I remember reading a few years ago about a company called Chronos Research who were touting systems boasting 20microsecond turnarounds using commodity hardware and their own custom lightweight Linux build. From what I remember the entry point was only about $25k. Obviously there are other larger costs to consider (exchange memberships etc) but it was interesting to see the initial barrier to entry was that low.
What Programming Language For Trading Jobs ?
- Thread starter Fundlord
- Start date
I remember reading a few years ago about a company called Chronos Research who were touting systems boasting 20microsecond turnarounds using commodity hardware and their own custom lightweight Linux build. From what I remember the entry point was only about $25k. Obviously there are other larger costs to consider (exchange memberships etc) but it was interesting to see the initial barrier to entry was that low.
I'm not going to pretend to be an expert, I just happen to know someone who was a developer for a few years in a true HFT firm and have talked about it.
As far as the lightweight linux build, I can think of a number of key things that are almost certainly changed from a standard linux system to reduce latency. I'd be surprised if any actually ran a standard linux kernel. In any case, it's not so much the hardware required that's the problem, it's the monthly expenditures (microwave access, exchange memberships, colocation).
Then there's the ongoing expenses that aren't fixed (clearing for example). To get good clearing agreements, you have to commit to substantial volume and have rather large capital (they don't just care about volume). Also because you're not going through a broker, there is not an API to use to get quote data and enter orders -- a lot of R&D in the work to deal with the messaging formats used by the exchanges (both for quote data and for orders) and these protocols do get updated and each exchange is different, so R&D has to be ongoing to keep up with these.
All of this is completely independent from the algorithms being used to determine when and what orders should be entered, which also need to be continually updated to keep with changing markets. It's really not a "go it alone" sort of thing. I really don't know that even an incredibly well funded individual could keep up with everything that needs to go in to running a true HFT operation -- they'd have to have a handful of people to assist in just keeping everything running as the exchanges themselves change.
As a sort of hobby project, I'm going to see how fast I can calculate the S&P 500 index from scratch (meaning no pre-computed values) from the time the last byte of the last quote comes in to the beginning of the first byte of the order is sent. I'm shooting for 2µs.
I'm not complaining, and I'm certainly not an "old timer" ... I'm 28. I just had my eye's opened to what that particular industry is looking for. I understand why they target the kids fresh out of college... But some of these MM firms are trying to offer there software engineers only about 100-130k in Chicago.... thats not enough, but may seem like a lot of money to a kid who's never seen a paycheck and has been eating ramen for the last 4-8 years .... The only explanation is these kids don't know what they are worth. But, they'll figure it out in a year or two, and move on. I haven't seen the statistics, but I'm guessing the turnover rate is high.
I'm not going to pretend to be an expert, I just happen to know someone who was a developer for a few years in a true HFT firm and have talked about it.
As far as the lightweight linux build, I can think of a number of key things that are almost certainly changed from a standard linux system to reduce latency. I'd be surprised if any actually ran a standard linux kernel. In any case, it's not so much the hardware required that's the problem, it's the monthly expenditures (microwave access, exchange memberships, colocation).
Then there's the ongoing expenses that aren't fixed (clearing for example). To get good clearing agreements, you have to commit to substantial volume and have rather large capital (they don't just care about volume). Also because you're not going through a broker, there is not an API to use to get quote data and enter orders -- a lot of R&D in the work to deal with the messaging formats used by the exchanges (both for quote data and for orders) and these protocols do get updated and each exchange is different, so R&D has to be ongoing to keep up with these.
All of this is completely independent from the algorithms being used to determine when and what orders should be entered, which also need to be continually updated to keep with changing markets. It's really not a "go it alone" sort of thing. I really don't know that even an incredibly well funded individual could keep up with everything that needs to go in to running a true HFT operation -- they'd have to have a handful of people to assist in just keeping everything running as the exchanges themselves change.
As a sort of hobby project, I'm going to see how fast I can calculate the S&P 500 index from scratch (meaning no pre-computed values) from the time the last byte of the last quote comes in to the beginning of the first byte of the order is sent. I'm shooting for 2µs.
I'm no expert either. I just go on what I read and hear. I do know that it's not a space I'd like to be in or be interested in moving to. Far too confined. All the HFT boys are running down an alleyway that keeps getting narrower.
I wasn't thinking of you at all. More some of the old boys I used to work with who really struggled with the way the industry was changing.
Here in London young techies with some industry experience start at about £80k. That's a reasonable wage given the profusion of fintech software talent available in the city.
80k is a good wage even in London, might be able to afford a house even on that kinda money.
Twice+ as expensive as here, joint wage 50k with the mrs, tough going on that.
Trading/quant role is my goal so I guess Ill start with python like others have advised.
Good point about not learning something just to get a job in a few years.
If I can test trading strategies and run automated trading strategies with python it will be a useful skill in my own trading. (if these things are possible with python)
You want to work in quant trading but don't know any language and don't know if "these things are possible with python."? I strongly recommend you consider a different career. You are simply not driven in this area. I mean it, you will simply get hurt badly by not being able to keep pace with everyone else around you. It sounds harsh but is the bare truth. Choose to take it or to ignore it. I hired many junior traders onto my desk and I think I know a thing or two spotting those who make it and those who break.
I have very basic math and programming abilities, but have yet to meet a programmer, or mathematician who can design an instrument pricing model like the one I use to trade independently. I have no doubt that most of them will remain employees the rest of their lives, and there is nothing wrong with that.