A question about designing a medium frequency system, 10k trades per day

Discussion in 'Automated Trading' started by mcgene4xpro, Jun 30, 2012.

  1. mcgene, without arguing what HFT is or is not I will say this. Everyone has different meanings for HFT. If you are trying to place 10k trades per day that implies that you are placing many more orders than filled trades. FX markets are open longer than equity markets so that is one thing you have working for you but it all comes down to what you are going to be trading. Is your symbol universe going to be 100 or 1000 or 50000 symbols? There is a big difference in executing 10,000 trades in 100 symbols versus 50k symbols.

    Just like the argument of whether or not your strategy will be HFT or not - how many symbols you scan will greatly impact code. If you are scanning 100 symbols you can build an easy, lightweight and fast program but if you are scanning 50k symbols you will be closer to medium frequency trading - but your code needs to be much more complex.

    An the link you provided (first post) for the FX exchange stated an average fill time of 3.4ms. That's slow by equity exchange standards where the NYSE matching engines are in the nano or pico-second range. And remember - an execution time is only the time it takes once they receive your order. Add on the latency it takes for your quotes & data to get to you, for the order (or cancel) to get to the broker/exchange and then add on how much time it takes for them to check your risk, which is not included in execution times but can be quite high.

    For a beginner who is not colocating I would look at the C languages vs. Java. Java can get pretty unruly and bloated if a novice programmer doesn't have the skillset to write efficient code. I would also strongly suggest you learn this yourself - or forever be at the mercy of hired programmers. Your paranoia over protecting your IP is valid - but it will be a massive hurdle to overcome. Even if you do find multiple coders to write for you - who is going to put it all together and compile the master source? Each programmer writes slightly differently... how are you going to address cross-coder loops & bugs? Your approach of splitting up the programming either needs to be accomplished by getting what you pay for and (as others have suggested) throwing lots of money at it - or learn it yourself because you will waste more time with finding multiple programmers to protect your IP than you would have if you simply learned it yourself.

    Fx is a bit like the wild west so I do think you have every right to be a bit paranoid over someone stealing a profitable strategy. I think that the only way to avoid something like that is to be fair and equitable and split equally (YES EQUALLY) among a small group. Look at it this way - you couldn't trade without your programmer and your programmer would have no strategy or wouldn't be able to execute without your ideas and capital.

    Either learn how to program (I say C# and C++ with tradelink) or find someone you trust and work with them.

    You have dismissed some very honest feedback in this thread. Yes it may have been harsh but it is true.

    Good luck
     
    #51     Jul 2, 2012
  2. WInston, I hate to blow your thoughts.

    First statement in this thread:

    It plays around 10,000 orders per day per instrument.

    This is 10k order PER INSTRUMENT per day, which - assuming 23 hours - is one order every 8 seconds, roughly.
     
    #52     Jul 2, 2012
  3. your post sounds you have the best intentions to help. But don't you think you put up some false hopes here? I can't even believe people discuss till now how a complete novice can come up with high frequency strategies, a testing platform, and a trading and execution framework. If he could do it starting at his current level of knowledge and be done with it in a year or two I and my quant degree from a top grad school, my many years at sell-side prop teams, my years of coding up my own systematic architecture, my years researching, profiling and developing ever evolving strategies should all be put to shame. I doubt this is gonna happen, though. The OP has stubbornly rejected to even consider mid-range frequencies (for no apparent reason) and naively believes he can push into this business with zero experience wanting to compete with the best of the best in this industry (programming wise, capitalization wise, and market internals knowledge wise).

    You are an optimist and I laud you for it, maybe it motivates him enough to push harder but I am willing to bet my farm that he is not gonna make it as fast as he wishes.

     
    #53     Jul 2, 2012
  4. He's got a ton of mis-info in here. There is no reference (that I found) to the number of instruments and the thread subject says "10k trades per day" which is what I was referencing while I typed that response. I'm not sure that the OP knows the difference between 10k trades and 10k orders.

    Either way I think it's craziness. 10k orders per one instrument isn't that much but if that will hold equal across a symbol universe then it gets into craziness all over again.
     
    #54     Jul 2, 2012
  5. Glad you noticed ;-)


     
    #55     Jul 2, 2012
  6. Great post i have read it twice and i am going to read it for the third time.
     
    #56     Jul 2, 2012
  7. gmst

    gmst

    mcgene4xpro,

    You are persistent for sure and have high ambitions also, which makes me curious as to what do you do for a living?
     
    #57     Jul 2, 2012
  8. With all due respect, i still feel declined from opening a real direct discussion with you.. Not because i am only stubborn! but also because i see our minds are totally on the opposite sides. I am pretty sure such a discussion will not go anywhere. This is not personal at all. I believe you are trying to give me a sincere advice. maybe!. and i accepted it and thanked you for. You might think i could not digest relatively harsh expression. I think it is untrue but the issue is i am seeing not what you are seeing. Simple like that.



    :D However, you actually come up with a very valid argumental subject when you said " how a complete novice can come up with high frequency strategies" and when you have done this conditional decision '' If he could do it starting at his current level of knowledge and be done with it in a year or two I and my quant degree from a top grad school, should all be put to shame."

    I expected after those assumptions , you will come to an end like that " I am willing to bet my farm that he is not gonna make it (period).

    But you came up with " .... as fast as he wishes". Which could be true.

    I think i could ease things in your mind if i manipulate your words a little bit by saying " how an extremely quant/algo novice can come up with a quite bit higher than usual frequency strategies"

    Does this make any change in your way of thinking?
     
    #58     Jul 2, 2012
  9. It is Ok,

    PhD in molecular genetics
     
    #59     Jul 2, 2012
  10. This certainly qualifies you to post on the internet...
    Though I find it very hard to believe.

    Anyone with a post-grad Science degree...
    Would exhibit a much higher degree of logical thinking.
     
    #60     Jul 2, 2012