• C++ Programming for Financial Engineering
    Highly recommended by thousands of MFE students. Covers essential C++ topics with applications to financial engineering. Learn more Join!
    Python for Finance with Intro to Data Science
    Gain practical understanding of Python to read, understand, and write professional Python code for your first day on the job. Learn more Join!
    An Intuition-Based Options Primer for FE
    Ideal for entry level positions interviews and graduate studies, specializing in options trading arbitrage and options valuation models. Learn more Join!

Algo trading

Joined
9/5/10
Messages
397
Points
38
I am interested in a career in Algo Trading and strategy post MFE?
Which courses would be relevant for that? Any program with particular strength in that?
 
I'd like to know that to, please. What electives at MFE I should be taking if I want to go algo trading route after graduation?
 
Time Series Analysis is an essential course.
I remember reading on wilmott forum a post by Dominic about new courses at Baruch MFE since Jim Gatheral's arrival. He basically said that Baruch has the "kernels of a specialist algo trading program".
Unfortunately, Wilmott removed that thread so i can't link to that post.

You can take a look at their new curriculum here Baruch MFE Curriculum | Master of Financial Engineering Program at Baruch College
MTH 9867 Time Series Analysis and Algorithmic Trading (Spring 2011; Instructors: Jay Damask & Jim Liew)

It's possible that other programs have introduced similar courses but I don't know.
 
Time Series Analysis is an essential course.
I remember reading on wilmott forum a post by Dominic about new courses at Baruch MFE since Jim Gatheral's arrival. He basically said that Baruch has the "kernels of a specialist algo trading program".
Unfortunately, Wilmott removed that thread so i can't link to that post.

You can take a look at their new curriculum here Baruch MFE Curriculum | Master of Financial Engineering Program at Baruch College
MTH 9867 Time Series Analysis and Algorithmic Trading (Spring 2011; Instructors: Jay Damask & Jim Liew)

It's possible that other programs have introduced similar courses but I don't know.

Time series and statistical arbitrage is taught by Rob Almgren at NYU-Courant as well. Some of the materials are available online.
http://cims.nyu.edu/~almgren/timeseries/
From the curriculum, the class may include only a small portion of algorithmic trading.
 
its a scam if someone says that they can teach algo trading in school. The only things you can learn in school are the tools needed like statistics. only market practitioners know how these things work and besides, what you learn in a class room is quite far from reality.
 
its a scam if someone says that they can teach algo trading in school. The only things you can learn in school are the tools needed like statistics. only market practitioners know how these things work and besides, what you learn in a class room is quite far from reality.

The course taught at Baruch is by practitioners. One works for a hedge/quant fund trading using algorithmic strategies.

The other was a statistical arbitrage trader for a hedge fund and recently went off to start his own quant fund sort of company with a few other people from other funds.

I don't think the point of the class is to teach you how to make money. It is how to develop effective strategies from an infrastructure point. That is one part at least. The second is to expose the students to several automated strategies that the students will develop themselves from scratch.
 
Time series and statistical arbitrage is taught by Rob Almgren at NYU-Courant as well. Some of the materials are available online.
http://cims.nyu.edu/~almgren/timeseries/
From the curriculum, the class may include only a small portion of algorithmic trading.

Nice find, unfortunately, he doesn't provide slides for the three of his Trading strategies lectures though, only for the times series part of the course.
 
The course taught at Baruch is by practitioners. One works for a hedge/quant fund trading using algorithmic strategies.

The other was a statistical arbitrage trader for a hedge fund and recently went off to start his own quant fund sort of company with a few other people from other funds.

I don't think the point of the class is to teach you how to make money. It is how to develop effective strategies from an infrastructure point. That is one part at least. The second is to expose the students to several automated strategies that the students will develop themselves from scratch.

if(heads)
buy();
else
sell();
 
Alexei, random walk doesn't necessarily imply 50-50 odds, I would go with
int result = rolldie();
if(result>2)
buy();
else
sell();
 
Currently there is no MFE that really aims to produce algotraders, though I have noticed a distinct pattern in the way Baruch is developing.

Latency has some of the truth, in that algotraders are very secretive, but we part company in whether it can be taught. Certainly it is hard, and you have to be realistic about the level you can attain by being taught.

You can teach econometrics, market microstructure, signal processing, C++, low level IP, game theory, various types of probability theory, gambling maths, pattern matching, dynamic programming, optimization and social engineering.

Supplement that core with options FPGAs, behavioural finance, linux kernels, technical analysis, FIX, and various bits of AI.

Whether you can do that in a year is highly questionable, regardless of the quality of teachers or students. Would be fun to try...

What you'd have (for those who survived a brutal learning curve), is the ability to understand what is going on. But that's like being able to read music, maybe even to sight read and play, but not to write your own tunes. Ultimately this is a creative endeavour, since by necessity there really isn't all that much utility in faithfully copying what others have done.

All the effort is wasted unless you happen to be creative.

I take it we have all seen the film 'Amadeus' ?
 
Currently there is no MFE that really aims to produce algotraders, though I have noticed a distinct pattern in the way Baruch is developing.

Latency has some of the truth, in that algotraders are very secretive, but we part company in whether it can be taught. Certainly it is hard, and you have to be realistic about the level you can attain by being taught.

You can teach econometrics, market microstructure, signal processing, C++, low level IP, game theory, various types of probability theory, gambling maths, pattern matching, dynamic programming, optimization and social engineering.

Supplement that core with options FPGAs, behavioural finance, linux kernels, technical analysis, FIX, and various bits of AI.

Whether you can do that in a year is highly questionable, regardless of the quality of teachers or students. Would be fun to try...

What you'd have (for those who survived a brutal learning curve), is the ability to understand what is going on. But that's like being able to read music, maybe even to sight read and play, but not to write your own tunes. Ultimately this is a creative endeavour, since by necessity there really isn't all that much utility in faithfully copying what others have done.

All the effort is wasted unless you happen to be creative.

I take it we have all seen the film 'Amadeus' ?

Dominic,

Would you say that an individual with a PhD or extensive experience would be better suited or have a better chance at algo trading? Or would a MFE be enough assuming said individual was creative?

Amadeus was a good movie but I read it wasn't very historically accurate, which given Hollywood isn't surprising.
 
>Would you say that an individual with a PhD or extensive experience would be better suited or have a better chance at algo trading?

Yes, if the PhD was in the right area.

>Or would a MFE be enough assuming said individual was creative?
In no case is an MFE adequate preparation.

An MFE on top of the right undergrad degree might do it, but MFEs simply aren't the right qualification. At the risk of over-simplifying, an MFE has net negative value if you are focused on algorithmic. Some hiring managers explicitly say they don't want MFEs, others feel that the content is largely irrelevant and that if you were smart enough to do this stuff, you could pick up Wilmott and teach yourself.

Also, no finance course anywhere teaches you C++ enough to get through a serious interview for these roles, and I include my own work in that set.

The typical MFE has inadequate coverage of game theory, and many don't cover market microstructure at all. Some MFEs look blankly at me if I talk about Kelly, which shows at the very least that they aren't properly prepared, and often that they aren't up for it.

But I'd like to clarify what exactly I'm talking about here...

I mean the cream jobs, where compensation may be explicitly unbounded, where you get to devise algos and play with real money.

Although algo teams tend to be smaller than in some other types of trading, there is any amount of work in managing risk, building components, data mining, testing & debugging, backtesting, and dealing with the golf players in central IT.

They're not bad jobs (usually), and there is a realistic prospect of moving to the front line. These are attainable by MFE grads.

One other thing I'd like to make clear is that doing a PhD for purely career reasons is all the way dumb. You're typically buying >4 years of misery and penury, unless you have a genuine interest in the topic you are pursuing.

I've toyed with starting a research group to do this, and even got the head of one university department to tentatively agree to host it. But it will probably remain a pipe dream because I have found that what people want is for me to explain to them how to be algotraders. Sure I know rather more about this than any other headhunter, but the key term is that I'm a HH who's done more different jobs than is good for him, not an algotrading veteran. The group would start with
main()
{
}
A pile of data to screw with, random people to pick the brains of, but no lectures or spoonfeeding at all.

I have been disappointed but not surprised to find at this point, interest is lost quickly.






Amadeus was a good movie but I read it wasn't very historically accurate, which given Hollywood isn't surprising.
 
You can teach econometrics, market microstructure, signal processing, C++, low level IP, game theory, various types of probability theory, gambling maths, pattern matching, dynamic programming, optimization and social engineering.

Supplement that core with options FPGAs, behavioural finance, linux kernels, technical analysis, FIX, and various bits of AI.

Whether you can do that in a year is highly questionable, regardless of the quality of teachers or students. Would be fun to try...

I doubt that even a PhD student can properly learn all those subjects in 5 - 6 years. I also doubt that majority of algo traders posses a deep knowledge of those subjects.

There is no program that teaches such a mix of subjects. A PhD student majoring in a quantitative field would have to take all those courses as electives if he/she decides on algo trading career. And taking a course or two doesn't count as having a thorough knowledge.
 
>I doubt that even a PhD student can properly learn all those subjects in 5 - 6 years. I also doubt that majority of algo traders posses a deep knowledge of those subjects.

I agree.
But this is a competitve game, if you want in to it, then these are things you will probably be asked about at interviews. My list was partly to illustrate how different algotrading is from traditional MFEs.

If anything, I may have understated the scale of the issue. I believe within the timescale of a PhD you can get a basic grasp of the things on my list. But you also need to good deep on at least one area, being mediocre across the whole range is probably not useful. That of course implies that a week of market microstructure may be easily enough if you are a whizz at signal processing.

You are also right that many current algotraders have big holes in the topics I list.

But the vast majority of ATs did not optimise their education with respect to getting that sort of job,
In some cases they were quite suprised that their skills happened to be rather valuable.

So, it is entirely possible to get an algotrading job with just two of the list above, if you do them very well. But not certain, and I am talking about how you get an AT job on purpose and how it might be taught to you. That's different from stumbling upon it.

Also, I am looking forward, not back. Some senior quants at some big banks only have an undergrad degree, no masters or PhD, one MD level quant used to be a computer journalist and no PhD/MFE either. But if you applied to them with just a degree, it would be very tough to get a quant job with them without experience or more education. The market has been subject to inflation in education.

Given the forum I am on, I therefore talk of what you should do to get from a position where you aren't qualified for a good algorithmic trading, to being educated in a way that gives you a good shot.
 
My list was partly to illustrate how different algotrading is from traditional MFEs.

So, it is entirely possible to get an algotrading job with just two of the list above, if you do them very well. But not certain, and I am talking about how you get an AT job on purpose and how it might be taught to you. That's different from stumbling upon it.

I see your points.

Here is Dominic's inventory of Algorithmic Traders' skills:

econometrics,
market microstructure,
signal processing,
C++,
low level IP,
game theory,
probability theory,
gambling maths,
pattern matching,
dynamic programming,
optimization,
social engineering.
FPGAs,
behavioral finance,
Linux kernels,
technical analysis,
FIX,
various bits of AI.

May I add Machine Learning.

Also, what about financial skills, don't ATs have to have finance knowledge to be able to devise new strategies?

Corporate Finance and Valuation are not that useful, I guess. Investments, Fixed Income, International Finance, Derivatives - these are useful, are they?

Does statistical arbitrage considered algorithmic trading?
 
Back
Top