Whats best for serous ATS: lightspeed vs rithmic vs DAS

Discussion in 'Automated Trading' started by rohan2008, Apr 16, 2013.

  1. rohan2008

    rohan2008

    Hi Everyone,

    I am looking at designing a medium frequency automated trading system. After reading these forums, I have a few unanswered questions and so any advice is highly appreciated.

    System Specs:
    • 50-100 trades/day (stocks initially)
    • Custom designed from scratch
    • Modified Linux kernel (cluster maybe) + Xeon cores + C/C++
    • Will eventually get collocated
    • Broker agnostic
    • Dev timeline: 8-12 months
    • Acc starting budget: 80-100K
    • Lowest latency possible (retail only)

    I am stuck with choosing the correct brokerage firm. I am VERY particular about using Linux as my OS and I have taken off IB from my list (of course latency of TWS as well along with order cancel fee). The other factor here is that I am not a pro trader, but a pro s/w architect type and I know there is a tremendous learning curve involved in the art of trading. MB trading is out since they don’t support Linux & C++.... again I am not sure about their FIX gateway product...

    I did consider Lime brokerage, but I feel that its too early to contact them since I don’t even have the system in place. I have currently zeroed on three brokers who offer good API solutions:
    - Vision financial (Rithmic) [ 250us execution with colo]
    - Lightspeed [ 2-3ms execution with colo]
    - Speed trader (DAS API) [unknown… but fast I believe]

    Vision Financial Markets seems to be the only brokerage firm that offers Rithmic for stocks. I did contact them and they are not willing to give out any information unless I come through an introducing broker (of course, they are assisting me in getting in touch with one). So, as such, I don’t know the commission structure.

    Lightspeed has given me good information about their BDK gateway product. I have seen their API and I think I am comfortable with it. They told me that if I can collocate my machine to one of their two location in NY(nasdaq), I can get an execution speed of about 2-3ms. It is definitely impressive, but not as impressive as what Rithmic claims.

    As for Speed trader, the sales reps I have spoken to have no clue about the technology part and so they have asked me to speak with http://www.dastrader.com/ . This website doesn’t even have a phone number? I did come across nice reviews about DAS Trader/APIs, but I don’t know what to expect here.

    So, here are my questions and any info is highly appreciated:
    1. Are there any other brokerage firms that I missed?

    2. Does any of you have experience in designing systems using any of the three: Rithmic, lightspeed & das apis?

    3. Are DAS apis linux compatible? I want to stay away from wine because of the tools I intend to use.

    4. What kind of commissions I should expect when I attempt to trade 50-100 a day? Lightspeed is promising that the commissions will reduce to under $0.0035 if I can make 250K shares a month, I think its doable, but definitely not during the first month.

    5. Datafeed… I am looking at doing the initial design/testing & small runs using nxcore, but I eventually need to move to some good level 1& 2 feeds. Since I intend to look at “stocks only”, I don’t need the options feed. Lightspeed is promising a good feed at their colo and claims that the latency is non-existent. How does this feed compare with rithmic, lime brokerage, Bloomberg and reuters? Lightspeed says that the feed is not one of these multi-thousand types. Can anyone share their experiences here?

    6. (Newbee alert) Whats this whole FIX gateway "industry standard" specs? MB trading seems to be offering... Can I directly program using the FIX protocol.... it doesn't make sense to me when rithmic and lightspeed are providing wrapper APIs. Looks like I am confused here. Do ATS designers program using the FIX specification or use rithmic/lightspeed apis?

    Any advice how I should proceed at this juncture…?

    Thanks,
    P2
     
  2. rohan2008

    rohan2008

    Ok, I have got more confusing information and I need your expert advice. It appears that Rithmic routs its orders to BATS where as lightspeed gives different routing options. Is this something that you consider as a serious limitation? As I understand BATS has about 12% of all the trading volume and so how likely is an order going to get executed in BATS... ? Is there any specific reason why some forum members (as I read) prefer EDGA vs BATS? Is this a hair raiser for that matter?

    The order execution time for Rithmic is around 250us compared to 3-4 ms for Lightspeed. Since their commissions are nearly the same (little higher for vision financials/rithmic) I was wondering if the 250us & BATS only method is going let me be more flexible with my strategies compared to the case with lightspeed with its 3-4ms & all ECNs? What would you choose if you were to choose and why?

    Any suggestions are highly appreciated. thanks.

    FYI: I am interested in stocks ONLY
     
  3. garachen

    garachen

    I think the issue is that it seems like you are approaching this from a software engineering perspective first and a economic/trading perspective maybe not at all.

    Instead of 'how do I beat almost all retail orders' try focusing on 'what's the minimum work I have to do to start making money'.

    Take, for example, your choice of OS/Platform. C++/Linux is fast, yes but you spend a lot of time in development to get very modest speed improvements. When I run c#/Windows I'm at 80-100 micros and I can do C++/Linux at around 12 micros tick to trade. So, yeah, once you get to that point it becomes important but you've got to figure you are spending 30K+ a month in data feeds and infrastructure costs before that matters at all and between 3K and 20K just for a network card that can handle the full exchange feed.

    'Non-existent' latency is total BS. Usually just means they don't want to measure it. Any feed you get that you are not getting direct off the exchange is consolidated. It is HARD to listed to exchange data directly and takes a specialized network card and clever tricks. It's just too much data. I'd be really surprised if you can get <1ms delayed/consolidated data. This alone makes your internal latency almost irrelevant.

    Fix is generally for sending orders. Other protocols for getting data.

    Of who you mentioned, I only have experience with nxcore and Lime. Nxcore is pretty good for what it is - not a bad choice for data. Just remember it's delayed. I kind of remember Lime running at $0.0015.

    Good luck. Sounds like fun
     
  4. rohan2008

    rohan2008

    Garachen,

    Thanks for the response. You are correct in some sense, but I want to approach this in a phased manner. As the first step, I am going to stick with nxcore and design the trading system in house, get it running and move it to a colocation. I shall phase out/replace nxcore data feed with a direct exchange feed. I don’t want a situation where I have to redesign my entire software as soon as I get to this phase and so I would like to go with Linux (besides, I eat, sleep & breath linux).

    So, as for the exchange feed, what kind of mbps are we looking at during day time operations? What type of network cards do you normally recommend for exchange data? Also, whats the cost of the exchange data…. 30K+ is outrageous!

    Any advice is highly appreciated. Thanks,
     
  5. garachen

    garachen

    Each stock exchange (BATS, ARCA, etc) costs around 1500/month for the basic feed. Some of them have FPGA feeds that are more. Futures exchanges are more around 3K/month, bonds are 6K-20K depending on the feed. But then it's the lines between the exchanges that can kill you. You can pay between 5K to 250K per month for a fast Chicago to New York line.

    You don't need to be super fast to make money. Your problem is mostly going to be finding a place that allows you to short the stocks you want. That was the main reason LIME wasn't working for us. Many places will have the same problem.
     
  6. Approximately how many trades (not shares, trades) have you done lifetime? It is an important starting point for anyone giving you advice.
     
  7. ofthomas

    ofthomas

    I have to agree with some of the posters on this thread, you are approaching this problem the wrong way.

    as an example, are you going to be rebuilding the order book? if you are medium frequency, what is the point of being colocated? all that stuff costs quite a bit of money... so not sure what your overall budget is, but if you are complaining about $30K for market data, you still have a lot to learn about the ULLT space that you are trying to enter...

    design and build your platform first, educate yourself on the markets and the required infrastructure, figure out your edge and what kind of funding you require for it, and go from there...

    specially, if you are talking about cash equities..
     
  8. bears21

    bears21

    have you found a firm that has a superior short list?
     
  9. rohan2008

    rohan2008

    ofthomas & Swan Noir

    Yes, as you have suspected, I do not have much trading experience and yes, I am a newbie. Here’s what I am trying to accomplish:

    Understand ULLT/colocation and use this knowledge to design a multi-strategy (strategy agnostic) trading system first. I shall start with 80-90K and run medium frequency trades(50-100/day) with a couple of strategies I have at hand. If PNL looks good after a few months, I will then collocate the system go for higher frequency strategies on the same trading system. All the discussions centered around ULLT/software… etc is to make sure that I take informed decisions and design the system in such a way that I don’t have to redesign it again whenever I get to the colocation stage… in my view, a multi-strategy system that works for medium frequency trades should also be able to process faster frequency provided we change the strategies and the data feeds. I hope I am thinking clearly on this.

    Yes, this is the problem when one enters a new field… its hard to figure out how much we know vs how much we don’t know and so these questions :(
     
  10. I know nothing about tech or system trading but it seems to me that designing a system when you don't have almost any trading experience is a very hard road to travel.


     
    #10     Apr 20, 2013