Algorithmic trading

You're talking about HFT, algorithmic trading is more than that. Also, try 2 microseconds, not 10 milisecconds.
We talk about this on various conferences, like cppcon.
You have 0 chance to beat us in HFT. In algorithmic trading, it would only take a couple prize-worthy mathematical advances for an individual to be better than the big guys. But you can definitely outcompete human investors and make some money, easily.

>cppcon
Why would this subject be in cpp conferences?

that's just for arbitrage

Why do you think all of the big investment banks have their headquarters so close to the New York Stock Exchange and Chicago Merchantile Exchange? It’s to get a competitive advantage.

I live in Europe, i wouldn't know that, but it's interesting in fact

There talks from various fields on cppcon, including HFT.
youtu.be/NH1Tta7purM

sweet

Just because everyone is making money doesn't mean your computer autotrading is good. It's really not worth the risk and better to just let people who get paid to worry about do it.

it would be immoral

I worked on this for a few months. It's kind of nice because it provides a certain kind of programming practice that's different than normal development especially the kind ordered by clients (unless you somehow are doing it for a living).

tell some stories

He's right.
I'm at Dow Jones right now and they have a product called newswires which is basically quick snippets/headlines, but the idea is that they come out right after an event happens. This information feeds into things like Blomberg terminals for other users/businesses to see.
Now, there used to be another product with was even lower latency but is not offered anymore due to a hard to penetrate market and diminishing returns. Many of the companies in this "low latency information" business even build their own towers to transmit the data faster (on a scale of milliseconds).

The companies that receive this information have advanced taxonomy systems that are able to categorize certain information from the information received (for example company name, any people involved, if it's "good" news or "bad" news, etc).This is then used to make an automatic trade.

So basically there is no way an individual or even small organization will be able to compete toe to toe with these large banks and investment firms. The best option is to just start dumping your money into mutual funds and indexes from an early age and watch the steady growth until 30-40 years down the road.

you work at Dow Jones?

just an internship

pretty nice. you have a CS education or what does your cv look like?

You should attend. It's worth it as the talks are hardly the most interesting part of cppcon.
You can get reasonably profitable algo in a month of playing. You start with a backtesting platform, get historical data and play it to your algo, see how it does. On the lower end, some basic stats is enough for a passive income that will buy you a house. If you want to delve into it, don't read anything about finance though - their textbooks are shit. Read something on DSP, lingebra, calc on manifolds, a bit of topo, obviously more stats and measure theory is always good; apart from the obvious followups, much interest has lately been in algebraic geometry because it's surprisingly easy to compute certain things within that framework.
Don't count on ever getting competitive with the big boyos, there are dozens PhDs working on this.

I'm studying finance and business analytics, but they are betting big on machine learning right now, so they're hiring a lot of cs students.

This is true. I go to an Ivy and I know tons of CS majors being hired by big firms to do machine learning and shit for them. They all told me that they're projects are turning out to be memes, but the CS majors who go work at banks tend to be brainlets anyways.

Too bad I'm too retarded to learn how to code/use computers lmao

Attached: arnold-schwarzenegger-32.jpg (610x402, 142K)

some stock exchanges let you colo a few machines so you don't always need a big building nearby

>I go to an Ivy
everyone goes into one nowadays

Just want to clarify one thing; there's a difference between banks and information companies. Dow Jones would fall into an information company, whereas something like Barclay's (obviously) is a bank.
It's an important distinction to make though because they have very different atmospheres. Information companies generally do not pay as much in salary, but the workplace environment is extremely relaxed and there are tons of benefits for employees. Banks are pretty much the opposite, where you do make a shit ton of money but you also have long hours and a high stress job.

The only interesting thing I could think of is discovering that the exchange I was using could match post-only orders at better than the order's price. I got suspended from the exchange for about a week though following that discovery. I'm still not sure if the weird executions caused my program to do something inappropriate or if it was a bug in their execution processing code and they didn't want someone who knew about it to trade until it was fixed. I guess that's the kind of area where actual industry experience would provide a no-brainer answer.

There is even a documentary about this. The game is rigged from the begging

Or I can just buy and hold like a sane person because I'm not retarded enough to trust the value of historical data or the many other variable information I'm missing. You people are just as deluded as day traders.

because that's already my full time job, and I work on infra R&D, not strategy/quant research.

decent talk, would advise anybody interested in high-perf C++ development to watch this.

you use C++ on the job?

yeah, plus assembler on rare case-by-case occasions.
strategy stuff tends to be written in higher level languages (including domain specific ones) that get transpiled, targeted to GPU offloading, etc., but C++11 and newer is the bread and butter of the industry for moderately long-lived pieces of architecture used across a broad range of trading platform components.

god damn

sounds very complicated

How to break into that kind of job?

through the front door

The industry thrives on specialization of labor, which is why neither I nor anybody else knows enough by him or herself to compete along the lines of OP's troll post. Our quants can write sim jobs, load them up to the farm, and click a button to schedule them that costs hundred if not a few thousands of dollars in computation every time they want to try something. My tiny niche is strategy agnostic infrastructure research, architectural analysis, and low-level x86 software design.

I had a string of pretty good luck in getting progressively more prestigious positions by job hopping, and proving myself in unexpected ways after I started somewhere. I am a way less focused person than average for my industry, but I've had a knack for identifying previously unnoticed design flaws and inventing (or at least rediscovering) new approaches that end up panning out. I don't have the best academic credentials (only a BS from a small but good engineering program) in a field full of MSs and PhDs from Stanford, CMU, etc., so my path is definitely not the standard way though.

Weird that this thread should appear now, just yesterday I downloaded Python and QuantiacsToolbox, but i've no idea what to do from here, apart from a few video tutorials i'll need to start working through. Is it just a meme to get desperate developers to write them trading algorithms to make them richer, or is there any possibility of some kind of a return from doing this? I know there's the carrot of prize money being offered for the top 3 performing algorithms but I haven't a hope in hell of getting to that level at this stage.

Any online trading system is going to copy the profitable strategies and run the strategies themselves. That's the whole goal of the business, to crowdsource strategy development for the profit of the owners.

Even the brokerages trade against traders, because most traders lose money.

Sure, that's understandable, but how much return can you expect from them running your code? 5%? 10%?

>Ask HN: Anyone making money through algorithmic trading?
news.ycombinator.com/item?id=16922538
One dude admits to losing $100k there.

this. however i have considered getting into stocks and/or cryto but it seems risky and also a fairly large commitment.

>buy and hold
Enjoy getting outperformed by rand(). A below-average applied math undergrad can make a profitable strategy. How long that strategy keeps being profitable is another story though.
There are better talks, but this one was the first that came to mind from last cppcon. It was interesting to see how their platform differs from our, at least the little he said or implied.
The platform stuff is incredibly interesting to me, i'm working on building the skillset to transfer from quant to platform SE because quant work is too stressful to me - too much competition forcing me into areas i don't give a fuck about, which is especially problematic since i came from pure math and most quant work doesn't touch that anymore.

Read the black swan, the signal and the noise and look up the efficient market hypothesis.
tl dr; good luck trying to predict the market, even with algorithms. Say goodbye to the money you invested and the time you wasted building it.

Now this has me piqued. Is/was there any way for testing offline? Seems more info could be wrung if the time was there. Then again, industry experience prolly plays into that angle, too.

I wouldn't even stress on that. Don't submit at all and run that shit on paper/virtually to test. Why hand anything over for a wish?

Is this legal..?

Attached: 1516800942994.gif (270x188, 1.78M)

Everyone knows this. Also, latency (ping time) = speed, not (MBps) throughput.

TOS has a history feature you can practice with.

This my account I use for algorithmic trading and anything based on computer predictions. I'd say I am doing alright.

Attached: Screen Shot 2018-07-14 at 12.41.33 PM.png (342x50, 12K)

is there a website where you can invest money?

link?

The pic is from Personal Capital. it is just a monitoring software. I link the software I built to Charles Schwab's software to complete the transactions.

All had the trade level data but I didn't know what to look for.

Literally none of those are systematic trading firms, just banks.

Also, enjoy getting poor. Systematic trading firms have dozens of economists and quantitative researchers with knowledge on financial math and stochastic modelling (and now are also falling for the machine learning meme). It's a field filled with very smart and knowledgeable people, and you still get that a lot of them (half) aren't able to turn in a profit bigger than the SP500 index.

t. did my Master's on this shit

Btw
This guy is correct is your strategy is supposed to be high frequency, taking advantage of market microstructure knowledge; you will get completely destroyed.

>us
Get me an interview plox, I am working as an IT Quant on an investment bank and I want to work on stochastic modelling.

Attached: le_pringles_face.jpg (393x342, 36K)

>topology
>algebraic geometry
Literally what the fuck can you grt from those fields that is useful for quant research that you can't just brute force with numerical methods from stochastic models? Genuinely asking, I'm curious.

You can be my replacement when i eventually transfer. What are your qualifications?
A new look on your data. Alg. geom. methods we explore allow us to see more. It's really just a toy currently though, not many quants (outside of rentec) can even work within that viewpoint, let alone do meaningful work.
I eventually gave up because i work in a more traditional company than rentec where there are no pure math grads apart from me (mostly physics and applied math), so my latest endeavor was into Sobolev embedding on domains without Lipschitz boundary - much friendlier for applied math grads. What we got from this was a big improvement in algorithmic complexity for a whole group of strategies.

Attached: 5672a3f3066c9.jpg (549x480, 20K)

BSc Physics

MSc Computational Finance fron world top 10 uni; thesis (distinction) on stochastic control theory in continuous time, replicating optimization patterns that lead to insider trading

Started in March as an IT Quant working for a french investment bank; working on market risk department though

(Not in the US)

Your own works are on algorithm optimization, though, or am I misinterpreting? It's also what I usually associate with algebraic fields, not about actually getting new dynamical implications from data.
On algebraic geometry, can you give me one example of a "new look" on data? I don't know anything about the field, only hear it as the recent buzzword in pure math (i am not from pure math). A link of a paper would be fine.

arbitrage or TA?

Polytechnique? Get a PhD and no problem. Otherwise my recommendation will only suffice for an interview.
And no problem about the location, i'm yuropoor too, they flew me out.
Well it's not that one-sided. I don't work on algos specifically, this was just the implication of my work (though i did participate in implementing it), i'm paid to prove theorems and explain how others can use them. My work in general results in better performing (both in the complexity sense and in the market performance sense, sometimes both) strategies.
You can see algebraic geometry as a "calculus of a pure mathematician" in that it became immensely useful in virtually everything a pure mathematician typically works with, just like calculus is immensely useful for an engineer. That's why you might feel like it's a buzzword, though how "recent" it is can be debated - it was *very* active ever since Grothendieck offered his interpretation.
As for its practicality, it's hard to conceptualize what exactly it can tell you about your data (my work was on understanding this), but the ultimate effect is better predictions with (hopefully) less computational complexity.
As for links, there's hardly anything specific because, only rentec really has the brainpower to play with it (and they do, through contacts in my field is how i know). But i can link some papers i used during the work, though i ultimately failed to find an elegant way to look at our data through alg. geometry so they are probably just useless noise to you - they certainly won't tell you how to make better predictions.
arxiv.org/abs/1012.3180
arxiv.org/abs/0710.0222

Fuck off fake news shill. Stop fucking shilling your fake news neo day trading lifestyle to the impressionable retards of Jow Forums

Stay poor and blame the jews again.

Attached: 1529657140516g.jpg (400x342, 39K)

>Yes, promising big riches and nextgen statistics and ai will definitely drive customers to come back to my online brokerage firm after day trading was show to be an utter retarded meme

Stop shilling

Works for us™. We won't stop making money just because some asses on the internet blogged about losing their lifetime savings.

Attached: chern.png (456x480, 64K)

you can make more money buying an ice cream truck and hiring someone to drive it around a park all summer
playing the stock market as a single investor with no inside info is just gambling

UCL
>get a PHD
That's basically 3 years, so yeah... For that I'd rather just work in the field and climb steadily while making money. Stopping my career and getting into Uni 3 more years is not exactly an immediate option.

Ok, thanks for your input; although I must say that I am skeptical on the practicalities of such abstract and topological intensive stuff in finance. I mean, even in physics, it's highly debatable if it's useful. Still, I will read more into algebraic geometry, even if not just to be a little less ignorant. I'll start with those two papers, even if not to get an idea of the names of the concepts and theorems.

Still me, by the way. Thanks for the availability, anyways. Talking to a guy on fucking Jow Forums, and being willing to give him a chance, you're a nice dude.

Yeah you're better-off grinding in yuropoor and perhaps applying few years later with a viable backup than trusting some internet shitter.
Hopefully you love your job, because it will eat you otherwise - i'm burned out 4 years in and i'm relatively free at work, as long as i provide results. The past month has been a nightmare though, never had this much of a brain fog. Good luck.
As for alg. geom. Yeah, pretty much everyone is skeptical. Only application fetishists like me are even working on applying it anywhere.
I wouldn't write it off just as easily though, it has proven useful enough to make it to condensed matter physics which is about as practical as theoretical physics (which uses plenty of basic alg. geom.) gets. Of course condensed matter physics is pretty far from finance, but hope dies last i guess? Anyway, i would really love to bring it "down" so that it's more easily accessible to people. Problem is algebraic geometry routinely works with tools that are very category-centric which makes it kind of esoteric and hard to get used to.
The papers are certainly not an easy reading. For algebraic geometry, strong background in commutative algebra is required (and also homological algebra), so Atiyah's Introduction to Commutative Algebra, followed by Shafarevich's Basic Algebraic Geometry I would perhaps be a good reading for sleepless nights.
Maybe more appropriate would be notes from mathematik.uni-kl.de/agag/mitglieder/professoren/gathmann/notes/alggeom/ as they don't assume much prior knowledge.
But i think ignorance in this case is a bliss, you'd do better to just focus on what you like and not take endeavors in these abstract spaces. If you want to deepen your knowledge of math in a useful way, there are topics in functional analysis that would be much more appropriate for finance, despite still being a territory of pure mathematicians.

hi reza

you can't compete with real quants

Thanks for those notes, killer.
At the moment I am studying stochastic analysis more in depth, my field was mainly stochastic processes/calculus, so I am trying to go deeper and I have been studying necessary auxiliary tools as the necessity to know them arises.
Also, theoretical physics is very broad, as you know, and most of it is fairy tales. I still love it.

Because it is my job to do this already. My employer has a lot more balance sheet than I do. I am fine being paid a percentage of the profit I generate.

what a cuck

I do, and did. But not for stocks, for forex.
Took fucking 5 years of my life.

Have been running it on a server computer hidden in inside my closet for almost a year now trading every currency pair except exotics through 4 brokers 24/5. GF thinks my "job" let's me sit at home all day, I actually don't have a job.

It's INSANELY difficult guys but its fucking well well well well well worth putting the effort in if you have a background in programming and trading.

how much you racking in

I have been hitting between 5-14% a month. I made about 12 thousand last month. I live very frugally so I have just been letting it compound with only small withdrawals to cover expenses.

but to earn that you need to have a lot of money first, no?

Being this unemployed

Kinda but not really, I used to make a living trading with a ~20k account manually which isn't very much. The key is just to have consistent growth don't aim for some rediculous 200% a month, just hit a small goal and watch it grow. Like 10% a month isn't much, but when you've grown to a large account it's a lot of fuckin money man.

i hope you have a bunch saved bruh

If you aren't lying, I don't understand how this is possible. Most professional firms, with hundreds of Ivy League PhDs, struggle to consistently beat the market by a meaningful margin. Meanwhile, some random autist on Jow Forums is pulling in 80% ~ 380% (5-14% compounded) per year through a server in his closet?

not him but see and
>Took fucking 5 years of my life.

He's clearly lying

A friend of mine is a trader at a trading firm. They mostly do arbitrage, or strategies that are close to arbitrage.
They have colocation (PCs inside the exchange) and they also don't use a normal internet connection. They use microwave signals, so they can be faster than fiber and save those milliseconds.
There is no way to compete with that, the reasonable thing would be to invest in an index fund, mirroing something like the SP500 and just let the money sit for a decade.

I was thinking, that maybe crypto trading could still be profitable, but not sure about that either.

Just buy the dip

Attached: 1506283726634.jpg (600x494, 30K)

>higher level languages
Any FPLs?

see

No idea. The documentation is at headercon in a different location.

i remember an interview with someone talking about these kinds of computers/programs which he claimed "control" the entire market. i don't know enough to believe him or not but if it is true why don't we see tons of compsci majors worth billions and billions? or is it just not that lucrative yet?

He may have been talking about how most trades on stock exchanges are from automated software, not humans with keyboards and mice clicking BUY and SELL.

yes

Heh.