Program Trading, what’s it all about?

Discussion in 'Strategy Building' started by RusselHarvey, Oct 28, 2006.

  1. September 24, 2006
    Posted by Dan Gallagher

    Up until recently I had no idea what Program Trading was all about, other than the fact that anyone who has worked in the area tends to be highly sought after due to the nature of the business.

    Program trading can mainly be found in sell side institutions and the main players are Credit Suisse, Deutsche Bank, Goldman Sachs, JP Morgan, Lehman Brothers, Merrill Lynch, Morgan Stanley and Citigroup. It offers products such as DMA (Direct Market Access), Algorithmic trading and Prop trading.

    The technology behind program trading is used to allow financial institutions to computerise or automate trades that were previously done by hand. With the advent of algorithmic trading banks can now apply an algorithim to the program to break up the trade in order to maximise profits.

    All this is made possible by having a very stong, reliable technology infrastructure. Banks allocate millions of pounds to creating cutting edge platforms for program trading. The majority tend to build their own in house applications and each system is made up of four key components; trading software, order routing, data feeds and risk/position management.

    The trading software is the user interface for the application and is known as the GUI. This is generally developed using front end technologies and it appears that C# or Java is the language of choice.

    Order routing applications allow banks to route trades to exchanges and ECNs. As program trading is reliant on speed of execution it is critical for these systems to be well built. It is in the space that we see FIX (Financial Informaiton Exchange) and messaging technologies such as Tibco RV.

    DMA is a method of looking at liquity across different markets and time zones and there are many products which can be used that allow this to happen. Companies such as NYFIX and GL Trade provide custom built solutions for financial institutions.

    Another key component for program trading is the handling of data. Data feeds handle real-time and historical data which are used to monitor and model market movements. This is what allows institutions to build algorithims which can then be tweaked in order to create market winning strategies.

    Risk and portfolio management is an essential component of any program trading set up. It allows traders to watch how the profit and loss potential from any trade as the market moves. Pricing engines or analytic engines are built using C++ that aggregate risk by trade or book. Complex data is stored in specialized databases such as KDb which allows fast access to information.

    From a recruting perspective what makes program trading so interesting is the need to find candidates who have very specialized experience in one of the above areas. My experience to date tells me that it is very hard to find people who have an indepth knowledge of all facets of program trading. However, for those who do work in the area it is one area within finance technology that needs continuous investment in order to keep up with your competitors.

    If you have experience in program trading and would like to explore new opportunities please feel free to contact me.
     
  2. what's this post all about?

    Are you looking to hire someone?
     
  3. thanks alot Russel.. very informative..
     
  4. This is like the snake oil of financial engineering.
     

  5. that site is interesting, and whether they provide a real solution that actually works is a totally different story, but what does that have to do with programme trading? NN may certianly have an application in all areas of life, but I've read that page and it seems they just offer some form of NN with continuel changing structure.

    Just to add to the original post, programme trading isn't exactly about predicting or forecasting etc as systematic, algorithim or systems trading may be. Programme trading is about managing a trade efficiently, executing large trades by breaking them down into smaller parts so that the market can absorb them with little market price impact. In a way, based on past data of transaction and the integral nature of market agents.

    On another subject, NN certainly has application guiding stock market forecasting - but only for the expert who knows how to use them.
     
  6. That's an actually worked system. Pertaining to continual changing structure - it's not actually true. The system changes it NN structure only during training session. That means user can retrain or enter new information that can be used as additional data that can be reflected in NN structure.



    In most cases you are right. However, in this case NN system Artificial Foreteller was developed to repeat trader activities. It uses fixed quantity of indicators to find extremum in prices. As soon as it finds this extremum it generates trade signal. The same set of indicators also is used during train.
    The problem is - when you change set of indicators you should retrain NN.
     
  7. I am not sure I know exactly what you meant by saying "program trading isn't about predicting or forecasting ..", based on what I heard it's definitely what the program trading is all about.

    In anther news clip I posted here,
    http://www.elitetrader.com/vb/showthread.php?s=&threadid=79698
    In last ten years, especially last 2 -3 years, the computing based, strategy based, quantitative trading style has definitely gained momentum and earned more respect from the traditional traders. Other than the fact the institutional fund model is much more scalable than individual trading strategy, there is any fundamental difference.
     
  8. Program Trading Definition...

    Computerized Selling or Buying a Basket of 15 or more Stocks with a value of $1,000,000.00... at one shot... bang... and not really caring if there is market impact or not... the buyer or seller is in or out immediately...



    Algorithm Trading Definition...

    (Algo Trading) Computerized selling or buying of large block of <b>A</b> as in ONE stock and slicing it up into 100 to 300 lots to sell or buy. The trades go off or not depending on a math based computer formula (algorithm)... that watches the market from nano second to nano second and they are random trades (in theory) and the execution hopes to create as little market impact as possible...

    There are also algo portfolios trading now but again the sells & buys are 100 to 300 lots...

    anyway that was the going def for both... maybe it no longer holds ?

    since there NOW seems to be a blurring of the use of the word Program trade and Algo trade...

    cj...

    :confused:


    HAVE STOP <img src=http://www.enflow.com/p.gif> WILL TRADE