4 ms·

Quantitative trading seems interesting because I would hazard that you need strong math background along with solid hacking skills. There are other problem dom

by tom_b 13y ago

Quantitative trading seems interesting because I would hazard that you need strong math background along with solid hacking skills. There are other problem domains with similar requirements, but I'm quite curious to hear stories from other HN'ers who have tackled quantitative trading either for themselves and in the day job. Specifically, if you can share a bit on your technical work in such a role, that would be cool to hear about.

shogunmike 13y ago
I was a quantitative developer working alongside a quantitative trader in a small 'proprietary trading fund'. This is roughly how all quant funds are separated. There is the "technology" side, which builds the data/trading infrastructure and then there is the "research" side that generates the trading strategies to run on the infrastructure. My job involved anything from hooking up to brokerage APIs to optimising MySQL replication topologies. Quite varied! In a way, it wasn't too different from your average startup, with the possible exception that you deal with a non-trivial amount of data from day #1 (hundreds of millions of rows are not uncommon). Open source has gained significant ground in funds these days. Python/R are now the "default" go to languages for quant trading research, with some MatLab too. Libraries such as NumPy, SciPy and pandas have really brought 'algo trading' to the 'retail' (algo) sector as well. .NET is still generally used quite a lot in investment banking, particularly C# for front-office GUI code, and C++ for any legacy number crunching libraries.
alexkus 13y ago
> Quantitative trading seems interesting because I would hazard that you need strong math background along with solid hacking skills. Strong maths knowledge can be good for general developer roles anyway. It's one (small) reason why I did a maths degree part time (Open University in the UK) to complement my existing Comp Sci degree. I'm sure you could get a better bang per buck picking and choosing specific areas rather than an entire degree but that wasn't my motivation (I'm equally interested in both in general). A good number theory course (even just plowing through Crandall and Pomerance, or Apostol) will definitely help with understanding/analysing asymmetric encryption.
- shogunmike 13y ago
  I completely agree with this. While you may not be solving partial differential equations in your average tech startup, there are plenty of instances where a maths degree can be directly applicable. Statistics is one instance, for A/B testing. Another example is the use of vector calculus in machine learning and "data science". How did you find the OU degree?
  - alexkus 13y ago
    > How did you find the OU degree? It was good. I took it almost as slowly as you can (typically one module per year) so it was 8 years from start to finish. In that time I got married, moved house twice and became a father so I wanted to avoid it taking over my life. Tutorials were local to home or work (I was lucky in that respect) but they did seem infrequent. Tricky to recommend now the fees have quadrupled though (I paid about £4k in total for my degree, it'd be closer to £15k now); unless it's your first degree and you're considering the OU over a traditional university; then £5k a year is quite cheap as you can be much more flexible with life and (part/full-time) work. I was considering either continuing with the OU with Maths on some of the Master's courses (those fees aren't subsidised in the same way that UG courses are), or switching to languages (French, German, Spanish) but the prices put me off those. For now I'll have a year or two off as a break.
    - bencollier49 13y ago
      I agree, the OU Maths degree is excellent, but it's been ruined by the new pricing arrangements. None of the really interesting people I met on the course would have been able to afford the new fees.
    - Tycho 13y ago
      I've been toying with the idea of doing an OU maths degree. Not for the credentials but because I'd like to be good at maths... I hate not being able to understand it well. However I'm dismayed to hear the cost is now £15k.
  - jaymzcampbell 13y ago
bachback 13y ago
tom_b, basically every shop cooks their own solution for just about everything. if you do mid- to low-frequency a Bloomberg will suffice. if you do high-frequency you are in the infrastructure game. you co-locate to the exchanges and battle for the milli- to microseconds. there are a few specialized databases in this area for dealing with TBs or rows. everything else in the area of finance is a complete waste of time IMO. basically the financial system is toast. I hope that bitcoin will improve some of the absurd structures we have set up. anyway, if you want to know some specifically you can go to nuclearphynance. There are some cracks around there. good traders would consider an MFE as the opposite of helpful, I'm sure. standard stochastic calculus is pretty much useless.
- makerops 13y ago
  There are fully managed solutions to receiving market data etc, that a lot of companies do use, both in the enterprise, and HFT realms (ie tickerplants).
  - bachback 13y ago
    Yeah, for 100k$ or more a year. So you need a lot of capital to get started on your won. Getting started with a private account is hard to impossible in my experience. Lime Brokerage offers good stuff for HFT. You need about 12 months to get really going.
- ruang 13y ago
  What is the difference between strategies used in mid/low frequency vs high-frequency? I understand high-frequency is mostly stat arb but I haven't heard of too many quant funds that do mid/low frequency. I've seen one present before that used small changes in portfolio optimization to get an alpha of ~0.2% over the benchmark but that didn't seem too useful.
nvarsj 13y ago
I worked at an HFT for several years as a developer on the quant lib team. I'd say the algorithm side is the hard and interesting part, while the math is relatively straightforward and solved. For example, it's well known how to price an option - but doing that efficiently and at large scale is the challenging aspect (think kd-trees, etc. for sophisticated caching, messaging protocols, dbs for organizing data). I also built a sophisticated graph algo for relating securities, which built the core of one of our most successful strategies. So technically, I found it very appealing. However, my experience wasn't so great otherwise. You really get the sense that people are only in it for the money - everything revolves around the year end bonus. And at least where I worked, this kind of huge money drove a lot of really nasty politics.

by tom_b 13y ago

shogunmike 13y ago

I was a quantitative developer working alongside a quantitative trader in a small 'proprietary trading fund'. This is roughly how all quant funds are separated. There is the "technology" side, which builds the data/trading infrastructure and then there is the "research" side that generates the trading strategies to run on the infrastructure. My job involved anything from hooking up to brokerage APIs to optimising MySQL replication topologies. Quite varied! In a way, it wasn't too different from your average startup, with the possible exception that you deal with a non-trivial amount of data from day #1 (hundreds of millions of rows are not uncommon). Open source has gained significant ground in funds these days. Python/R are now the "default" go to languages for quant trading research, with some MatLab too. Libraries such as NumPy, SciPy and pandas have really brought 'algo trading' to the 'retail' (algo) sector as well. .NET is still generally used quite a lot in investment banking, particularly C# for front-office GUI code, and C++ for any legacy number crunching libraries.

alexkus 13y ago

> Quantitative trading seems interesting because I would hazard that you need strong math background along with solid hacking skills. Strong maths knowledge can be good for general developer roles anyway. It's one (small) reason why I did a maths degree part time (Open University in the UK) to complement my existing Comp Sci degree. I'm sure you could get a better bang per buck picking and choosing specific areas rather than an entire degree but that wasn't my motivation (I'm equally interested in both in general). A good number theory course (even just plowing through Crandall and Pomerance, or Apostol) will definitely help with understanding/analysing asymmetric encryption.

shogunmike 13y ago
I completely agree with this. While you may not be solving partial differential equations in your average tech startup, there are plenty of instances where a maths degree can be directly applicable. Statistics is one instance, for A/B testing. Another example is the use of vector calculus in machine learning and "data science". How did you find the OU degree?
- alexkus 13y ago
  > How did you find the OU degree? It was good. I took it almost as slowly as you can (typically one module per year) so it was 8 years from start to finish. In that time I got married, moved house twice and became a father so I wanted to avoid it taking over my life. Tutorials were local to home or work (I was lucky in that respect) but they did seem infrequent. Tricky to recommend now the fees have quadrupled though (I paid about £4k in total for my degree, it'd be closer to £15k now); unless it's your first degree and you're considering the OU over a traditional university; then £5k a year is quite cheap as you can be much more flexible with life and (part/full-time) work. I was considering either continuing with the OU with Maths on some of the Master's courses (those fees aren't subsidised in the same way that UG courses are), or switching to languages (French, German, Spanish) but the prices put me off those. For now I'll have a year or two off as a break.
  - bencollier49 13y ago
    I agree, the OU Maths degree is excellent, but it's been ruined by the new pricing arrangements. None of the really interesting people I met on the course would have been able to afford the new fees.
  - Tycho 13y ago
    I've been toying with the idea of doing an OU maths degree. Not for the credentials but because I'd like to be good at maths... I hate not being able to understand it well. However I'm dismayed to hear the cost is now £15k.
- jaymzcampbell 13y ago

bachback 13y ago

tom_b, basically every shop cooks their own solution for just about everything. if you do mid- to low-frequency a Bloomberg will suffice. if you do high-frequency you are in the infrastructure game. you co-locate to the exchanges and battle for the milli- to microseconds. there are a few specialized databases in this area for dealing with TBs or rows. everything else in the area of finance is a complete waste of time IMO. basically the financial system is toast. I hope that bitcoin will improve some of the absurd structures we have set up. anyway, if you want to know some specifically you can go to nuclearphynance. There are some cracks around there. good traders would consider an MFE as the opposite of helpful, I'm sure. standard stochastic calculus is pretty much useless.

makerops 13y ago
There are fully managed solutions to receiving market data etc, that a lot of companies do use, both in the enterprise, and HFT realms (ie tickerplants).
- bachback 13y ago
  Yeah, for 100k$ or more a year. So you need a lot of capital to get started on your won. Getting started with a private account is hard to impossible in my experience. Lime Brokerage offers good stuff for HFT. You need about 12 months to get really going.
ruang 13y ago
What is the difference between strategies used in mid/low frequency vs high-frequency? I understand high-frequency is mostly stat arb but I haven't heard of too many quant funds that do mid/low frequency. I've seen one present before that used small changes in portfolio optimization to get an alpha of ~0.2% over the benchmark but that didn't seem too useful.

nvarsj 13y ago

I worked at an HFT for several years as a developer on the quant lib team. I'd say the algorithm side is the hard and interesting part, while the math is relatively straightforward and solved. For example, it's well known how to price an option - but doing that efficiently and at large scale is the challenging aspect (think kd-trees, etc. for sophisticated caching, messaging protocols, dbs for organizing data). I also built a sophisticated graph algo for relating securities, which built the core of one of our most successful strategies. So technically, I found it very appealing. However, my experience wasn't so great otherwise. You really get the sense that people are only in it for the money - everything revolves around the year end bonus. And at least where I worked, this kind of huge money drove a lot of really nasty politics.