Sure, there are grunt programmers out there to build a platform / connectors etc. Lots of those guys. But someone you can pay to build a strategy .. not so much.
Java API for startup hi-speed algo trading
- Thread starter afx111
- Start date
Lots of thsoe, too - the question is what you consider "build a strategy". Come up with the algo? No. Most are PROGRAMMERS that wil lbe paired up with Quants finding the stuff to trade.
This thread is referring to and in the context of getting setup. Not developing strategies. Hence my response to being able to find consultants that specialize in this sort of thing (infrastructure, coding bridges, etc).
If the original poster is indeed referring to finding help in developing a strategy, then it was far from clear.
I agree with the posts above.
Guess I'm trying to say, that without a few "good" strategies, don't even bother to build the infrastructure. The grunt work to build HFT co-lo'd infrastructure today is reasonably straight forward, but still expensive, and *importantly* it does not give you any edge, in and of itself.
Strategy is king. Building the infrastructure is an annoying cost.
Guess I'm trying to say, that without a few "good" strategies, don't even bother to build the infrastructure. The grunt work to build HFT co-lo'd infrastructure today is reasonably straight forward, but still expensive, and *importantly* it does not give you any edge, in and of itself.
Strategy is king. Building the infrastructure is an annoying cost.
I would not be THAT hard. Decent infratructure also depends on WHAT you want to trade and WHAT you consider to be HFT - for real HFT do not bother with Java or most frameworks.
But decent infrastructure for non true HFT (i.e. millisecond range) on the CME is a 180 or so USD per month server. I Know
I get one in 2 weeks. Decent intel core, 16gb RAM, 120gb SSD, 10tb traffic (ok, that was not needed -just the smallest packet they have). FInally moving (virtual) machines to Chicago from Germany (cutting around 125ms latency).
But decent infrastructure for non true HFT (i.e. millisecond range) on the CME is a 180 or so USD per month server. I Know
I get one in 2 weeks. Decent intel core, 16gb RAM, 120gb SSD, 10tb traffic (ok, that was not needed -just the smallest packet they have). FInally moving (virtual) machines to Chicago from Germany (cutting around 125ms latency).I see that as being a problem more than a solution.
Because it's that cheap for the space / box, there's not much competitive advantage in it.
Still some cost if you pay guys to build the connectors, etc. (if the OP does).
So, just saying, all there is is strategy, and that can't really be bought.
Because it's that cheap for the space / box, there's not much competitive advantage in it.
Still some cost if you pay guys to build the connectors, etc. (if the OP does).
So, just saying, all there is is strategy, and that can't really be bought.
Depends. In my case it takes latency from 125 to less than 5ms
One way - that is a LOT of advantage. And it actually allows me to shut down the CLsuter in germany and pull all the machines there into my basement Being exchange local (as in same city) is a HUGH advantage - puts you ahead of everyone not there. On top if ....
YOU are not in the USA to start with that is hugh."puts you ahead of everyone not there" -- Agreed.
But, a lot of people are there.
..
I know an exchange that ran out of colo space a few years ago. After that, no matter what you did, you couldn't get in.
Good times for the few firms that were there from the start.
But, a lot of people are there.
..
I know an exchange that ran out of colo space a few years ago. After that, no matter what you did, you couldn't get in.
Good times for the few firms that were there from the start.
Well Everyone is kind of hart. 80% are likely not Tons of retail people
Point being - real HFT is hard. THen you can not even use FIX sort of - or only with special libraries. Because you need to be extremely fast. THat is expensive like hell.
Everyone is kind of hart. 80% are likely not Tons of retail people Point being - real HFT is hard. THen you can not even use FIX sort of - or only with special libraries. Because you need to be extremely fast. THat is expensive like hell.
Depends on the exchange & the API.
Sometimes it's also easier to avoid FIX, and build the algo right into the exchange connector. (And with less of that silly architecture & platform overhead).
details details ..
Sometimes it's also easier to avoid FIX, and build the algo right into the exchange connector. (And with less of that silly architecture & platform overhead).
details details ..