low latency trading

[siphenx]

Hello all, I just started my career and have a seemingly good option at hand. But I have some concerns and would like you to give me some feedback.

I have an opportunity at a 2nd tier IB as a trading strategist. But it's basically an IT role where I work with quants and traders to help implement, optimize, test and monitor their algorithms. I'll do a lot of very low level hardware related stuff to shave down latency and achieve high performance. C++, Python, bash, sockets, multithreading, parallel processors ... will be my friends.

However, my background is Physics in both undergrad and grad school. As a result, I have very good critical thinking ability and quantitative skills. On top of that, I'm heavily interested in finance, especially investment. I also did an internship at a hedge fund a while ago and am working towards CFA. So I do not really want to do IT for the rest of my life, although I don't mind starting off as an IT guy as long as I can switch to real trading/quant trading/algorithmic trading in a year or two.

But here is the dilemma. I think I'll probably get to talk to some people in the field every once in a while. But I'm not sure if I'll have the time to network as my role is highly demanding and stressful and most of my time will be spent picking up technical skills. I did a lot of computational stuff over the course of grad research and have very solid knowledge on DS and algorithms. But I've never taken any course on computer architecture or network. So if I take this role, I'll have to spend at least the entire first year boning up on those topics. And if I jump ship in a year or two, all the time spent will go wasted.

Should I give it a shot or pass it up? Any feedback would be appreciated.

 

[Ian Kaplan]

If you want to do quantitative/algorithmic trading and you don't like being deeply immersed in all issues involved with software development, testing and deployment, in my opinion you should not be trusted with money.

Most statisticians will tell you that 80% of their work involves data. The other twenty percent is statistical mathematics. But most people who do real, complex, statistical analysis are mired in data and mundane data issues like data quality and cleaning.

Finance is the same way. It is very easy to make a mistake and not know it. It is very easy to have unreliable software. This makes software development skills a critical skill set.

Quantitative/algorithmic trading groups in many ways resemble software groups.

In this work I don't think that you're going to get a lot of change to exercise your "quantitative skills" outside of implementing and testing them in software. There are a few exceptions, but they are generally people like options geniuses.

 

[siphenx]

If you want to do quantitative/algorithmic trading and you don't like being deeply immersed in all issues involved with software development, testing and deployment, in my opinion you should not be trusted with money.

I don't disagree with you that quantitative/algorithmic trading involves a lot CS stuff. But I don't think it's the same as pure IT development. When I was interviewing with the firm, they asked me multiple times if I had any intention to become a trader/quant because although they work closely with traders and quants (they actually sit together), they are a pure IT team and what they do is to automate and optimize strategies designed and code written by traders and quants and to make sure all network servers function well. My gut feeling is this is not a position that can lead me into trading.

I guess my real question is that if what you said is true that algo trading resembles software development a lot then why there's a pure IT group in the trading team where once you get in you can hardly switch into trading?

 

[Ian Kaplan]

There have been some spectacular software blowups. The most dramatic being Knight Capital which lost $400 million in 40 minutes. There was also a software error at Axa Rosenberg, which lost about $200 million. Barr Rosenberg, who did seminal work on portfolio theory and factor models, will, sadly, be remembered for not being honest about this error. As a result, he is banned from the securities industry for life.

Just because quantitative trading groups resemble software companies and should act this way does not mean that they do. Nor does it mean that they treat software engineering with the respect and focus that it deserves.

This said, there is more complexity in what you write. A large scale trading operation has many components. For example, trades have to be entered, the trade book has to be visible, there has to be risk visibility. In addition there has to be back-office resolution of trades, capital and expenses (e.g., interest). And many other pieces that I'm sure I've left out.

This software is complicated and has high requirements for quality. But it has nothing to do with trading.

Then there is trading model development and testing. The groups that do this are, indeed, different from the groups I've described above. But the demands seem to me to be greater. Numerical mistakes are harder to catch than straight algorithmic errors. Or at least I find this to be true. If it were my call, I'd have two independent groups developing the same mode and then compare the results. I know that some "shops" do this with options models. I think that it would be worth doing with any quantitative model.

I don't know how easy it is to move from developing all of the support software to the trading software. It may be difficult because the groups don't interact that much, especially at the level of software developers. On the other hand, it's hard getting jobs in finance these days. So it might be worth taking the job if you don't have an offer for a trading position.

 

[mhy]

I don't disagree with you that quantitative/algorithmic trading involves a lot CS stuff. But I don't think it's the same as pure IT development. When I was interviewing with the firm, they asked me multiple times if I had any intention to become a trader/quant because although they work closely with traders and quants (they actually sit together), they are a pure IT team and what they do is to automate and optimize strategies designed and code written by traders and quants and to make sure all network servers function well. My gut feeling is this is not a position that can lead me into trading.

I guess my real question is that if what you said is true that algo trading resembles software development a lot then why there's a pure IT group in the trading team where once you get in you can hardly switch into trading?

What did you say when they asked you that question?