Personally I would just identify a technical niche that is highly valued by certain industries, and specialize in that.
For example in high-frequency trading, there is a small niche developing ultra-low-latency FPGAs or ASICs for trading systems. The work in question is fairly straightforward compared to other industries, but most of the people with the right skillset tend to go to aerospace or other big user of sophisticated electrical engineering instead.
The number of positions in the relevant niche is low, but given that there are so few people, they won't care where they are, could sponsor a green card and will pay very well.
I run a job board for FPGA people and hoo WEE do the HFT firms compensate amply. Talking two to three times the salary of a comparable position at a defense or chip company.
I worked in financial trading for a few years in the 1990s. It was the same deal then: pay was great, industry was a boys club run by assholes, work environment was toxic. After a few years I was very much burnt out and had to change careers from sysadmin work to software development as my internal reaction to being asked to do sysadmin had become waves of nausea and hostility. Took me years to get over it.
So if you are tempted to do financial industry work, make sure to price in the risk of lost work time, years of therapy, etc. It was a fascinating experience and I learned a ton, but financially it was not a net win for me in the long term.
No. A large majority of trading strategies don't require the low latency that FPGAs help provide.
There is some truth to this statement though, if such a firm is trying to pursue a certain kind of strategy - to compete on latency, some of your business must be in hardware.
Not really. There are only a few financial instruments worldwide where very low latency matters, and there you'd be competing against people who have invested millions in custom transceivers, so it's unlikely FPGA work on a readymade board is really going to make much of a difference.
But like all industries, there are many factors that contribute to success. FPGA work might help improve things a bit even if you're not the best.
When young, I was asked to take over reporting for a couple weeks while someone went on vacation. It was a full time position that required 7 solid hours a day of work.
I had the position fully automated by the time person got back.
Thankfully they took it well and they got more interesting assignments.
2 weeks of my time saved several years of salary for company. I did this many times.
They are not replacing "needed traders", i.e. people that used to execute trades required by real people or companies, like "buy some Apple stock". These firms are playing their very own game in the markets, ultimately extracing money from the economy, with the supposed public benefit of "creating liquidity" (not needed for real investors who can very well wait a day or two before closing a deal) while actually distorting the markets and obscuring the real value).
The way the market works is that there are people who want to buy liquidity (the buy side) and people that are willing to sell that liquidity to them (the sell side).
It doesn't refer to buying or selling lots but rather initiating/fulfilling orders regardless of side.
The buy side are typically pension or hedge funds that act based on predictions they have about the future to maximize long-term the value of their portfolio.
The sell side are either banks or specialized HFT firms that only know the instantaneous price of things and their short-term correlations (and not how they might evolve long-term), whose goal is to collect the difference between bid and ask price, tabling on the fact they can sell back their inventory to someone else before the price goes against them. They expose themselves to the risk that price moves before they can do that, and the difference in price between the bid and ask reflects that uncertainty.
The sell side is essential for the buy side to function, and the competition between them leads them to them providing the tightest possible margins and therefore the best price for investors.
The HFT players provide the tightest prices by being very fast to react to market changes and get out of their risk, which is why it's highly technology-driven.
Thinking they have no value shows lack of understanding of market dynamics. Without HFT firms, investors would just be paying large commissions to banks instead.
Low latency trading is so crucial to the economy that the market mostly shuts down at 4pm every day and takes weekends and holidays, adding 2^47 nanosecond delays every week without much outcry.
And whoever wins by 20ns gets rewarded the same as the previous guy who was winning by 21ns. The magnitude of latency isn't rewarded, just coming in first, so more and is spent on less an less absolute improvements, none of which have been relevant to human reaction times for decades.
> The magnitude of latency isn't rewarded, just coming in first, so more and is spent on less an less absolute improvements, none of which have been relevant to human reaction times for decades.
So why are the HFT firms earning profits? There is no law that says people are obligated to do business with them. There is no law that says people have to buy at the prices HFT firms are selling at and sell at the prices HFT firms are buying at.
If someone come in and beats them by .001ns they get all their profits. Is .001ns really worth that reward? The utility provided is mostly detached from the reward, it is just winner gets it.
Worth what reward? Once an HFT firm buys a security, they now have to sell it again. Or vice versa if they sell a security. Do you think they earn a profit on every trade? There is risk in every trade, sometimes they overpay, sometimes they underpay.
It makes no material difference to the profits of a buy and hold investor. If it did, buy and hold investors would not trade with HFT firms.
The liquidity provider reward or index/underlying arbitrage reward, etc. Yes there is a stochastic component and various other things, I'm only talking about the edge from lower latency.
Not all markets are deterministic, in some there is some level of jitter so fastest guy doesn't always win anyway.
Some markets also take steps to prevent the speed game too in how they define the rules.
While for obvious big moves it might be all about latency, most events are more about the quality of your pricing, incorporating correlated information, edge requirement and risk management that decides what you go for and what you don't, and for those smarter trades you usually have microseconds to spare.
> Low latency trading is so crucial to the economy that the market mostly shuts down at 4pm every day and takes weekends and holidays, adding 2^47 nanosecond delays every week without much outcry.
To this I would say: grocery stores (which serve as a sort of middleman to a whole host of necessary goods and are clearly crucial to the economy) close overnight without much outcry.
(To be clear, I think grocery stores are far more important to the economy/society than HFTs, but I think your argument for why is poor)
One grocery store doesn't bury another by being humanly imperceptibly faster at returning carts than the other by a microsecond, and if they were forced by the structure of the market to spend tons of money on that before investing in being open 24/7 which real customers care far more about and provides far more value, it would seem really weird and like something was really messed up with market incentives in the grocery store industry.
Sure, but at that point I would argue you metaphor is not very useful.
One convenience store would likely "bury" another if it were one block closer to its customers than the other. I wouldn't say the food store industry was "really messed up with market incentives" because of it.
A better example would be two stores were equidistant and the market structure made it make economic sense for one store to pay enormous amounts to have itself jacked up and moved one millimeter closer.
A block closer is massively human perceivable, but we're talking about stuff far below human reaction times.
Anyone who buys organically is buying after the HFT firms have made their buying decisions on the entirety of the market, it's not possible to "wait a day or two". The deal will be incrementally worse as these firms extract value, your price will be the one that all of these firms deemed "meh".
It's like walking into a grocery store and the banana you really want it 1$ and you'll buy. Because that's the price that highly sophisticated middlemen between the store and you have determined to be good.
The liquidity argument is abstract and unquantifyable enough to be used as a fig leaf for the industry. Real investors who are in it for fundamental values, dividends or strategy don't need to sell in a nanosecond. Now that you can sell in a nanosecond your deal is bad, as in the moment you hear of the Volkswagen scandal all the trading algorithms are done with their work already.
> Now that you can sell in a nanosecond your deal is bad, as in the moment you hear of the Volkswagen scandal all the trading algorithms are done with their work already.
You can only sell in a nanosecond because someone else is willing to buy in a nanosecond. And if your goal is to earn money trading in time horizons of minutes or hours or even days, you probably are not equipped to do that as a retail investor.
Why should you have the right to dump VW after the scandal and not someone else? What about an investor that did not read the news until the next day?
> The deal will be incrementally worse as these firms extract value, your price will be the one that all of these firms deemed "meh".
No one owes you arbitrage opportunity. If you think the banana is worth $1, then it is worth $1. Maybe a nanosecond ago you could have bought it at $.999999, but as long as you still think it is worth $1, then what is the problem with paying $1?
Not the right place to debate further but it seems this topic seems to be underdiscussed or I present my points not well enough. Certainly my goal is neither to earn money in horizons of days, nor to have arbitrage opportunities at all. All of that is nonsense and would assume the legitimacy of people who sit in front of computer screens looking at stock charts. Value creating is a long term endeavour and so should be investing.
If we let go of the mentality to "dump VW after the scandal before someone else" we find ourselves with bad news about VW and new conclusions about a good value for the stock and people buying and selling accordingly after all have ready their morning paper. A small Tobin tax or other technical measures can prevent unethical actors from taking the banana out of your shopping cart and pricing it within nanoseconds.
The supply-and-demand maximalism is a holdover from anticommunist thinking and ignores market distortions like Zillow's real estate buying.
I wonder if this stuff pays well though. I recall someone showing me a position at Jane Street (since this thread was discussing HFTs) with at was looking for a VBA developer to develop some sort of integrations between their APIs and Excel sheets I guess someone in the company loved.
My first though was “there’s no way this is paying what you typically see for Jane Street developer salaries”.
Do you mean meaningful in the bull** jobs way or the moralistic way? I find the moralistic version confusing. I usually hear it applied to CEO pay, which last year was 350x the average worker. But what will ever feel like a fair or just differential? 7x? 3x? And what would be a fair metric? Are dentists less or more than social workers? What about dangerous jobs, are lumberjacks worth more than coal miners?
They don't really create liquidity, but they did narrow the bid-ask spread. The "liquidity" they provide dries up when volatility spikes anyways (although that was always true, so not really a criticism of HFT).
A bit of both. The market is white hot right now, so there's a better than usual chance of getting something. Of course there's still a lot of applicants so it's useful to have the right degree.
The thing to do is talk to headhunter firms. I've always maintained a number of relationships with the leading firms in my area (quant trading) and they'll call now and again. But if you don't know them, get in touch and they'll usually tell you if you're useful to them.
You can learn the basics of FPGA development on your own (I did), but you need to go to a company that uses them to actually get good.[0] Which probably means a defense contractor or maybe a telecom company. A finance firm isn't going to hire some random person who knows FPGA dev unless they have done real world projects.
[0]: I'm sure there are exceptions of people who self taught themselves and are great, but it is significantly harder to break into than software.
It can be useful to have a high-level idea of what people do so that you can show you understand what's asked of you and you can more easily find the right role.
For example in high-frequency trading, there is a small niche developing ultra-low-latency FPGAs or ASICs for trading systems. The work in question is fairly straightforward compared to other industries, but most of the people with the right skillset tend to go to aerospace or other big user of sophisticated electrical engineering instead. The number of positions in the relevant niche is low, but given that there are so few people, they won't care where they are, could sponsor a green card and will pay very well.