News algorithms?

Discussion in 'Automated Trading' started by dancingphil, Feb 3, 2006.

  1. sccz97

    sccz97

    Of course a certain amount of discretion goes into the input parameters but I thought that was implicit. If the parameters were set up to effect an opposite logic thne the ratio would almost be reversed. The systme is mechanical only in the resepct that the transactions are all handled by application.
     
    #21     Mar 17, 2006
  2. malaxi

    malaxi

    I don't know if this is the kind of thing you're looking for, but for the last few months I've been looking for correlations between words and phrases in the news and stock prices. You can see what I've got so far, here:

    http://www.stockprompts.com/

    Disclaimer: it's got Google Ads on it.

    I'm always tweaking things, but the results of the first 9 versions of the algorithms I'm running are updated daily, along with what they think the stocks/indexes will do each day.

    I have no idea how to make money with it yet (hence the Google Ads :), but it's fun to play around with.
     
    #22     Mar 17, 2006
  3. promagma

    promagma

    malaxi,

    That is pretty cool, nice work!
     
    #23     Mar 17, 2006
  4. malaxi

    malaxi

    Thanks, promagma - now I just need to figure out what to do with the data...

    Believe it or not, I haven't even done the paper trading version of using these predictions yet. I think I'll try to run that this weekend and see if I would have made anything if I had started with, say, $1K for each symbol.

    If you have a symbol you want me to start tracking, let me know!
     
    #24     Mar 17, 2006
  5. sccz97

    sccz97

    interesting. However, you will have more luck making a profitable system if you have make it more reactive to real time news instead of predicting potential moves off online news articles that will most likely be biased one way or another.

    One thing I will say is that you have access to a number of fast news sources, say bb/reuters/mni such that you'll be getting the news as fast as any trader in the world, you can have something as crude as a word lookup to make instant trade decisions. Interpreting certain market moving news like the google cfo comments would be difficult but smthing like the fruad investigation into bae last year or the bid attempt and subsequent rejected for game was very simple. Of course there are numerous things to consider such as:
    1. How do you make sure you know exactly which company you're dealing with
    2. Is there anything you can do in terms of confirmation before the trade goes through (make sure it hasn't already gapped for example)
    3. Do you want to dive straight into the underlying stock or perhaps play it safe with perhaps some options
    4. If you're too slow in getting filled is that particular opportunity over? Perhaps not, depending on the severity of the news and the placing of the company wrt mktcap within its industry sector/group it might have a knock-on effect to similar companies that you can exploit e.g. yahoo/google, gm/ford etc.

    Just a few starting hints as to how you might try and achieve an extremtly profitable, low risk news trading strategy. Please note, that you need some significant technical infrastructure to do this
     
    #25     Mar 18, 2006
  6. malaxi

    malaxi

    Your point about biased news sources is a good one. In theory, though, it shouldn't really matter how biased the news is for what I'm doing. In a nutshell, I'm simply counting things like words and phrases and then looking for correlations between those and stock prices. The content of the news can be biased, as long as it's consistently biased - I'm currently not using the meaning of the content, just measurements of how much there is. If that makes any sense.

    There is definitely a lot to do and to think about beyond my level of expertise. I don't have any plans in mind for an automated system, I just thought that what I've been playing with might be interesting to those who are looking for cues from the news.

    As far as trading strategies go, I've run the numbers on the symbols that stockprompts is tracking right now, and if you were able to just buy plain shares at open and sell at close following the predictions you would definitely not always make back the commissions.

    So whatever the strategy is, it's not as simple as that. Whether it's options or something else, I'm hoping someone will tell me!

    Also, I imagine there are more volatile symbols than the ones I'm following right now (I just added QQQQ on a recommendation). I'm a novice, but I would imagine that more volatility would mean more potential profit for day trades. I'd love to hear tips on other symbols to track.

    (please use the contact form at stockprompts.com for suggestions - I'm more likely to see them there)
     
    #26     Mar 18, 2006
  7. malaxi

    malaxi

    Just wanted to clarify that - I ran the numbers assuming $1000 per stock to start with. Obviously, the more money you put into it, the easier it is to make back the commissions.

    I'll put the plain-jane equity buying version of the paper trades up on the website sometime today if I can.

    -mal
     
    #27     Mar 18, 2006
  8. Hi,

    Are the percentage accuracy numbers based on if the stock increased in price from one close to the next, or from the open to the close?

    It seems like you might just be finding positive news articles and then finding a higher open. For example:

    "NASDAQ futures are higher this morning on bullish earnings reports by CSCO and MSFT."

    Best,
    Morgan
     
    #28     Mar 18, 2006
  9. malaxi

    malaxi

    Hi Morgan,

    I use the daily open and close numbers for calculations.
     
    #29     Mar 18, 2006
  10. Thanks
     
    #30     Mar 18, 2006