The Age of the Main Street Quant Is Upon Us
I suppose it was only a matter of time before regular guys with a head for algorithms found a place to try their luck on Wall Street. Internet start-up Quantopian exited beta last month and promises to democratize algorithmic trading for the masses.
Emerging from beta in January, Quantopian said it has built a browser-based algorithmic trading platform where anyone "with a mind for finance" can find the tools and infrastructure to learn, create, and test trading strategies, according to its site. Traders can post their quantitative results, measure their results, show their code and allow someone to clone their algo.
You can read more about the company here. I think this is a pretty cool tool, and it makes sense that something like this is finally available to the public. I can see it becoming a sort of GitHub where amateur quants can give the pros a run for their money.
In a Jan. 23 press release, the firm’s founder and CEO John ‘Fawce’ Fawcett, said: “For way too long, Wall Street has kept quantitative finance to itself by hoarding information and providing little transparency or accountability. Fawcett went to say that his firm’s goal is “to dispel that secrecy and grow the quant community by a thousand fold. We welcome talented people from around the world into our community by providing access to the data , infrastructure and mentorship necessary to participate.”
No word on how you keep your secret sauce secret if you're collaborating on the algo with others in the community, but I'm betting the founders have given that plenty of thought.
So what do you guys think? Will this be the birthplace of a blue collar RenTech? Or is this just a place for CS guys to humble themselves in the markets? Any WSO quants gonna throw their hat in the ring and test out a homegrown algo? Can't do worse than Knight Capital, right?
Comments
The FAQ page has info on
The FAQ page has info on keeping your algos secret. Thank you for posting this.
Market efficiency will be a
Market efficiency will be a self fulfilling prophecy.
"A man generally has two reasons for doing anything. One that sounds good, and the real one." - J.P. Morgan
BTbanker: Market efficiency
Market efficiency will be a self fulfilling prophecy.
My 'proof' against efficient market hypothesis: EMH requires common knowledge of some information X, game theoretically defined as an infinite descent of "everyone knows that everyone knows that everyone knows ... X". Each level in this hierarchy is a separate item of knowledge; there are therefore an infinite number of items of knowledge necessary for EMH. Each item of knowledge utilizes a discrete and nonzero number of neurons, of which we have a finite number; therefore no human can accumulate sufficient knowledge such that would imply the results of EMH. And this is even assuming that information is freely available to anyone who would wish to acquire it.
The obvious route of counterargument comes from symbolic logic: while we can't possibly 'know' all that is required, symbolically we have arrived at the conclusions this would imply, so if each individual market participant consented, we could have a market that obeyed this model (notwithstanding the many issues in modeling a discrete reality with continuous models). But this isn't the case -- I myself could cause a violation of EMH if I wanted.
Not to mention that I could barely understand the ramifications of even "everyone knows that everyone knows that everyone knows that everyone knows X" -- I don't think most humans can understand beyond 3-4 levels of indirection.
Edit: more objections to EMH (as you can probably tell by now I have a problem with it)
1) Assumes rational participants
2) More particularly, assumes rational participants with identical goals
3) Assumes knowledge of the future -- i.e. what impact their actions will have
4) Assumes ubiquitous information reaching all participants at the same time
I'm sure amateur quants might
I'm sure amateur quants might be able to write sophisticated algorithms just as the pros do, but I think the problem here is that the average joe will never be able to compete with the infrastructure the pros have. The big banks have entire teams of developers just working to improve the infrastructure, for example, the speed with which the algorithm sends and receives orders to and from the exchange. And we're not even talking about physical advantages. I'm sure the average joe can't rent office space in the same building as the exchange in order to get faster execution. Anyway, this sounds like a cool idea and it will be interesting to see what people come up with.
nontarget kid: I'm sure
I'm sure amateur quants might be able to write sophisticated algorithms just as the pros do, but I think the problem here is that the average joe will never be able to compete with the infrastructure the pros have. The big banks have entire teams of developers just working to improve the infrastructure, for example, the speed with which the algorithm sends and receives orders to and from the exchange. And we're not even talking about physical advantages. I'm sure the average joe can't rent office space in the same building as the exchange in order to get faster execution. Anyway, this sounds like a cool idea and it will be interesting to see what people come up with.
This is assuming people are going to try to compete on the super short time frames. I doubt that what most will be trying to do. I know plenty of guys (including myself) that used to trade on short time frames that have abandoned that to the HFTers. Now they hold hours to weeks.
spoonfork: nontarget
I'm sure amateur quants might be able to write sophisticated algorithms just as the pros do, but I think the problem here is that the average joe will never be able to compete with the infrastructure the pros have. The big banks have entire teams of developers just working to improve the infrastructure, for example, the speed with which the algorithm sends and receives orders to and from the exchange. And we're not even talking about physical advantages. I'm sure the average joe can't rent office space in the same building as the exchange in order to get faster execution. Anyway, this sounds like a cool idea and it will be interesting to see what people come up with.
This is assuming people are going to try to compete on the super short time frames. I doubt that what most will be trying to do. I know plenty of guys (including myself) that used to trade on short time frames that have abandoned that to the HFTers. Now they hold hours to weeks.
Oh yeah, I didn't consider holding for longer periods of time. I think that's definitely a possibility then.
Eddie, the problem is that
Eddie, the problem is that most of the people who can understand this stuff are either doing it somewhere if they're interested in it, or doing something other than finance if they're not interested in it.
The people in the world who can both understand what an autocorrelation function is as well as how to run the primal and dual linear programs number probably about 100,000.
50K of these people work in finance and love it; 50K of these people don't work in finance and love that.
Work hard, play hard.
IlliniProgrammer: The people
The people in the world who can both understand what an autocorrelation function is as well as how to run the primal and dual linear programs number probably about 100,000.
50K of these people work in finance and love it; 50K of these people don't work in finance and love that.
No, the number of people who CAN understand what an autocorrelation function (and other stuff) is numbers at least around 1000x the number you listed.
However, probably only around 100,000 understand it at the moment because others do not care, are lazy, never heard about it etc.
This is more likely to impact
This is more likely to impact quants' careers, long term: http://arxiv.org/pdf/1112.6209.pdf ("Building High-level Features Using Large Scale Unsupervised Learning", by Google Research).
No, the number of people who
No, the number of people who CAN understand what an autocorrelation function (and other stuff) is numbers at least around 1000x the number you listed.
Only about 10% of the people in the country actually understand Calculus. This is based off of the fact that 1/3 of 18-22 year old Americans actually attend college and perhaps 1/3 of those take a calculus course. 10% of those will likely get nothing out of it leaving us with .9*1/9=10%.
That is the base, base prereq to doing this stuff. Then you tack on calculus-based probability and linear algebra, as well as multivariable calculus. Now we're going to get into methods for generating empirical statistical relationships and evaluating time series', stochastic calculus, some moderately advanced economics, and of course all of the algorithms behind that.
I think a hard working guy with a 125 IQ can understand 2/3 of this, but he won't be able to understand it implicitly and holistically on all of the levels he needs to when he's looking at the problem.
So for an IQ distribution with mean 100, stdev 10, only about 0.6% of the population is capable of understanding everything required to generate trading models.
I don't think this changes much. It maybe means a few physics and engineering guys who love research but are very very bored and moderately interested in the markets can do something with their spare time.
I think there's a much better IRR on learning programming than there is on learning stats.
Work hard, play hard.
EURCHF parity: This is more
This is more likely to impact quants' careers, long term: http://arxiv.org/pdf/1112.6209.pdf ("Building High-level Features Using Large Scale Unsupervised Learning", by Google Research).
As well as all traders', for that matter. This machine learning stuff is pretty darned neat.
In thirty years, the federal reserve and US economic policy will also be run by machine learning algorithms.
Work hard, play hard.
IlliniProgrammer: No, the
No, the number of people who CAN understand what an autocorrelation function (and other stuff) is numbers at least around 1000x the number you listed.
Only about 10% of the people in the country understand Calculus. This is based off of the fact that 1/3 of 18-22 year old Americans actually attend college and perhaps 1/3 of those take a calculus course. 10% of those will likely get nothing out of it leaving us with .9*1/9=10%.
I think this is more a matter of what people choose to do, and where they come from, than what they're capable of. The main tools of calculus outside of proofs are integrals and derivatives, which are just a matter of pattern matching and memorization. Applied linear algebra is largely memorizing definitions and algorithms. I think almost anyone who wanted to could apply calculus / linear algebra procedurally (contrast to developing theory), but I agree that those capable of creating new theoretical knowledge are a much smaller set.
nontarget kid: I'm sure
I'm sure amateur quants might be able to write sophisticated algorithms just as the pros do, but I think the problem here is that the average joe will never be able to compete with the infrastructure the pros have. The big banks have entire teams of developers just working to improve the infrastructure, for example, the speed with which the algorithm sends and receives orders to and from the exchange. And we're not even talking about physical advantages. I'm sure the average joe can't rent office space in the same building as the exchange in order to get faster execution. Anyway, this sounds like a cool idea and it will be interesting to see what people come up with.
I think websites like the pirate bay, 4chan, and others that operate with minimal hardware yet can handle enormous amounts of data is proof against this argument.
I will be writing a post about TPB sometime this week.
Check out my Blog
I think this is more a matter
I think this is more a matter of what people choose to do, and where they come from, than what they're capable of.
But our choices and where we come from ultimately define what we're capable of.
For instance, if you don't learn a language before ~20, you'll always speak it with a significant accent.
I mean, calculus is a language in this case. And it's harder to learn and you don't understand it as well if you see it for the first time at 25, 30, or 35. The neural pathways simply aren't there, they take longer to develop, and when they do develop, they're weaker.
A decent quant- one who can get hired on Wall Street- is ultimately a polyglot of mathematics, statistics, programming, and finance. Basically, he can speak ten different languages fluently and without an accent in the same conversation. (mind you that mathematics and stats have several sub-languages). On top of that, he brings insight and creativity and an intuitive understanding to the financial problem.
Not everyone can do that. On the language front, I can speak English, German, and maybe two other languages fluently (with a mild Arianna Huffington-like accent) if I really set my mind to it. I can't speak ten. When I was 18, I could have learned maybe one or two more languages than that, but not ten other languages. I did extremely well on my Verbals for the SAT and GRE, but I'm not as verbally smart as a decent quant is quantitatively smart.
So the bottom line is that:
1.) This probably reduces the value of prestige. If you went to MIT, Harvard, Yale, Princeton, Stanford, or Berkeley, big deal. Can you beat the Math PhD from Clemson?
2.) Prestige didn't matter that much in the first place on the street. If you were brilliant, you got hired as a quant. If you weren't brilliant, you didn't get hired. If you were somehow still brilliant and couldn't get hired by Wall Street but wanted to work in finance, Chicago would clearly take you.
3.) So if you were capable of this stuff in the first place, and interested in finance, you were probably already in industry.
Net-net, this gives the academic or researcher a few more options to satisfy his interest in the markets, but I don't think it changes much else. Who wants to spend a good 4000 hours studying math, stats, programming, finance, accounting, and econ when one can spend 2000 hours studying CS and earn just as much working for Google?
Work hard, play hard.
awawgoian: BTbanker: Market
Market efficiency will be a self fulfilling prophecy.
My 'proof' against efficient market hypothesis: EMH requires common knowledge of some information X, game theoretically defined as an infinite descent of "everyone knows that everyone knows that everyone knows ... X". Each level in this hierarchy is a separate item of knowledge; there are therefore an infinite number of items of knowledge necessary for EMH. Each item of knowledge utilizes a discrete and nonzero number of neurons, of which we have a finite number; therefore no human can accumulate sufficient knowledge such that would imply the results of EMH. And this is even assuming that information is freely available to anyone who would wish to acquire it.
The obvious route of counterargument comes from symbolic logic: while we can't possibly 'know' all that is required, symbolically we have arrived at the conclusions this would imply, so if each individual market participant consented, we could have a market that obeyed this model (notwithstanding the many issues in modeling a discrete reality with continuous models). But this isn't the case -- I myself could cause a violation of EMH if I wanted.
Not to mention that I could barely understand the ramifications of even "everyone knows that everyone knows that everyone knows that everyone knows X" -- I don't think most humans can understand beyond 3-4 levels of indirection.
Edit: more objections to EMH (as you can probably tell by now I have a problem with it)
1) Assumes rational participants
2) More particularly, assumes rational participants with identical goals
3) Assumes knowledge of the future -- i.e. what impact their actions will have
4) Assumes ubiquitous information reaching all participants at the same time
Computers do not have a finite number of "neurons" therefore your 'proof' is false.
"A man generally has two reasons for doing anything. One that sounds good, and the real one." - J.P. Morgan
I'm sure amateur quants might
Work hard, play hard.
BTbanker: Computers do not
Work hard, play hard.
IlliniProgrammer: BTbanker:
I think websites like the
Work hard, play hard.
BTbanker: awawgoian: BTba
Edmundo
Work hard, play hard.
Hey cool, I and a relative
YOU JUST GOT TROLLED
http://www.troll.me/images/red-foreman322/dont-you...
Hey cool, I and a relative
Work hard, play hard.
Most of the strategies on
"I swear, by my life and my love of it, that I will never live for the sake of another man, nor ask another man to live for mine."
Edmundo
"A man generally has two reasons for doing anything. One that sounds good, and the real one." - J.P. Morgan
IlliniProgrammer: EURCHF
"I swear, by my life and my love of it, that I will never live for the sake of another man, nor ask another man to live for mine."
IlliniProgrammer: UFO hit me
YOU JUST GOT TROLLED
http://www.troll.me/images/red-foreman322/dont-you...
I have a feeling there are
"The power of accurate observation is commonly called cynicism by those who have not got it." - George Bernard Shaw
IlliniProgrammer: No, the
jmayhem: I have a feeling
Xepa: jmayhem: I have a
nice reply remember, i was
Work hard, play hard.
IlliniProgrammer: I think
About Google -- is the salary
Work hard, play hard.
IlliniProgrammer: SVP is very
YOU JUST GOT TROLLED
http://www.troll.me/images/red-foreman322/dont-you...
I don't really get how this
Turbo leverage for capital explosion -- BD Capital
My WSO Blog
[quote=IlliniProgrammer]
Illini Programmer- Excellent
Gomez Addams: Illini
Work hard, play hard.
IlliniProgrammer: Gomez
Gomez Addams: True, but it
Work hard, play hard.
as long as their are liars in
"Everything comes to those who hustle while they wait."
-Thomas Edison
Sounds like someone works for
Work hard, play hard.
IlliniProgrammer: Gomez
Gomez Addams: Sure, you can
Work hard, play hard.
IlliniProgrammer: Gomez