Super fast trading - all automatic

Discussion in 'Automated Trading' started by trillenium, Jan 17, 2004.

  1. Hi,

    I have read an article in a Traders magazine recently (german) about some small firms who created huge databases of the order book of various futuresmarkets. They analyze every change in the order book and develop trading strategies for so called "super fast trading" . Supposedly they already implemented computers that trade very very frequently every day and act like market makers -- these computers constantly buy and sell depending on the change of the order book.

    Has anybody heard about these small firms and how successfull they are in doing that ?
  2. cosmic



    you could even employ neural network classifiers to scan for nonlinear patterns in these electronic orderbooks...if you can take the costs down, this would be a "master" version for an active scalper...

    btw: I read also the article in Traders & was instantly fascinated! :D

    But I also do not know the firms involved...

  3. hmm. i looked into this, for eurex, but the problem was that nobody has the technology to give you the depth and eletronic execution required. sure, you can see the depth but they've not got it in an api yet so that you can feed it into code. this will be available in about 3-6 months from most futures brokers though.

    i had also thought about this too ;)

  4. Maybe you could get a bit more into
    depth of the article ?
    Which techniques were described etc...

    Was something available somewhere on the WWW ?

    Would be very kind. Thanks.
  5. Banjo


    There is a group in Australia that has been doing this for a couple of years now. There are some guys from Hong Kong involved with them. They are basically a group of quant heads that employ their own ever evolving algorythms applied to index futs on the worlds bourses. They use unix machines and about 50 million usd cash for daytrading. The partners had to put up 8m apiece when they started. The strange part is that they created the system for horse racing, apparently some Asian horse tracks have huge payoffs. Racing is how I became aware of them. The point is that people are doing this around the world, the tech/info are readily available, the rest is pattern hunting within the framework of the larger reality of the moment and betting that micro history repeats itself to the x degree. There are some big players out there
  6. ig0r


    I believe abogdan has created a system (and has been running it very profitably for some time) doing similar analysis and "super fast trading" on stocks, after analyzing MM/SIZE relationships to movements in price.
  7. Banjo i think their application is different. As far as i'm aware you can't actually get historical market depth (bloomberg and reuters offer some for somethings but it's not very accurate). The people in Asia are using pattern matching techniques to determine market direction, probably from 3 bar data. It's actually very effective - I know someone who has developed something similar - it's basically just saying "how many times has this pattern occurred in the past and how many times did it go up and how many down". The guy I know uses daily data and high, low, open, close and stores this as a vector and creates a type of candlestick out of it. He uses it on fx.

    Of course, the technology could be used on this but the problem is that the computing requirement then becomes massive because there are, say, 10 levels of depth for both the bid and offer then you've got to do some quick analysis of recent bid offer depth and make your judgment - and remember people are pulling and adding bids and offers all the time so they you've got to recompute everytime that happens. Thats a lot of computations. I believe a different type of model would be better for that task. Just my 2p.

  8. Please state the title of the article and author as well as the month of the magazine the article appears in.
  9. if anyone has a proven system and needs
    an API
    Software to be written
    please PM me
    #10     Jan 17, 2004