want to help the world. Fuck this shit. I want straight cash! And if you think people at the top of the tech industry dont think like this ur just naive as fuck.
Whats not cool about a bot farming money? Its like the best thing ever looool.
Has offshoring ever been a significant factor in this arena? Given the profoundly proprietary nature of the systems being built, I would think that unlikely. Offshoring development is a good way to find your custom software reskinned and sold to a competitor without much legal recourse.
Also, is 'mainstream' programming that common of a need? I had read (here on a HN comment thread) awhile ago that many organizations use commercial languages that simply aren't available outside of the financial services sector (I forget the name, it was 1 or 2 letters... KQ or something like that, an array programming language whose code looks like APL). I wondered how people broke into such a market. It's always seemed like an environment with interesting challenges which would have pretty objective goals and practices out of need.
Honestly, OCaml is hardly esoteric. It's not a major language, but it's not a specialty language either. I'd put it in the same category as Elm, Erlang, or even something like Julia. Scala takes a ton of inspiration from OCaml, to the point where you pretty much feel like you know one if you know the other (at least speaking for myself that's been true).
Yes you are exactly correct! Kdb and the Q language that back it. I thought that was really prevalent in the trading arena just based on other postings I'd come across on HN.
Additionally, financial institutions are getting geared up for major participation in the cryptocurrency markets. Will be interesting to see how that changes the landscape.
To be fair, quant vs. programmer is a false dichotomy since most quants did degrees like mathematical finance, actuarial science, etc. where computer science and numerical programming courses are part of their degrees.
Depends where you go. Some firms have a very bright line between "quants" and "programmers", and you don't want to wind up on the "programmer" side of it.
Agree 100%. First firm where I worked the programmers were treated like monkeys, and the traders, some of whom weren't even that bright at the time, were almost godlike.
Some other person wrote on here about traders being angry assholes. That firm had a ton of angry asshole traders because they'd push and push and push on deadlines, didn't understand the system at all, got hacked up buggy code, never invested in unit or system tests. It was an absolute disaster.
Little anecdote: First year I was out with a more senior trader on my team over drinks. He was talking about the glory days of 08 where every single trader earned a million dollar bonus or more. I asked about the one infrastructure developer who came up with a system to use multicast to simultaneously publish our EuroStoxx signal to DAX along with a dozen other instruments. Nothing earth shattering in 2017, but it quadrupled our P&L and market share at the time.
> Eighteen are now enrolled in the program, including Korn.
GS has 34k employees. Thousands of traders and probably >10k in IT. 18 people isn't really a revolution. Devs still are second class citizens in banks - the highest earning ones switched to become traders and just do some python/R/VBA on the desk. There are a few hedge funds (Millennium, Two Sigma) where devs are nearly treated as well as traders - would be interested to hear of any place else.
> There are a few hedge funds (Millennium, Two Sigma) where devs are nearly treated as well as traders - would be interested to hear of any place else.
Software engineers are treated very well, if not as well as research scientists, at essentially any large or successful quantitative hedge fund. This includes firms like DRW, Optiver, Tower Research, Two Sigma, RenTec, Citadel, etc.
Counter view: I only work in the field because it's the only way in tech to pay rent where I live. I actually hate it, it's of zero value to society and frankly the trading/business side are real assholes under pressure
> if your firm created no value for society, it would go out of business.
This is silly wishful thinking. There are plenty of businesses that are sustainably built on screwing people over or simply extracting money from them.
> What value are financial services?
I don't think he's talking about financial services like retail banking. I think he's talking about HFT activities other than market making. Legal front-running, etc.
It’s literally incorrect that businesses can sustainably provide no value. It may be the case that they can sustainably offer deals you wouldn’t take, but how we get from there to the business being immoral, useless, or deserving of a ban is more nebulous.
If people are systematically taking deals you wouldn’t, there might be a broad societal problem (why is tht crappy situation anyone’s best option?) or an information problem (why don’t they know better?) but the business offering the deals is probably not a logical or useful target if you want to address the larger problem.
> It’s literally incorrect that businesses can sustainably provide no value.
That depends entirely on the definition of "value".
If a business exists solely to move money/stock around, that action does not create "value", in the sense that juggling oranges does not create "value".
Trading stocks is an effort to take advantage of the fluctuating relative value each stock has, so that individuals can take home the "profit". Such "profit" is not created, but sifted out of the economy. Is that 100% worthless? No, but it's darn close.
The size of the financial sector swelled to about 9% of GDP in 2010 from around 5% in in 1980. Perhaps that is what the thread was referring to. Nearly doubling in size when finance doesn't produce anything real. They are a middleman. Would certainly like to see nationalization of large parts of finance such as simple checking accounts, insurance, and the big one, lets double down on Social security so that people can have dignified retirements regardless of what their work was. We dont need a big financial sector to have a great economy; Its a brain drain. Smart people should be working on real world problems like diseases, automation, efficient cheaper construction, environmental remediation, etc. that improve the standard of living for society.
A lot of that is simply busy work and can in turn be automated. A huge bank like Goldman is essentially a middleman - and middlemen can also be trimmed down. For example, I was just talking to a friend who works at Goldman as a trader. 75% of his job is to answer the phone and deal with clients issues (eg “why is my account frozen?”). His job honestly sounded like a customer service job. He’s also the person who signs off on wire transfers, something that could easily be automated.
So much of the “value” Goldman supposedly creates is due to them creating an artificial need for it.
I think you're slightly confused at the distinctions between prop shops, hedge funds, asset managers, and model driven quant desks at investment banks. Having worked at at least 3 of these, I can tell you again:
On the sell side, it's just pricing and hedging out each other's increasingly complex structures.
HFT Market makers have some merit in providing tight liquidity and low spreads.
Don't even get me started on buy side speculation.
Flash boys is mostly incorrect and at points misleading, and this is coming from somebody whose entire job is trading systems to avoid hft "picking off" my institutions orders/market activity. I can write more once off mobile but there's good material online about why that's better written than what I would write.
I'm in the range of 3-5 years out of undergrad, working at a quant firm. My total gross comp for this year will be about 400-450k, majority of that is bonus.
By quant firm I mean either a proprietary trading firm (no outside investors) or a hedge fund (takes outside money), but not a bank.
I have a CS background but have self-studied in stats, and received mentorship from people with strong math backgrounds.
Main advantages of quant shops (doesn't apply to banks) over working at Google/FB/Netflix etc:
- small company size, startup style working environment, but big corp pay (or more)
- more varied/interesting work, e.g. mix of distributed data processing, high performance numerics, or performance optimization in latency sensitive code
- compensation paid purely in cash, no stock options/RSUs
Regardless of what industry you are in, the closer you are to money, the more honestly you will be paid. This is why in many companies some of the best paid people (sometimes more than executives) are in sales. It is quite difficult to tell an employee that you can only manage a small raise if they can say "I brought in $15 million in business this year." If they have no discernible way to determine what they have earned for the company, they can be had for a song.
This was one of the reasons I left the agency world. I was leading the strategy part and pitching for new business (not to mention scoping the work and contract) and won 7 figures in new business we would not have won otherwise. Didn't get a dime in commission or bonus (or even a raise that year) because I wasn't officially on the sales team and was salaried. Learned my lesson there.
Where is that in the pay distribution? Near median? Realistically, where, comp-wise, would the average SWE plateau?
How many hours are you guys putting in a week?
What was the interview like? Did they go deep into stats/stochastic-calc/derivates-pricing/etc ?
---
I'm currently in tech, making near the low end of your range. Getting promoted to the next pay scale (Staff SWE) is a lot more difficult. I'd seriously consider jumping to fintech if 400-450 was below median.
This sounds really interesting to me. Are there other downsides to watch out for? I imagine the security environment is very restrictive: no work from home? No root access on your workstation?
What kinds of skills do you need to get into this area?
The security varies a lot. At one hedge fund I interviewed at, security was almost MI5-like - no phones in the office, no direct internet access, etc. At the place I ended up working, it's rootapalooza. Not in prod, obviously - there it's sudo-a-gogo.
I'm a senior at an HYPSM school, currently looking at offers from one of {Google, FB} and one of {Jane Street, Old Mission, Five Rings, HRT, Virtu}.
The offers are about the same at about 200-250k depending on bonus and stock performance -- the big tech company upped their stock grant to meet the expected bonus at the trading company. It's not obvious to me which one will be more in a few years, and it seems like hours in tech are better. Can anyone weigh in?
It may be close for the first year or so due to the signing bonus, but 3-4 years in if you're performing at the expected level at a trading firm you should be making double the 200-250k number, and I doubt the tech firms match that kind of compensation (at least for most people).
Another factor you may wanna consider is the opportunity cost -- how hard would it be for you to work at GOOG/FB if the trading firm job doesn't work out? And vice versa?
Face the market you work market hours. Tech doesn't have that limit. Both are cyclic, but FB is probably here to stay. Finance can be very dynamic: think LTCM.
A friend works at Virtu and 60 hour weeks are the norm if not more. They're very aggressive wrt managing headcount. Pay is great in absolute terms but pretty low as a % of revenue.
Doesn't mean it's a bad place to work. I have a ton of respect for the business they've built. Their discipline on costs is saving their asses in historically slow market conditions while many competitors are merging or shutting down.
As a Norwegian software engineer, my mind boggles when reading the salary discussions of engineers based in the big cities in the US.
Most engineers here are paid a small fraction of the numbers you cite, even though we are decidedly a first-world country. I guess good ML people will probably get over 100k here, but I haven't heard of anyone near 200k. Any thoughts as to what causes the massive difference?
The dollar is massively over inflated in major US cities. Studio apartments in Manhattan sell for $800-950k. A loaf of decent bread costs $6-8. A beer costs $8 (plus tip). I know that Norway is also expensive, but is it this bad?
This isn't the whole explanation, unfortunately (I really want to understand this relationship). A beer costs more in Norway, a good loaf of bread would be $6 (although you could probably get a cheap one for $2-3), a studio apartment in Oslo would be $300k-ish.
So the biggest price difference is in the rents, or the housing prices. I would probably increase my savings by a factor of five or more if I had a salary like the ones cited in this thread :)
Obviously it's supply and demand, I just wonder which side of the equation is skewed and why.
That isn't it; all the things you mentioned are paid through very heavy taxation, and all numbers we've been discussing are gross salary.
You could probably attribute a 14% difference through the "employer tax" which the US doesn't have (which is taken out of your paycheck before the gross salary number is cited), and arguably another 4% due to mandatory retirement savings. But it doesn't make up for it when the difference is 50%-75%.
I've seen claims of $1M base salaries in NYC, and seven figure bonuses are certainly possible (in HFT). $300K+ (incl. bonus) is no problem, but ML is less of a factor than FPGA and C++ experience.
I work in prop trading and I've never seen that kind of guaranteed salary offered. Maybe as a one time signing bonus to poach someone from a competitor. Typical base salaries for traders/quants/engineers are in the low six figures.
I'm head of a small team running high frequency strategies. I control parameters/portfolio/risk, design/fit predictive models, code C++ for simulation and live trading, and spec out designs for our FPGA developers. I make around 200 base with average bonus in the high 6 figures, good years over a million. At firms with uncapped contractural payouts, bonuses can be far higher for very profitable traders.
But if you come into the industry expecting that, prepare for disappointment. The markets are super efficient so it's hard to make profits. A lot of good ideas are done to death already. You might be stuck on a bad team or not get the opportunity to advance. Making mid 6 figures all-in on a consistently profitable team at a profitable diversified firm is a very good outcome. Modal outcome is making low 6 figures for a couple years and being pushed out.
Also while some enlightened firms realize that infrastructure is a competitive edge and pay big bonuses to engineers, there are many where they're second-class citizens. Only take front-office roles at such firms, not just because of the money, but because you won't be respected. IMO they should be avoided entirely because they'll eventually fall behind and lose.
Thanks for the details. Your depiction is in line with what I've seen personally, but the million dollar base promises did filter in after I left; not for me, but to pass along to people like you, probably.
Complete nonsense. Plus note the intersection between FPGA dev and ML is likely to be practically zero and equally useless at quant firms unless you're in an AI firm building ML on chip. Far easier to go from FPGA to ML than the other way round.
Yeah not really, many HFs will match that 400k, but many will not. I have received a few offers from these quant funds, the offers have been good, but nothing I couldnt get working in tech. My understanding is that the average finance professional makes less than or equal to tech, but the top 1 - 5% of finance easily make 7 figures.
> My understanding is that the average finance professional makes less than or equal to tech, but the top 1 - 5% of finance easily make 7 figures.
This is more or less true, but an interesting thing to think about is that the number of people (with technical backgrounds) making 7 figures in financial firms is smaller than the number of people making 7 figures in technology firms. This is probably obvious to you if you work in finance, but I think most outsiders don't realize that.
If you include people with sales backgrounds like investment bankers then the numbers might shift in favor of finance.
I'm not sure how to answer the question. How would you answer the equivalent question about finance? It's not like working in finance guarantees you 7 figures. I know plenty of people on both the engineering side and the trading side who never made more than 2-300k.
If you want specific examples from tech then the Waymo article from a while back is helpful, but that doesn't mean that you need to work at Waymo or Google to get there, although a large portion of those roles are at Google-tier companies.
Your average agency will not be hiring pure developers and paying them 7 figures, although they might to someone who is generating revenue and bringing in enough business to justify it.
the op made a claim so I am assuming the op knows how to answer this question. The claim is that many more engineers make 7 figures in tech than in finance. My question, specifically, is, what fraction of engineers in tech make 7 figures, let's say what fraction of engineers with 10 years experience and what are the types of positions at those google tier companies that pay this much. I think this is pretty specific
My comment was specifically about the "I mean what sort of position do you need to have to get there?" part of your question.
> My question, specifically, is, what fraction of engineers in tech make 7 figures
I doubt that OP meant that they literally have data showing that it's 1-5%. That's a wide enough range that it's meaningless. I suspect the meaning was more "way above average, but reasonably possible".
I think only distinguished engineers get that outright at amazon of which there are only like 10-15.
However, I bet most principal engineers and up who have been here 3+ years are hitting that lately with how good our stock is doing and that amazon pays higher ups mostly in RSU's.
If literally an engineer, not an executive that used to be an engineer or a famous engineer, this is accomplished by winning the stock options lottery.
It seems plausible to me at least that there are more ordinary engineers that have won the stock options lottery across all of tech than the total that are earning equivalent amounts in finance.
Well, there are more engineers working in tech than in finance. The relevant question is the odds of a particular engineer winning the options lottery than 7 figures in finance
Ah that matches what I have heard. HFT/HF is paying up but overall, as someone who is coming up on making the decision, it's hard to see the upsides of finance.
Losing all visibility/ability to publish/being a first class contributor and not second class dev..
I wouldnt get into HFT, its commoditized. The future of finance is in processing huge volume (internet scale) unstructured data and analyzing it in real time. HFT of website data if you will, right now we are in the infancy
But, I agree, finding alpha through alternative means at internet scale has been around for at least a decade, but things are getting more interesting now.
How large do you think that population of people in finance making 6 figures is in raw numbers? When you compare that to the tech industry, there are certainly high level engineers and managers that make 7 figure salaries all things included (especially given the growth in stock price at certain companies).
"How large do you think that population of people in finance making 6 figures is in raw numbers?" you mean 7 figures? if you are a top engineer in finance with many years of experience, you are likely making close to 7 figures.
Depends on what you mean. If you're looking at tech trading of stocks from home, quantopian is a great place to start. wallstreetoasis forums for a traditional roles for juniors and people trying to get in.
And just remember its a shrinking area. There are lots of traditional traders that have been laid off and are looking for work.
Personally, i am working on building a cryptocurrency trading platform, because the API's are easily accessible - but the markets entirely different in behavior. Still, rolling my own implementations of technical indicators has been pretty insightful, and backtesting has shown whats worked and what hasn't in given market conditions.
but i realize you're probably talking about more traditional markets like FOREX
Fee isn't limited to what you're paying the exchange, you have to take into account slippage - the price you use in your backtests is NOT the price you could actually realistically buy at because as soon as you buy, you move the price (especially in size).
Strictly speaking, you should be incorporating level 2 order book data to properly gauge fees and simulate real world profits from your backtests.
poloniex has historical data you can use. I trade on bittrex, but their data doesn't go back very far - enough to do calculations in the now, but not nearly enough to backtest against.
Say “getting people to click on ads” if you like it better. In the end it’s just making money as well. I guess working at a loss making company can be much more satifactory from a moral point of view but it’s not sustainable...
Yes, but step back and consider the immense intellect and creativity being devoted to getting laid, or speculating about what's going to happen on Game of Thrones, or thinking up clever things you should have said half an hour ago, or complaining about what's wrong with Star Wars these days.
Or arguing on the internet.
Just a teeny portion of our species' intellectual output winds up creating or discovering anything of real value, but that's fine, we muddle through nonetheless. The finance industry has its role to play in ensuring the lights stay on and there's food in the supermarket.
I don't think intellect is much of a factor in getting laid. My bet here is that intellect, or at least use of intellect, has an inverse effect on success in that area :).
The biggest difference between my example and your examples is that it's a job. This talent pool is using their working hours (and quite often most of their waking hours) pursuing a goal that provides almost nothing to society.
The "finance industry" that we're talking about here is self-serving and in most cases is having no positive relationship with the lights or the supermarket.
The transportation industry, by comparison, has a profit motive but ultimately turns human creativity and intellect into things that arguably make life better for a large number of people.
This is unfair, finance is basically an abstraction over our resource constrained world. We didn't get to choose not to live in a world with limited supplies of everything, and removing money from the equation doesn't change that constraint.
Given that we've chosen money as our method of resource allocation, it makes sense people who want access to more resources would try to optimise their profit building strategies, and smarter people are going to make better choices. It is why financial hubs, such as London or New York, have more resources to play with and so can provide a higher quality of living.
People aren’t talking about money itself, they’re talking about hedge funds and prop shops and traders, which didn’t exist prior to the 70s and even Wall Street traders weren’t trading pure securities until the early 1900s (it was slaves and real shares before then) Economy worked just fine without the quants, and ever since trading has been a major segment of Wall Street they’ve completely crashed the markets three times and had to be bailed out by the fed (1929, 1973, and 2008). If you take the capital allocation justification at face value they suck at it. They’re losers underwritten by the government not winners smartly allocating capital.
Supermarkets were full of food and lights were on before hedge funds and prop shops existed. The first supermarkets weren’t even financed by wall street but a real estate tycoon with his own money.
Wall Street is a welfare program for traders, underwritten by the taxpayers and paying for the financial sector’s total losses in 1929, 1973 and 2008.
Are there any automated trading firms that are specifically focused on exploiting the unintended combined behavior of multiple systems operating in an environment where they modify both their own and each others environment (the stock market being that environment)? While it might be tractable for an organization to construct their system such that it can project its own consequences to its own sphere of interest, once you mix in other systems operating in the same area all bets are off. And once you have the majority of trades being performed by automated or rules-based systems, you ought to detect their parameters through assuming reasonable things about those systems (processors which are limited to a certain number of operations, bounding the latency on their response to change, limitations on the complexity of the systems or rules being used due to needing to be comprehensible to human analysis in its source form, etc) and looking for predictable patterns created primarily by the interactions between those automated systems. Everything I have read about automated trading systems (which of course is always very light on details due to everything being extremely valuable and proprietary) seems to indicate that most automated systems are designed as if they are operating in an environment where the other traders are primarily humans. If anyone has come across indication that they are treating the movements of the market as the output of deterministic software systems rather than 'natural' trading, I'd be appreciative of a link!
Depends how you mean exploit. Not a lawyer, but to a first approximation, doing anything in markets to elicit a response from others rather than for bona fide economic reasons is illegal market manipulation so active exploitation is rarely done at serious firms. It's usually individual day traders or offshore boiler room type groups.
Passive exploitation, as in detecting predictable future actions, is probably common, if unintentional. Most automated traders are searching for anything that predicts an imminent price change, on average. Predicting future actions and predicting prices are roughly equivalent. As you move along the continuum from straightforward arbitrage to esoteric black box quant stuff, it becomes more and more difficult to understand how or why a trade even works. I'm sure there are models that predict prices very well solely because they've locked on to some combination of tells unintentionally leaked by another algo.
Here's an example of someone who built a model to detect algorithmic iceberg orders reloading: https://mechanicalmarkets.wordpress.com/2015/04/30/market-da... - he makes an interesting point that any quant data mining order book event sequences that predict strong price movements would find the same pattern easily.
Hi.
Well, yes, lots of people has come across that fact. But the guys able to exploit it are market makers. They look for kinds of repeated patterns in the form of orders, stops, time of day... And they later trick on them until they stop working.
As you noticed, knowing you'll get someone to buy your thing at a relative price at some specific time and day can be very profitable.
Still, it's difficult to read something "worthwile" about this.
I wonder if this will see an ability to more easily switch into fintech/banking as a developer working in various other industries. I worked with GS via a mobile agency for a while and really respected the developers and what they were doing from an engineering PoV. I had always viewed that industry as essentially closed unless you got in straight after graduating.
Has anyone ever made the switch into an engineering role within a bank? Is it hard to get past the lack of specific "financial" experience if you have X years of software engineering behind you in general?
+1 for an answer to this question. Fintech seems to me to be as walled off to outsiders as game companies are, only tending to hire amongst their own. The job descriptions seem to require more specialized "arcane wizardry" skills than the normal above-average C++ or mobile developer. If you're a commoner who broke into the biz, I'd be interested in learning your story.
Not working on a fintech but FWIW I work on trading platforms for a well-known investment bank. Getting in was not particularly difficult but even with some experience you may need to network with someone on the inside to get your resume in front of the right recruiter, simply submitting your resume via the usual channels appears to be a crapshoot (though I had never applied here because I wasn't aware they were in town)
My salary is quite average for the role/seniority level and AFAICT it's resented by many developers who work here because many will trade their "in" on their industry into more interesting fintech/banking work elsewhere somewhere down the line for higher compensation.
Knowledge of financial transactions is definitely highly specialized and I don't have nearly enough of it yet but it has started to sink in
Would I be correct in guessing from your username that you work with FrontArena? It seems in-house coustomisation of those kinds of systems isn’t paid very well (although I know some murex guys who get paid quite a lot as consultants).
What does the day to day look like working on these beasts? Are you in Europe? Most of the developers I know working on these platforms seem to be quite silod and not really involved with the programmer ‘scene’ as it were (all using windows, no opinion on editors, never heard of hn, don’t care about this weeks $shinynew etc which I know shouldn’t be any kind of grading scheme but it always strikes me as a bit strange).
Would like to hear your opinions on it: are most people who end up working here turned into these (generalisation warning, I know not all) unmotivated souls by the sheer pain of working on this stuff or does it just attract candidates who don’t really care that much about tech?
The username I have carried for some time since I used to play this bomberman clone called Teslatron.
I am not in Europe but a large chunk of the team I'm working with is. Divulging the name of the company is frowned upon, but it's one of the large full-service investment banks
Day-to-day is a lot like other jobs but projects are better managed than I have previously experienced resulting in less meetings for developers. Development can be faster-paced as a direct result.
News of $shinynew is discussed daily by developers (e.g.: React's license, React 16, Vue, etc.) but projects themselves tend to be more conservative as a matter of company policy. I can say the average developer on my team is more capable than the average developer at other companies I have worked with previously. People who are unmotivated will be moved to projects nobody wants (my observation)
Regarding editors, environments, tech events, HN, StackOverflow, etc.: participation is likely there but self-identification is a huge no-no
TLDR: It's not all rainbows and there are institutional challenges, but the tech side is not naive
There are plenty of CRUD positions on Wall Street. If one wanted to get into a "strats" type role maybe it would be easier to just join as a regular Sr. Engineer. Get paid above avg salary to write web services or whatever, learn domain knowledge in a less stressful manner, and figure out a way to get into the group you really want internally.
Banks aren't the only financial services organizations out there. Look to get into the sector through smaller firms that do IR, asset management, etc. You'll still be paid competitively, and you can pick up knowledge about the sector. They might even foot the bill for an entry-level cert like the Canadian Securities Course (I don't know of any equivalents abroad but I'm sure they exist). With these under your belt, you will have a leg up in applying to a larger bank.
Crypto is doing really well these last two years, it seems like that is the startup equivalent to the investment banking/quant jobs. However with Crypto, it does feel like the beginning of the dot com era. Dot Com == ICO
Well it's a lot easier for a trader to learn enough programming to implement specific algorithms than it is for a programmer to learn about financial markets.
There was an amazing article about one of the programmers responsible for inventing CDS and I can't find it again :( it included such gems as putting offal in a grinder and getting steak out of it, anyone remembers it?
I would think it's pretty good. I get recruiters calling me all the time, a lot of them are quoting £150-200k total comp. That's for a guy with over a decade's experience in financial coding. By contrast on StackOverflow the salaries that you can see tend to be much lower for the non-finance things.
It's not surprising that it's higher than the average job on stackoverlfow. That's about what I would expect. I'm not saying it's horrible, but it's not exactly higher than top paying tech companies.
I'd say the typical for a senior engineer is around 70, ranging up to about 100 or so depending on expertise/company. Going to lead/architect you're looking around 80-120 total comp (there's a huge range...). Finance will pay much better like you said, 150-200, but the tech will be pretty bad. People making over 150 out of finance are generally contracting or working for a US company.
I'd put it like this, the bulk of positions are around what I quoted. That's for senior people who aren't managing large teams, just run-of-the-mill squaddies.
There's also a few jobs where the pay is unlimited. Typically you need a track record but you'll also get a cut of the profits.
And then there's management, which is another level as well.
I made the remark primarily because the article discusses career trajectory as one of the motivations for emphasizing the combination of “traders who code.” In essence, they’re advertising that your pay won’t get capped out at the level of “a very good programmer.” Instead, the cap will the higher number of “a very good trader.”
I liked this article because the roll of the programmer on wall street has been going down recently from my perspective.
Programmers used to work along side quants and were well respected. This was generally because practicing quants didn't used to be able to program. This was especially true in the 80s and 90's where programmers and quants where almost valued identically as one required the other to make things happen.
Somewhere around mid 2000's most quants started to be able to program pretty well. Not necessarily with C++, though lots (especially former physics) could use C++, but certainly with mat-lab, etc. This really shrunk the required programmer head count,
This, coupled with the fact that few programmers were trained in fields like stochastic calc, and the advent of offshoring meant that programming started to be come a lower status job around this time.
Some relatively small fields like HFT kept good programmers in need, but these jobs are relatively few and far between compared to the sell side bank jobs.
Now, programmers are much more valuable/respected once again, due to the requirement for top notch software, not even as a competitive advantage, but just table stakes to compete in the markets.
And like most fields programming in finance has really started to specialize:
- lockless programming, low level Linux, user mode network drivers, etc for HFT
- distributed systems, low latency messaging for quote dissemination in large systems like what most sell side banks run.
- real time risk limits, web UIs for helping to visualize and alert portfolio risk
- "production" modeling of quantitative models, for those that can span the programming/math boundary
> This, coupled with the fact that few programmers were trained in fields like stochastic calc,
Anecdotally though, I've met plenty of coders who did this to themselves by refusing to learn more about the domain or the business.
They basically silo'd themselves, then wondered why their colleagues didn't treat them as an integral part of the team when they couldn't speak the language or understand the concerns.
Taught myself enough stochastic calc to change and validate pricing models while on the job.
Coders who understand their domain well are immensely valuable, and are easily worth at least 3x, because they don't over- or under-engineer a solution, and because they can conjure up with technological solutions for problems that nobody else in the company can even conceive of.
> Coders who understand their domain well are immensely valuable, and are easily worth at least 3x, because they don't over- or under-engineer a solution, and because they can conjure up with technological solutions for problems that nobody else in the company can even conceive of.
This is a crucial point to me. Being able to code is good, but being able to code well while also having an intermediate or expert understanding of the domain that you are working within is what is going to set you apart in the long run.
In the industry, developing software can usually be seen as the process of creating a tool that will help to better solve a problem. It seems strange to me that you would want someone who doesn't have a solid understanding of the nature of the problem (or the solution) to be designing and building the tool, just because they claim to be good tool-makers.
> It seems strange to me that you would want someone who doesn't have a solid understanding of the nature of the problem (or the solution) to be designing and building the tool, just because they claim to be good tool-makers.
I couldn't agree more and this is my main objection to "AI will replace [insert any job here]".
A lot of problems are inaccessible to programmers just because not enough of them seem to be willing to understand them. Complex rules about accounting in large companies seems to be one of them for example. This is why I think AI will certainly overtake many professions but mostly professions that are required on a large scale and that are easy to understand: driving, cleaning a house, building a house, ironing, cooking, assembling objects, etc. For complex accounting or legal issues, I am a lot more skeptical. You don't need complex AI to run accounting systems, they can easily be done with regular software. However in large companies you still have huge accounting departments booking and maintaining these complex logics by hand (excel), a lot could be automated now. And the same reason why they haven't been automated in C# now will apply to AI in the future.
Re your AI thesis, I think it’s the exact opposite. Deterministic, rule-based processing applied to lots of similar data are perfect for AI-fication. It’s all in the computer, and you can kick out to a human-driven exception process when required.
All that physical, real-world stuff you mentioned (cleaning, cooking etc) is vastly more complicated, riddled with experience-driven subjective judgements (is this paper on the floor trash or should I put it back on the desk?) and highly situational. Extremely difficult to generalize and handle with an AI. Not to forget it’s messy stuff in the real world, with engineering and hardware and grease and failures.
Most of the complex accounting and legal questions that these departments are doing manually aren't completely trivial either (we are not talking about just adding assets). But the determination of whether the paper on the floor is trash or not is something that pretty much any AI professional can apply his mind on (and will likely do given the size of the market), whereas very few will apply their mind on how IFRS should treat a tax asset in consolidation when the corresponding tax treatment under USGAAP is a still in discussion.
If you look at self driving cars, they are already doing the sort of wizardry of your trash example when analysing the other objects in sight.
And again it's not so much a matter of technical complexity, it's a matter of the domain being inaccessible to those who can best automate it.
>Most of the complex accounting and legal questions that these departments are doing manually aren't completely trivial either (we are not talking about just adding assets).
These are exactly the types of problems that machine learning is perfect for. Problems which involve chains of nebulous decisions that would be difficult to program traditionally.
Couldn't agree more. I have met so many programmers in banking who seem to actively avoid having to learn anything about a bank, not even complex math.
In theory, project managers are supposed to bridge the gap between the business and programmers, however my experience is that they usually understand neither the business nor the IT side, and are only adding more confusion.
People with both knowledge of both worlds are incredibly rare and valuable.
what then is the best approach to learning how a bank works? where would one such programmer start?
Work for a bank.
I could not have formulated this better myself.
All but a few programmers I have met have tried to understand the underlying reason for their employment and the businesses issues they are employed to solve.
This approach have caused programmers to loose a lot of influence and status.
10 years ago companies had developers develop their own customized CRMs, which can really benefit a company if their domain-specific needs and processes can be integrated to such core-business software.
Nowadays, Adobe Campaign have a cake walk selling their bloated, general and non-tailored “CRM” solution to business.
How can they Adobe etc do this? They approach the marketing and sales team and offer a nice slide show.
Why does not IT or programmers even get involved before the purchase? Because no one likes talking to the idiots in IT that does not understand the businesses needs. They only nag about some “technical debt” (that no matter how much money is poured on the teams never disappears) and the need for re-writes of fully functioning parts of the system that introduces new bugs.
>Anecdotally though, I've met plenty of coders who did this to themselves by refusing to learn more about the domain or the business.
Anything outside of learning enough to do your job should be at the business's expense.
1) some fields are not good for self-teaching. It leaves the possibility open to misunderstanding and writing code to that misunderstanding
2) the business will probably have a good idea of conferences and workshops that impart the knowledge the business is looking for
3) this is more fuel for "unpaid studying outside of work hours"
Based on posts like this, it seems really dumb to enter software first -- better to get a degree in some other technical field and just teach yourself to code.
In many environments, you probably can do a lot of that learning at work, just by talking more than strictly necessary to people having the knowledge.
> Anything outside of learning enough to do your job should be at the business's expense.
There's no "should" here — it's an open market with a lot of agents. You're free to pursue certain working conditions and then look for employer that will agree with them. And another developer is also free to go an extra mile and ask something extra in return.
This is why I'm against any kind of unionizing for developers, by the way — too many different situations and opinions about what our working conditions should be.
If it increases your value and allows you to demand a higher salary, that is at the businesses expense. You are a business expense.
It'd add even more value if companies spent more on proper training. Maybe even training geared for software devs learning the field from that perspective if that's available.
But if they think I can become a domain expert by reading junk on the internet with no one really verifying my skills, I'm happy to take their paycheck.
I apologize for the off the wall comment. Every time I read about wall street programming, I've got this subliminal voice in the back of my head wondering how anyone could find the motivation to do this work. I'm obviously revealing my biases against the financial industry here. Also, I'm unable to make myself do work if I don't see it contributing to some greater vision or cause. Are there people out there who love finance and learn these types of specializations on the job, or are they hired away from other industries lured by big comp packages?
I guess it's just a case of lack of empathy on my part. I just can't understand why anyone would want to put so much effort and skill into making bots for farming money for a bunch of questionable people.
I don't feel too guilty for saying this. Every time a story about bitcoin is posted to HN, people go off topic and start complaining about everything that is wrong with bitcoin. I don't see the same when we talk about finance in general, even though similar sentiments apply, maybe more so.
I think it helps to understand that the financial sector provides capital that serves to energize virtually every other sector in every industry.
Without finance, there is no biomedical research (private especially). Without finance, there is no aerospace advancement. Without finance, there is no alternative energy. Being able to facilitate a more liquid, more transparent financial sector is, in my opinion, a calling worthy of any programmer who seeks to make the world a better place.
Your assignment of people as 'questionable' is likely an artifact of the 2007 recession, and that kind of thing is a personal choice. I can guarantee you that there are 'questionable' people in every sector and in virtually every company. Sometimes people are great, and sometimes they aren't. Their field of employment has very little to do with what makes them that way, and I think that it's a little ignorant to be so casual (and I say this because you are certainly not alone in your judgment) with how you view people.
Remember how, as children, we are taught to not judge books by their covers? Why do the same with a programmer who works in finance? Sure, some of the work may be a little pedantic, like squeezing an extra millisecond out of an algorithm, but hey, that's computer science in general, and I liken that to an F1 racing team spending hour upon hour sculpting the perfect frame for their car.
Sorry for the rant, I've spent a decent amount of time working on financial algorithms, and what you said touched a nerve.
Have a great day!
WELL said! :)
I stumbled into finance a few years ago and I’ve had overall the best bosses and work env ofanyone I know. I’m also doing a more audacious program of research engineering/ computer science than I could likely do anywhere else.
I like to sometimes describe my work as some combination of “making the systems that fund aerospace at least as reliable as commercial aircraft”, as transparent as intergalactic vacuum, and easy to use to boot! Or at least that’s the idea :)
On the hft front I recently started at poking at how to get accurate time on a dev computer, which gets fiddly the moment you want really interesting accuracy on commodity hardware :)
Thanks for the praise, and for the insight per your experience. Sounds like a great gig!
I looked you up, and see that you are likely working exclusively in Haskell, but thought you might derive some value from Carl Cook's presentation on optimizing HFT code at the most recent cpp convention. I recognize the languages are distinctly different, but he provides some interesting theoretical points along the way that might be advantageous to you.
Here's a link: https://www.youtube.com/watch?v=NH1Tta7purM&feature=youtu.be
Also, I can sympathize with your plight to run scalable sims. I've done some work in bioinformatics, and am building a small cluster at home so that I can learn to write code that'll scale to more massively parallel systems that are de rigueur in that domain.
Cheers!
I’ll have a gander! Thx for the link. I do code in other stuff when it makes sense ;)
Yeah building stuff that works easily in the small and sanely in the large is a fun challenge. Also hard.
Possibly because of your original comment, or perhaps unrelatedly, I’ve been lately saying “finance done right is the lock free wait free scheduling algorithm for moving society’s resources around” —— all the other stuff folks think of as finance is really just icing and fancy wrapping on top of that core truth
As someone who works (hard) in the part of finance you're describing, it's a mix of a few things. From a purely intellectual standpoint, it's a very hard, very interesting quantitative/modeling problem to solve (ie. predict the price N seconds in the future is non-trivial). It's an extremely competitive field (you vs. other teams like you in the world competing for a slice of the pie), and you get results back from your experiments very quickly, so you know exactly where you stand. And finally, to your following point --
>Also, I'm unable to make myself do work if I don't see it contributing to some greater vision or cause
This simply is not something I experience. I've never felt any particular satisfaction from that particular aspect of a job, it's just something that never crosses my mind and incentive me in any way.
There's actually quite a lot of puzzle solving involved that makes it interesting. I think there are a lot of misconceptions of the industry at large and it suffers from bad PR after the global financial crisis. It's an interesting software problem to try and write algorithms that can earn alpha regardless of what is happening in the market. If anything computers have made our markets more efficient and it is great for price discovery.
It may not be the same as creating a product or service that people love or solves a need, but it is an interesting problem to tackle.
See that's the thing. I don't know what "earning alpha" means and why I should care, and working on "price discovery", which is also something I am unaware of, sounds more like a punishment than a reward.
alpha is a return in excess of what is expected. If you invest in a low cost index fund you expect alpha of 0 as compared to the index the fund is mirroring. If instead you put money in a hedge fund, alpha is the difference (positive or negative) between the returns the fund gets vs some benchmark like the S&P 500. Negative alpha is bad, positive alpha is good. Earning alpha means you are beating "the market".
There's also beta, which despite being the next greek letter, isn't the same at all! Beta is a measure of the price correlation between an asset and some benchmark in terms of volatility. Beta of 1 means that the asset moves in lockstep with the benchmark. Low beta means the asset doesn't move as much as the benchmark and high beta means the asset moves a lot compared to the benchmark. Tech IPOs are often high beta compared to the S&P 500.
Anyway, finance has a lot of intellectually challenging problems that happen to overlap well with people trained to build mathematical models that attempt to describe stochastic processes. CS, Physics, Stats, some branches of pure Mathematics -- all have direct applications in the work they do to some aspect of being able to understand and predict markets.
Appreciate the explanation. Thank you.
Price discovery means that anyone can find a fair price for some asset. For example, algorithms that provide liquidity in agriculture products ensure that farmers can easily liquidate their inventory, as well as hedging their risk through derivatives.
I wouldn’t close your mind to it. Finance attracts “questionable people” but finance is also very interesting on many levels and is a central part of life. Work through some finance math. Calculate an IRR. See how it works. If anything there’s a lot of great human history in it.
Thanks. I'm trying to be diplomatic here with my criticism and appreciate the positive and informative responses despite my negativity. You are right, finance is critical to social function and perhaps I've only been focusing on the negative aspects.
Finance is a vast world and unfortunately too often reduced to pure speculation.
First you have basic lending and borrowing activities. The bank can afford to pay you interest (well not now that interest rates are at zero) only because your money is invested, usually in loans.
Then you have payments, cash management for companies.
Then you have more complex services offered to larger companies, raising capital (equity and debt) on the market, helping them manage their risk with derivatives (interest rate, currency, commodities (oil, metals, livestock, etc prices), inflation, etc) plus some advisory on corporate actions (mergers, IPOs, etc).
You have insurance, reinsurance.
And then you have asset management, which you can call speculation if you want, but which is essentially people managing investment portfolios on behalf of savers, pension funds, insurance premium for long term risks, etc. With a wide variety of approaches from very conservative money market funds which invest in highly safe and liquid assets, to trackers who try to replicate a benchmark by over performing it by a little, to investments in illiquid assets (property, non listed companies) or very aggressive/leveraged strategies (hedge funds). Fundamentally what they are doing is to ensure your money gets invested wisely (at least they try).
And of course all sorts of intermediaries: consultants, brokers, market data providers, research (sort of financial journalists), prime brokers (who provide financing when you leverage liquid financial assets), system providers, auditors, rating agencies, etc.
Now you could call everything speculation: when a bank lends you money, it bets on your financial soundness. When an investor takes the opposite side of a commodity (say oil) transaction, it bets on the price of oil. But that allowed you to get money to run a business or to offset an oil price risk that could have made you lose money if oil increased (say your are an airline company). There is no real dinstinction between risk taking / risk transfer and speculation. But in most instances it does have a social benefit.
My feeling is that most criticisms of finance are based on a combination of a lack of understanding of what finance is and does, and indignation about how much money some people in the industry get paid. It's any easy target. Absolutely there are 'questionable people' in it, but are programers focused on getting users to click on more ads or spend time on a given app any better? Is tech really saving the world? There is so much more to tech than that, and most people in tech don't feel that's a fair representative of what they do. I think finance is similar, but I don't work in it so maybe I'm wrong...
Cash motivates plenty of people, as does the joy in solving a problem. Whether the value of that solution to society's benefit is positive or negative does not always factor into a solution architect's moral calculus. This is how I explain your observed point to myself when I see folks stoked to work for $largetechco or $adtechco.
> as does the joy in solving a problem
This is it for me. Futures trading is a zero sum game (negative sum if you factor in fees), so writing a program to automatically extract value from an ever changing market is a nontrivial and ongoing project. But unlike a game, getting good at it can pay the bills in a real way. I understand that in the big picture long term, I'm not making much of a mark on society, but nevertheless it satisfies my personal and financial needs.
Markets are not zero-sum game it is a common misconception.
http://www.businessinsider.com/the-stock-market-is-not-a-zer...
That's why I specifically called out futures trading. From [1]:
> In the financial markets, options and futures are examples of zero-sum games, excluding transaction costs. For every person who gains on a contract, there is a counter-party who loses.
[1] https://www.investopedia.com/terms/z/zero-sumgame.asp
That quote is misleading. He's talking about perceived price. The actual price of an equity is the price at which the execution occurs. And for every execution there are two sides, hence the zero sum game is the only game in town.
You are completely ignoring the enormous value that financial activities provide for society. Efficient capital markets, risk pricing and exchange are fundamental ingredients to an effective economy. Do you think these businesses get paid for nothing?
You’re comfortable enough to be ethical about your capital, for me it would be a big constraint. I’m sure I’ll get to be picky someday but that’s years from now. And at least there’s cool maths :)
Some make a whole lot less than you and still feel ill at ease working for the finance industry when it seems to soak up so much capital and contribute so little to society
Doesn't this sentiment preclude you from almost every programming job?
It's almost always going to be about finding the optimal way to funnel money toward whoever is in control of your employer.
Yes, it took me about 20 years to find an employer I am happy with, and that included myself for a while in the list of those I wasn't happy with.
May I ask what industry you ended up in? I assume it's something that you feel is adding value to the world?
Most of the quants and quant devs I work with are out-of-work physicists who want to provide for their families. Most of them would rather be tenured. I work in this industry because I cannot find employment in my field of engineering near my friends.
I just can't understand why anyone would want to put so much effort and skill into making bots for farming money for a bunch of questionable people.
Be aware that a large part of the Internet industry consists of persuading people to click ads, in order to enrich the Saudi sovereign wealth fund.
https://www.nytimes.com/2017/11/06/technology/unsavory-sourc...
Does that float your boat?
You wouldn’t be able to type this comment on your computer (and neither would I) were it not for financial markets to organize funding for the company that produces your device. Or you might not have gotten a mortgage to buy your home.
Financial markets serve a hugely important social function of allocating capital (intermediating between people who have money and those who need money).
>farming money for a bunch of questionable people.
The "questionable people" include grandparents, firemen, teachers, etc with pensions and retirement funds.
>I don't see it contributing to some greater vision or cause. [...] I guess it's just a case of lack of empathy on my part.
Your sentiment is common and I think it's caused by people not connecting the dots from grandparents' retirement pension to the hedge fund trader.
As one example, at the micro level, grandparents are depending on a pension.[1] They also don't want their _real_ purchasing power to diminish. The "real" not "nominal" purchasing power is an important distinction because if the price of bread is $2 in 2017 but becomes $4 in 2037, they want to have adequate future money to buy food. In other words, they want their money to grow and keep up with inflation.
Your grandparents stashing money in a savings account paying less than 1% will not protect their purchasing power from inflation. Likewise, the pension manager that oversees a $1 billion retirement fund to pay retired teachers will declare it bankrupt if it only gets a 1% return by buying Treasury Bills. The pension's assumptions to avoid insolvency might require 8% or better annual returns.
Therefore, all these participants in the economy are looking for better returns. You can't get 8% from savings accounts and t-bills. That's where money managers like private equity, hedge funds, and VCs in Silicon Valley come in. They sell their ability to generate higher returns. This expertise attracts capital from entities like pension funds.
So basically, a bunch of programmers are fine tuning algorithms because [...connect a bunch of dots...] your grandmother doesn't want to be a starving destitute and live under a bridge in her retirement years. If you multiply millions of everyday people not wanting to be destitute (macro level), you end up with hedge fund managers with billions to play with.
That said, can we "reconfigure society" so that all this competitive activity of shuffling money from one pile to another isn't done? Maybe. It would involve fundamental changes that we all can't agree on. For example, if you stop inflation you can stop the incentive to generate returns that try to beat inflation; but then some would say that "deflation" would lead to people hoarding money and the economy would crash. Inflation is the current winning ideology for the modern economy. It's hard to see how we could alter society such that Wall Street traders go away.
[1] for example, the $300 billion California Public Employees pension fund provides these benefits: https://en.wikipedia.org/wiki/CalPERS#Benefits
> That said, can we "reconfigure society" so that all this competitive activity of shuffling money from one pile to another isn't done? Maybe.
Your proposals boil down to ending deficit spending. Aside from the authority issue (no single entity has authority to do this, not even congress, definitely not the president), that means the US Federal government has to cut $650 billion dollars in spending.
Or to put it more directly: We currently have pensions, social security, an army (meaning having an army at all). Pick 2. The other one gets destroyed.
Still sounds reasonable ?
I do HFT and think about this a lot. I am essentially a middleman that facilitates trading across:
1. Time - I quote buy and sell prices continuously in Microsoft stock trying to earn the spread between them. You want to buy stock and my offer to sell is the cheapest in the market. We trade. Later, someone else comes along to sell and I have the highest bid, so I buy the stock back from them.
In between I'm exposed to price fluctuations, so sometimes I make money and sometimes I lose, but I make a profit on average by charging a spread between my buy/sell prices and predicting small price movements. The two of you could have met had you waited, but I let you lock in a sure thing by bearing risks you didn't want. Think of this like insurance. Odds are you spend more on car insurance than you're expected to receive in payments, otherwise the insurers would go out of business. But you probably prefer spending $1000 a year vs. spending $0 most years and $30000 once in your life.
2. Place - You're in the US and want to buy Nokia stock. Someone in Finland wants to sell for a cheaper price. You don't have access to the European exchanges and she doesn't have access to US exchanges, but I do. I trade against both your orders and earn a couple cents per share in profit. I helped you meet halfway across the world and also keep prices efficient to reflect global supply and demand.
3. Product - You want to invest in an index ETF that holds 40 different stocks. I quote an offer to sell based on where they're trading in the market plus a small profit margin. You buy my offer. I turn around and buy the stocks to hedge my risk.
I'm better at trading than you are and have very low costs. Even considering my profit margin, it's cheaper and less risky for you to buy the ETF from me than 40 individual stocks. I keep the ETF and stock prices aligned and also helped people who wanted to sell stocks get their orders filled. Win win win.
I think the work I do clearly benefits the market and natural investors. All of these trades are competitive too, so I can only earn razor thin margins per trade, or someone else will undercut me.
I do sometimes question how the market picks winners and losers, typically based on speed. Is someone whose system is hundreds of nanoseconds faster more deserving of arbitrage profits? Do high technology costs naturally lead to consolidation and reduced competition? https://faculty.chicagobooth.edu/eric.budish/research/HFT-Fr... this paper has some interesting thoughts on the topic.
FWIW I do think continuous price-time priority trading is flawed in that it creates an arms race in speed, but it's the least bad option. Other mechanisms like batch auctions are even worse. As a thought experiment, consider the #3 ETF trade I described. If I can't immediately hedge my risk and instead go into a batch auction for each stock, how will that affect the price I offer in the ETF?
PS: I do think it's a fun job. You get to play with cool technology and it's like a game that gets harder every day against very skilled opponents. Even the arms race isn't all bad. HFT firms have effectively bankrolled modern innovations in high speed ethernet, FPGAs, low latency kernel bypass networking, etc. that have knock-on benefits for other applications.
> As a thought experiment, consider the #3 ETF trade I described. If I can't immediately hedge my risk and instead go into a batch auction for each stock, how will that affect the price I offer in the ETF?
My idea for this is that when you rest an order, you should be able to attach a set of hedge orders to it that will initially be inactive, but will fire when the owning order is filled. This would let everyone get hedges done at better-than-HFT speed, and make things like speed bumps or batches a lot more tolerable for market makers.
You'd want some kind of pro-rata, or multiple orders at fill thresholds, so you could show big size and respond sensibly to partial fills. You might want alternative hedges to cope with price moves in the hedge target. You might want some mechanism whereby the exchange will pull the owning quote if the hedge disappears. There's all sorts of fun you could have.
For this to work really well, you'd want the hedges to be processed synchronously with the fill of the owning order, so there's no HFTable window of vulnerability between fill and hedge. That couldn't be done across different matching engines in the typical architecture current exchanges use. For futures, though, it might be workable - i think all the different expiries of a given underlying, and their spreads, are already handled synchronously, so that the exchange can do implied matching. That might not include some of the more exotic constructs which aren't in the implied chain, though.
This sounds similar to OneChronos, though I don't know if they've launched anything: https://www.onechronos.com/
You're right that there's precedent for this:
-Implied engines as you mentioned where traders can atomically express a view on the term structure without legging risk. For the benefit of non-finance readers: Say you believe the fair price difference between Dec and March futures is 2 index points, you could quote in the spread between the contracts (long one, short the other) buying 1.75 selling 2.25 without being fastest to respond to every shift in the index itself. Your order in the spread will create implied orders in the underlying contracts themselves and you can only be filled if both your long and short trade.
-IEX's D-Peg effectively runs a high frequency pricing model inside the matching engine to predict when an adverse price tick is likely.
-I think D.E. Shaw has a patent on a domain-specific language for complex orders where they can be priced off any other instrument(s), ratios between instruments, features of the order book, etc.
The real problem with batch auctions is coordination. So long as you have other exchanges trading correlated assets in continuous time, there will still be a race to adjust orders in the batch auction market after a pricing signal occurs. Even with a random delay or crossing time, it's still beneficial in statistical expectation to be faster.
You'd need to trade most similar products worldwide on a single platform. Since exchanges are a network effect business, this would likely lead to monopoly profits for the exchange operator.
I'm not sure "no high-speed arbitrage/monopoly exchange" is net better than "some high speed arbitrage/competing exchanges." I suppose you could have the government run the exchange as a utility, but that would stifle innovation. A lot of things we now take for granted started in experimental ATS and ECN startup markets.
> how anyone could find the motivation to do this work
There a lot of challenging problems in the finance. Some people like the problem solving and $$
> I just can't understand why anyone would want to put so much effort and skill into making bots for farming money for a bunch of questionable people
Its making money for yourself as well. Many jobs in Wall Street at % of profits based rather than a fixed salary.
I don't work in finance, but I see the modern global financial system as one of the greatest achievements of humankind and one of the main reason behind humanity's prosperity. I would love to work there if I had the skills or patience to gain them (instead, I work in gamedev, where I acquired immense domain knowledge over the years).
To add an angle from someone who shares your concerns about the financial industry, the high pay makes it so you can donate large amounts of money to charitable organizations of your choice.
It's actually a recommended route in the effective altruism community for those with the aptitude.
Of course people like finance.
We are not all some hippies who want to help the world. Fuck this shit. I want straight cash! And if you think people at the top of the tech industry dont think like this ur just naive as fuck.
Whats not cool about a bot farming money? Its like the best thing ever looool.
Has offshoring ever been a significant factor in this arena? Given the profoundly proprietary nature of the systems being built, I would think that unlikely. Offshoring development is a good way to find your custom software reskinned and sold to a competitor without much legal recourse.
Also, is 'mainstream' programming that common of a need? I had read (here on a HN comment thread) awhile ago that many organizations use commercial languages that simply aren't available outside of the financial services sector (I forget the name, it was 1 or 2 letters... KQ or something like that, an array programming language whose code looks like APL). I wondered how people broke into such a market. It's always seemed like an environment with interesting challenges which would have pretty objective goals and practices out of need.
Huh. I see C++, Java, and Python as the go-to languages with my wall-street programming and modeling acquaintances. Very little esoterica.
Jane Street's famous for its obsession with OCaml.
Honestly, OCaml is hardly esoteric. It's not a major language, but it's not a specialty language either. I'd put it in the same category as Elm, Erlang, or even something like Julia. Scala takes a ton of inspiration from OCaml, to the point where you pretty much feel like you know one if you know the other (at least speaking for myself that's been true).
https://en.wikipedia.org/wiki/K_(programming_language)
You are thinking of Q and Kdb which are used to process high throughput tick data.
Yes you are exactly correct! Kdb and the Q language that back it. I thought that was really prevalent in the trading arena just based on other postings I'd come across on HN.
Additionally, financial institutions are getting geared up for major participation in the cryptocurrency markets. Will be interesting to see how that changes the landscape.
To be fair, quant vs. programmer is a false dichotomy since most quants did degrees like mathematical finance, actuarial science, etc. where computer science and numerical programming courses are part of their degrees.
Depends where you go. Some firms have a very bright line between "quants" and "programmers", and you don't want to wind up on the "programmer" side of it.
Agree 100%. First firm where I worked the programmers were treated like monkeys, and the traders, some of whom weren't even that bright at the time, were almost godlike.
Some other person wrote on here about traders being angry assholes. That firm had a ton of angry asshole traders because they'd push and push and push on deadlines, didn't understand the system at all, got hacked up buggy code, never invested in unit or system tests. It was an absolute disaster.
Little anecdote: First year I was out with a more senior trader on my team over drinks. He was talking about the glory days of 08 where every single trader earned a million dollar bonus or more. I asked about the one infrastructure developer who came up with a system to use multicast to simultaneously publish our EuroStoxx signal to DAX along with a dozen other instruments. Nothing earth shattering in 2017, but it quadrupled our P&L and market share at the time.
He got 10 grand.
Thanks for the detailed response. A lot of good stuff to Google.
I can program way better on average than a quant, but what does that matter to a quant
> Eighteen are now enrolled in the program, including Korn.
GS has 34k employees. Thousands of traders and probably >10k in IT. 18 people isn't really a revolution. Devs still are second class citizens in banks - the highest earning ones switched to become traders and just do some python/R/VBA on the desk. There are a few hedge funds (Millennium, Two Sigma) where devs are nearly treated as well as traders - would be interested to hear of any place else.
> There are a few hedge funds (Millennium, Two Sigma) where devs are nearly treated as well as traders - would be interested to hear of any place else.
Software engineers are treated very well, if not as well as research scientists, at essentially any large or successful quantitative hedge fund. This includes firms like DRW, Optiver, Tower Research, Two Sigma, RenTec, Citadel, etc.
Jane Street is probably an exception.
Counter view: I only work in the field because it's the only way in tech to pay rent where I live. I actually hate it, it's of zero value to society and frankly the trading/business side are real assholes under pressure
> zero value to society
What value are financial services? Do you
- have a bank account? checking? savings? money markets?
- monitor your accounts online? do transfers?
- have an investment account? save for retirement?
- have life or disability insurance? pool risks and hedge against them?
- have a city or country that needs to raise money, borrow from others?
- have a company that needs to issue stock for capital investments?
People earn, save, invest, borrow, and build their dreams. Finance makes it possible.
I'm sure you just want to pay the rent, but finance is very competitive and if your firm created no value for society, it would go out of business.
> if your firm created no value for society, it would go out of business.
This is silly wishful thinking. There are plenty of businesses that are sustainably built on screwing people over or simply extracting money from them.
> What value are financial services?
I don't think he's talking about financial services like retail banking. I think he's talking about HFT activities other than market making. Legal front-running, etc.
>screwing people over or simply extracting money from them.
People enter these transactions because there's something in it for them, even if you think it's too little.
So you are upset they didn't say "approaches no value" instead of "no value"?
It’s literally incorrect that businesses can sustainably provide no value. It may be the case that they can sustainably offer deals you wouldn’t take, but how we get from there to the business being immoral, useless, or deserving of a ban is more nebulous.
If people are systematically taking deals you wouldn’t, there might be a broad societal problem (why is tht crappy situation anyone’s best option?) or an information problem (why don’t they know better?) but the business offering the deals is probably not a logical or useful target if you want to address the larger problem.
> It’s literally incorrect that businesses can sustainably provide no value.
That depends entirely on the definition of "value".
If a business exists solely to move money/stock around, that action does not create "value", in the sense that juggling oranges does not create "value".
Trading stocks is an effort to take advantage of the fluctuating relative value each stock has, so that individuals can take home the "profit". Such "profit" is not created, but sifted out of the economy. Is that 100% worthless? No, but it's darn close.
The size of the financial sector swelled to about 9% of GDP in 2010 from around 5% in in 1980. Perhaps that is what the thread was referring to. Nearly doubling in size when finance doesn't produce anything real. They are a middleman. Would certainly like to see nationalization of large parts of finance such as simple checking accounts, insurance, and the big one, lets double down on Social security so that people can have dignified retirements regardless of what their work was. We dont need a big financial sector to have a great economy; Its a brain drain. Smart people should be working on real world problems like diseases, automation, efficient cheaper construction, environmental remediation, etc. that improve the standard of living for society.
A lot of that is simply busy work and can in turn be automated. A huge bank like Goldman is essentially a middleman - and middlemen can also be trimmed down. For example, I was just talking to a friend who works at Goldman as a trader. 75% of his job is to answer the phone and deal with clients issues (eg “why is my account frozen?”). His job honestly sounded like a customer service job. He’s also the person who signs off on wire transfers, something that could easily be automated.
So much of the “value” Goldman supposedly creates is due to them creating an artificial need for it.
So, why do you think it hasn't been done yet? You make it sound like there's a ton of money to be made in this "automating Goldman Sachs out" thing.
You are equating optimization to making money.
When it's about optimizing costs - yes, I am. That's exactly what optimizing costs is about.
I think you're slightly confused at the distinctions between prop shops, hedge funds, asset managers, and model driven quant desks at investment banks. Having worked at at least 3 of these, I can tell you again:
On the sell side, it's just pricing and hedging out each other's increasingly complex structures. HFT Market makers have some merit in providing tight liquidity and low spreads. Don't even get me started on buy side speculation.
If you want to learn more I’d also suggest reading a book called Flash Boys. It’ll give you huge insight into what OP is talking about.
Flash boys is mostly incorrect and at points misleading, and this is coming from somebody whose entire job is trading systems to avoid hft "picking off" my institutions orders/market activity. I can write more once off mobile but there's good material online about why that's better written than what I would write.
Could you please link to the counterargument to Flash Boys?
Thank you, I’d be very interested in learning more!
Can anyone comment on how competitive salaries are there actually?
Right now, for top tier PhD graduates in machine learning, it's about 300-400k at big tech companies starting out, and ~200k for master's degrees.
From what I read and heard, banks are not really willing to pay up competitively for software?
I'm in the range of 3-5 years out of undergrad, working at a quant firm. My total gross comp for this year will be about 400-450k, majority of that is bonus.
By quant firm I mean either a proprietary trading firm (no outside investors) or a hedge fund (takes outside money), but not a bank.
I have a CS background but have self-studied in stats, and received mentorship from people with strong math backgrounds.
Main advantages of quant shops (doesn't apply to banks) over working at Google/FB/Netflix etc:
- small company size, startup style working environment, but big corp pay (or more)
- more varied/interesting work, e.g. mix of distributed data processing, high performance numerics, or performance optimization in latency sensitive code
- compensation paid purely in cash, no stock options/RSUs
> compensation paid purely in cash, no stock options/RSUs
Hands down my favorite part of working in finance. Better salary, no bullshit.
Regardless of what industry you are in, the closer you are to money, the more honestly you will be paid. This is why in many companies some of the best paid people (sometimes more than executives) are in sales. It is quite difficult to tell an employee that you can only manage a small raise if they can say "I brought in $15 million in business this year." If they have no discernible way to determine what they have earned for the company, they can be had for a song.
This was one of the reasons I left the agency world. I was leading the strategy part and pitching for new business (not to mention scoping the work and contract) and won 7 figures in new business we would not have won otherwise. Didn't get a dime in commission or bonus (or even a raise that year) because I wasn't officially on the sales team and was salaried. Learned my lesson there.
I wish we had more salary threads for quants.
I started 4 years ago at a quant shop fresh out of college and expect to make around 250k.
quant at a quant shop, first year made 200-250; expect to be at least 50% higher this year.
How much math did you self-teach and what did you use to learn it? Aside from mentorship.
A few questions, if you don't mind answering:
> 400-450k
Where is that in the pay distribution? Near median? Realistically, where, comp-wise, would the average SWE plateau?
How many hours are you guys putting in a week?
What was the interview like? Did they go deep into stats/stochastic-calc/derivates-pricing/etc ?
---
I'm currently in tech, making near the low end of your range. Getting promoted to the next pay scale (Staff SWE) is a lot more difficult. I'd seriously consider jumping to fintech if 400-450 was below median.
What does it take to make 7 figures at your tech company ?
This sounds really interesting to me. Are there other downsides to watch out for? I imagine the security environment is very restrictive: no work from home? No root access on your workstation?
What kinds of skills do you need to get into this area?
The security varies a lot. At one hedge fund I interviewed at, security was almost MI5-like - no phones in the office, no direct internet access, etc. At the place I ended up working, it's rootapalooza. Not in prod, obviously - there it's sudo-a-gogo.
How did you change fields, what was your transition like? Was it hard to interview, being self-studied and without a degree relevant to the field?
May I ask what is the area you work on? HFT, derivs etc? Also are you a developer or trader?
Sounds interesting and challenging - my contact details are in my profile if you'd like to chat
I'm a senior at an HYPSM school, currently looking at offers from one of {Google, FB} and one of {Jane Street, Old Mission, Five Rings, HRT, Virtu}.
The offers are about the same at about 200-250k depending on bonus and stock performance -- the big tech company upped their stock grant to meet the expected bonus at the trading company. It's not obvious to me which one will be more in a few years, and it seems like hours in tech are better. Can anyone weigh in?
It may be close for the first year or so due to the signing bonus, but 3-4 years in if you're performing at the expected level at a trading firm you should be making double the 200-250k number, and I doubt the tech firms match that kind of compensation (at least for most people).
Another factor you may wanna consider is the opportunity cost -- how hard would it be for you to work at GOOG/FB if the trading firm job doesn't work out? And vice versa?
Face the market you work market hours. Tech doesn't have that limit. Both are cyclic, but FB is probably here to stay. Finance can be very dynamic: think LTCM.
At least two of the financial firms on OP's list are known for working you more than market hours, the average is well into the 60+ hours per week.
Which ones?
A friend works at Virtu and 60 hour weeks are the norm if not more. They're very aggressive wrt managing headcount. Pay is great in absolute terms but pretty low as a % of revenue.
Doesn't mean it's a bad place to work. I have a ton of respect for the business they've built. Their discipline on costs is saving their asses in historically slow market conditions while many competitors are merging or shutting down.
As a Norwegian software engineer, my mind boggles when reading the salary discussions of engineers based in the big cities in the US.
Most engineers here are paid a small fraction of the numbers you cite, even though we are decidedly a first-world country. I guess good ML people will probably get over 100k here, but I haven't heard of anyone near 200k. Any thoughts as to what causes the massive difference?
The dollar is massively over inflated in major US cities. Studio apartments in Manhattan sell for $800-950k. A loaf of decent bread costs $6-8. A beer costs $8 (plus tip). I know that Norway is also expensive, but is it this bad?
This isn't the whole explanation, unfortunately (I really want to understand this relationship). A beer costs more in Norway, a good loaf of bread would be $6 (although you could probably get a cheap one for $2-3), a studio apartment in Oslo would be $300k-ish.
So the biggest price difference is in the rents, or the housing prices. I would probably increase my savings by a factor of five or more if I had a salary like the ones cited in this thread :)
Obviously it's supply and demand, I just wonder which side of the equation is skewed and why.
In Norway you get everything for free, education, healthcare, retirement, etc. I'm sure that's part of the difference
That isn't it; all the things you mentioned are paid through very heavy taxation, and all numbers we've been discussing are gross salary.
You could probably attribute a 14% difference through the "employer tax" which the US doesn't have (which is taken out of your paycheck before the gross salary number is cited), and arguably another 4% due to mandatory retirement savings. But it doesn't make up for it when the difference is 50%-75%.
Yeah, but when you retire in Norway, you will be taken care of. Not so in the US, so you need to save for that from that extra 50%.
supply and demand
I've seen claims of $1M base salaries in NYC, and seven figure bonuses are certainly possible (in HFT). $300K+ (incl. bonus) is no problem, but ML is less of a factor than FPGA and C++ experience.
I work in prop trading and I've never seen that kind of guaranteed salary offered. Maybe as a one time signing bonus to poach someone from a competitor. Typical base salaries for traders/quants/engineers are in the low six figures.
I'm head of a small team running high frequency strategies. I control parameters/portfolio/risk, design/fit predictive models, code C++ for simulation and live trading, and spec out designs for our FPGA developers. I make around 200 base with average bonus in the high 6 figures, good years over a million. At firms with uncapped contractural payouts, bonuses can be far higher for very profitable traders.
But if you come into the industry expecting that, prepare for disappointment. The markets are super efficient so it's hard to make profits. A lot of good ideas are done to death already. You might be stuck on a bad team or not get the opportunity to advance. Making mid 6 figures all-in on a consistently profitable team at a profitable diversified firm is a very good outcome. Modal outcome is making low 6 figures for a couple years and being pushed out.
Also while some enlightened firms realize that infrastructure is a competitive edge and pay big bonuses to engineers, there are many where they're second-class citizens. Only take front-office roles at such firms, not just because of the money, but because you won't be respected. IMO they should be avoided entirely because they'll eventually fall behind and lose.
Thanks for the details. Your depiction is in line with what I've seen personally, but the million dollar base promises did filter in after I left; not for me, but to pass along to people like you, probably.
I'm in the industry. Out of curiosity, which firms do you think do a good job of not treating back-office engineers as second-class citizens?
Finally, some common sense and a realistic response.
Perusing the financial statements of publicly traded trading firms might be an interesting experience for anyone who's not familiar with the industry.
1. How many people work at your chosen company? How big is the entire tech heavy trading industry?
2. What is the average total compensation?
3. Which way have revenues been trending? Revenues per employee?
For anyone who's interested in comparing it to the glory days, look at the historical numbers in the GETCO/Knight S-4.
Complete nonsense. Plus note the intersection between FPGA dev and ML is likely to be practically zero and equally useless at quant firms unless you're in an AI firm building ML on chip. Far easier to go from FPGA to ML than the other way round.
You're right about the lack of intersection between ML and FPGAs. I should have used stronger language.
Yeah not really, many HFs will match that 400k, but many will not. I have received a few offers from these quant funds, the offers have been good, but nothing I couldnt get working in tech. My understanding is that the average finance professional makes less than or equal to tech, but the top 1 - 5% of finance easily make 7 figures.
> My understanding is that the average finance professional makes less than or equal to tech, but the top 1 - 5% of finance easily make 7 figures.
This is more or less true, but an interesting thing to think about is that the number of people (with technical backgrounds) making 7 figures in financial firms is smaller than the number of people making 7 figures in technology firms. This is probably obvious to you if you work in finance, but I think most outsiders don't realize that.
If you include people with sales backgrounds like investment bankers then the numbers might shift in favor of finance.
are you sure? how likely is it that an engineer in a technology firm makes 7 figures? I mean what sort of position do you need to have to get there?
I'm not sure how to answer the question. How would you answer the equivalent question about finance? It's not like working in finance guarantees you 7 figures. I know plenty of people on both the engineering side and the trading side who never made more than 2-300k.
If you want specific examples from tech then the Waymo article from a while back is helpful, but that doesn't mean that you need to work at Waymo or Google to get there, although a large portion of those roles are at Google-tier companies.
Your average agency will not be hiring pure developers and paying them 7 figures, although they might to someone who is generating revenue and bringing in enough business to justify it.
the op made a claim so I am assuming the op knows how to answer this question. The claim is that many more engineers make 7 figures in tech than in finance. My question, specifically, is, what fraction of engineers in tech make 7 figures, let's say what fraction of engineers with 10 years experience and what are the types of positions at those google tier companies that pay this much. I think this is pretty specific
My comment was specifically about the "I mean what sort of position do you need to have to get there?" part of your question.
> My question, specifically, is, what fraction of engineers in tech make 7 figures
I doubt that OP meant that they literally have data showing that it's 1-5%. That's a wide enough range that it's meaningless. I suspect the meaning was more "way above average, but reasonably possible".
I appreciate the insights but you are not in any way answering my very specific question
I think only distinguished engineers get that outright at amazon of which there are only like 10-15.
However, I bet most principal engineers and up who have been here 3+ years are hitting that lately with how good our stock is doing and that amazon pays higher ups mostly in RSU's.
If literally an engineer, not an executive that used to be an engineer or a famous engineer, this is accomplished by winning the stock options lottery.
It seems plausible to me at least that there are more ordinary engineers that have won the stock options lottery across all of tech than the total that are earning equivalent amounts in finance.
Well, there are more engineers working in tech than in finance. The relevant question is the odds of a particular engineer winning the options lottery than 7 figures in finance
Ah that matches what I have heard. HFT/HF is paying up but overall, as someone who is coming up on making the decision, it's hard to see the upsides of finance.
Losing all visibility/ability to publish/being a first class contributor and not second class dev..
I wouldnt get into HFT, its commoditized. The future of finance is in processing huge volume (internet scale) unstructured data and analyzing it in real time. HFT of website data if you will, right now we are in the infancy
HFT has matured.
But, I agree, finding alpha through alternative means at internet scale has been around for at least a decade, but things are getting more interesting now.
How large do you think that population of people in finance making 6 figures is in raw numbers? When you compare that to the tech industry, there are certainly high level engineers and managers that make 7 figure salaries all things included (especially given the growth in stock price at certain companies).
"How large do you think that population of people in finance making 6 figures is in raw numbers?" you mean 7 figures? if you are a top engineer in finance with many years of experience, you are likely making close to 7 figures.
So how does a mid-level software engineer get into trading? I do paper trading on my own for fun but I'd like to make it a career.
Depends on what you mean. If you're looking at tech trading of stocks from home, quantopian is a great place to start. wallstreetoasis forums for a traditional roles for juniors and people trying to get in.
And just remember its a shrinking area. There are lots of traditional traders that have been laid off and are looking for work.
Is quantopian a viable platform or are you just going to food for the bigger fish in the pond?
Personally, i am working on building a cryptocurrency trading platform, because the API's are easily accessible - but the markets entirely different in behavior. Still, rolling my own implementations of technical indicators has been pretty insightful, and backtesting has shown whats worked and what hasn't in given market conditions.
but i realize you're probably talking about more traditional markets like FOREX
What are you using to backtest? Data you gathered yourself?
Hardest part with backtesting in my experience has been trying to simulate successful buy/sell orders and the fees.
You can query Bittrex for some data, or you can find some data and a script from here https://github.com/gcarq/freqtrade/tree/master/freqtrade/tes...
Couldn't you bake the fee into the application logic e.g. https://github.com/gcarq/freqtrade/blob/master/freqtrade/mai...
> Couldn't you bake the fee into the application logic
You can (and I do as well) - although some proprietary bots have had notable problems with figuring them out.
Fee isn't limited to what you're paying the exchange, you have to take into account slippage - the price you use in your backtests is NOT the price you could actually realistically buy at because as soon as you buy, you move the price (especially in size).
Strictly speaking, you should be incorporating level 2 order book data to properly gauge fees and simulate real world profits from your backtests.
poloniex has historical data you can use. I trade on bittrex, but their data doesn't go back very far - enough to do calculations in the now, but not nearly enough to backtest against.
Same way you get into any industry. Get a job at a place that trades.
Like any industry it’s best if you have contacts there to get you past the HR gauntlet but even if not they advertise jobs like anywhere else.
I am also interested in the answer to this question.
Step back and consider the immense intellect and creativity being devoted to just making money.
Or rather reallocating money, when you really think about it...
Same could be said for "just finding cat videos."
Hardly. How many people, MIT grads, for example, spend 10 hours of paid work day "just finding cat videos"?
Say “getting people to click on ads” if you like it better. In the end it’s just making money as well. I guess working at a loss making company can be much more satifactory from a moral point of view but it’s not sustainable...
Yes, but step back and consider the immense intellect and creativity being devoted to getting laid, or speculating about what's going to happen on Game of Thrones, or thinking up clever things you should have said half an hour ago, or complaining about what's wrong with Star Wars these days.
Or arguing on the internet.
Just a teeny portion of our species' intellectual output winds up creating or discovering anything of real value, but that's fine, we muddle through nonetheless. The finance industry has its role to play in ensuring the lights stay on and there's food in the supermarket.
I don't think intellect is much of a factor in getting laid. My bet here is that intellect, or at least use of intellect, has an inverse effect on success in that area :).
The biggest difference between my example and your examples is that it's a job. This talent pool is using their working hours (and quite often most of their waking hours) pursuing a goal that provides almost nothing to society.
The "finance industry" that we're talking about here is self-serving and in most cases is having no positive relationship with the lights or the supermarket.
The transportation industry, by comparison, has a profit motive but ultimately turns human creativity and intellect into things that arguably make life better for a large number of people.
This is unfair, finance is basically an abstraction over our resource constrained world. We didn't get to choose not to live in a world with limited supplies of everything, and removing money from the equation doesn't change that constraint.
Given that we've chosen money as our method of resource allocation, it makes sense people who want access to more resources would try to optimise their profit building strategies, and smarter people are going to make better choices. It is why financial hubs, such as London or New York, have more resources to play with and so can provide a higher quality of living.
People aren’t talking about money itself, they’re talking about hedge funds and prop shops and traders, which didn’t exist prior to the 70s and even Wall Street traders weren’t trading pure securities until the early 1900s (it was slaves and real shares before then) Economy worked just fine without the quants, and ever since trading has been a major segment of Wall Street they’ve completely crashed the markets three times and had to be bailed out by the fed (1929, 1973, and 2008). If you take the capital allocation justification at face value they suck at it. They’re losers underwritten by the government not winners smartly allocating capital.
Supermarkets were full of food and lights were on before hedge funds and prop shops existed. The first supermarkets weren’t even financed by wall street but a real estate tycoon with his own money.
Wall Street is a welfare program for traders, underwritten by the taxpayers and paying for the financial sector’s total losses in 1929, 1973 and 2008.
Are there any automated trading firms that are specifically focused on exploiting the unintended combined behavior of multiple systems operating in an environment where they modify both their own and each others environment (the stock market being that environment)? While it might be tractable for an organization to construct their system such that it can project its own consequences to its own sphere of interest, once you mix in other systems operating in the same area all bets are off. And once you have the majority of trades being performed by automated or rules-based systems, you ought to detect their parameters through assuming reasonable things about those systems (processors which are limited to a certain number of operations, bounding the latency on their response to change, limitations on the complexity of the systems or rules being used due to needing to be comprehensible to human analysis in its source form, etc) and looking for predictable patterns created primarily by the interactions between those automated systems. Everything I have read about automated trading systems (which of course is always very light on details due to everything being extremely valuable and proprietary) seems to indicate that most automated systems are designed as if they are operating in an environment where the other traders are primarily humans. If anyone has come across indication that they are treating the movements of the market as the output of deterministic software systems rather than 'natural' trading, I'd be appreciative of a link!
Depends how you mean exploit. Not a lawyer, but to a first approximation, doing anything in markets to elicit a response from others rather than for bona fide economic reasons is illegal market manipulation so active exploitation is rarely done at serious firms. It's usually individual day traders or offshore boiler room type groups.
Passive exploitation, as in detecting predictable future actions, is probably common, if unintentional. Most automated traders are searching for anything that predicts an imminent price change, on average. Predicting future actions and predicting prices are roughly equivalent. As you move along the continuum from straightforward arbitrage to esoteric black box quant stuff, it becomes more and more difficult to understand how or why a trade even works. I'm sure there are models that predict prices very well solely because they've locked on to some combination of tells unintentionally leaked by another algo.
Here's an example of someone who built a model to detect algorithmic iceberg orders reloading: https://mechanicalmarkets.wordpress.com/2015/04/30/market-da... - he makes an interesting point that any quant data mining order book event sequences that predict strong price movements would find the same pattern easily.
Hi. Well, yes, lots of people has come across that fact. But the guys able to exploit it are market makers. They look for kinds of repeated patterns in the form of orders, stops, time of day... And they later trick on them until they stop working. As you noticed, knowing you'll get someone to buy your thing at a relative price at some specific time and day can be very profitable. Still, it's difficult to read something "worthwile" about this.
A datapoint for the argument that "technologists can learn business easier than business folks can learn technology" camp.
I wonder if this will see an ability to more easily switch into fintech/banking as a developer working in various other industries. I worked with GS via a mobile agency for a while and really respected the developers and what they were doing from an engineering PoV. I had always viewed that industry as essentially closed unless you got in straight after graduating.
Has anyone ever made the switch into an engineering role within a bank? Is it hard to get past the lack of specific "financial" experience if you have X years of software engineering behind you in general?
+1 for an answer to this question. Fintech seems to me to be as walled off to outsiders as game companies are, only tending to hire amongst their own. The job descriptions seem to require more specialized "arcane wizardry" skills than the normal above-average C++ or mobile developer. If you're a commoner who broke into the biz, I'd be interested in learning your story.
Not working on a fintech but FWIW I work on trading platforms for a well-known investment bank. Getting in was not particularly difficult but even with some experience you may need to network with someone on the inside to get your resume in front of the right recruiter, simply submitting your resume via the usual channels appears to be a crapshoot (though I had never applied here because I wasn't aware they were in town)
My salary is quite average for the role/seniority level and AFAICT it's resented by many developers who work here because many will trade their "in" on their industry into more interesting fintech/banking work elsewhere somewhere down the line for higher compensation.
Knowledge of financial transactions is definitely highly specialized and I don't have nearly enough of it yet but it has started to sink in
Would I be correct in guessing from your username that you work with FrontArena? It seems in-house coustomisation of those kinds of systems isn’t paid very well (although I know some murex guys who get paid quite a lot as consultants).
What does the day to day look like working on these beasts? Are you in Europe? Most of the developers I know working on these platforms seem to be quite silod and not really involved with the programmer ‘scene’ as it were (all using windows, no opinion on editors, never heard of hn, don’t care about this weeks $shinynew etc which I know shouldn’t be any kind of grading scheme but it always strikes me as a bit strange).
Would like to hear your opinions on it: are most people who end up working here turned into these (generalisation warning, I know not all) unmotivated souls by the sheer pain of working on this stuff or does it just attract candidates who don’t really care that much about tech?
The username I have carried for some time since I used to play this bomberman clone called Teslatron.
I am not in Europe but a large chunk of the team I'm working with is. Divulging the name of the company is frowned upon, but it's one of the large full-service investment banks
Day-to-day is a lot like other jobs but projects are better managed than I have previously experienced resulting in less meetings for developers. Development can be faster-paced as a direct result.
News of $shinynew is discussed daily by developers (e.g.: React's license, React 16, Vue, etc.) but projects themselves tend to be more conservative as a matter of company policy. I can say the average developer on my team is more capable than the average developer at other companies I have worked with previously. People who are unmotivated will be moved to projects nobody wants (my observation)
Regarding editors, environments, tech events, HN, StackOverflow, etc.: participation is likely there but self-identification is a huge no-no
TLDR: It's not all rainbows and there are institutional challenges, but the tech side is not naive
There are plenty of CRUD positions on Wall Street. If one wanted to get into a "strats" type role maybe it would be easier to just join as a regular Sr. Engineer. Get paid above avg salary to write web services or whatever, learn domain knowledge in a less stressful manner, and figure out a way to get into the group you really want internally.
Banks aren't the only financial services organizations out there. Look to get into the sector through smaller firms that do IR, asset management, etc. You'll still be paid competitively, and you can pick up knowledge about the sector. They might even foot the bill for an entry-level cert like the Canadian Securities Course (I don't know of any equivalents abroad but I'm sure they exist). With these under your belt, you will have a leg up in applying to a larger bank.
Where are the engineering recruiting channels? Can't find anything except corpspeak blah blah that isnt even meant to be read
Crypto is doing really well these last two years, it seems like that is the startup equivalent to the investment banking/quant jobs. However with Crypto, it does feel like the beginning of the dot com era. Dot Com == ICO
Well it's a lot easier for a trader to learn enough programming to implement specific algorithms than it is for a programmer to learn about financial markets.
Can someone give me any information on what skill set can one develop for a quant job as a programmer, and what to know for the recruitment process.
I'm coming back to NYC....
There was an amazing article about one of the programmers responsible for inventing CDS and I can't find it again :( it included such gems as putting offal in a grinder and getting steak out of it, anyone remembers it?
This is a native recruitment piece for Goldman and other banks. It was probably orchestrated between GS and BBG.
That said, I think it’s genuine. Computer Scientists who work on Wall Street should not view their role as simply “a person who can write code.”
Lastly, this is from October.
Not only is it a native recruitment piece, but it's telling that they didn't talk about total compensation.
I would think it's pretty good. I get recruiters calling me all the time, a lot of them are quoting £150-200k total comp. That's for a guy with over a decade's experience in financial coding. By contrast on StackOverflow the salaries that you can see tend to be much lower for the non-finance things.
It's not surprising that it's higher than the average job on stackoverlfow. That's about what I would expect. I'm not saying it's horrible, but it's not exactly higher than top paying tech companies.
Any idea what the tech firms pay in London? I don't have many data points for them.
I'd say the typical for a senior engineer is around 70, ranging up to about 100 or so depending on expertise/company. Going to lead/architect you're looking around 80-120 total comp (there's a huge range...). Finance will pay much better like you said, 150-200, but the tech will be pretty bad. People making over 150 out of finance are generally contracting or working for a US company.
The trading-coding positions they’re referring to in that article have total comps that can be far greater than 200k pounds.
I'd put it like this, the bulk of positions are around what I quoted. That's for senior people who aren't managing large teams, just run-of-the-mill squaddies.
There's also a few jobs where the pay is unlimited. Typically you need a track record but you'll also get a cut of the profits.
And then there's management, which is another level as well.
Yeah, I think that’s accurate.
I made the remark primarily because the article discusses career trajectory as one of the motivations for emphasizing the combination of “traders who code.” In essence, they’re advertising that your pay won’t get capped out at the level of “a very good programmer.” Instead, the cap will the higher number of “a very good trader.”
Is it a cost cutting exercise? Because GS is ruthless in cutting costs.