Integrating News Into Automated Trading

Discussion in 'Automated Trading' started by Norm, Sep 27, 2005.

  1. Norm


    I would like to automate my intraday trading. However, I do pay attention to news and take it into account when trading.

    Is there a way to account for news when automating trading/

  2. Choad


    I remember one of the writers over at Real Money describing his (proprietary) automated news/sentment grabber.

    It would go to various places on the web and gather board posts and news articles, and then parse them (it probably had a list of keywords and added up a sentiment "score").

    I think I remember it was a contrary system. When there were too many bullish words, phrases and articles it would indicate time to sell/short. It wouldn't be very hard to program up something like that, but it would probably take quite a while to fine tune it. And then I'm not sure how useful it would be anyway.

    But it's an idea...
  3. Norm



    The news/sentiment grabber is certainly an interesting idea. I suspect what I will wind up doing is setting up the system to accept an externally supplied bias factor. Then, as I become aware of news, I will enter the bias factor that I think is appropriate. This scheme reduces the level of automation, but for what I do, I think the human factor is needed.

    However, I do like the idea and I am open to other ways to handle the integration of news with an automated trading system.

  4. Wow, that's a tough one. Some how you need to translate imprecise language into precise computer logic. If I remember my AI courses correctly(seriously, that's a stretch), you'll be going down the road of AI and machine learning. Unless you build thing really amazing, I think the best you can hope is a "guide", rather than something terribly useful.

    Like Choad eluded to, maybe a simply statistical based system is a good start. I just can't see anything simply being worth anything.

    Sorry, I wish could help more.
  5. maxpi


    The trading software from MyTrack has a column that shows if there is news for an issue or not. They have an API as well, it might be possible to know simply if there was news or not directly in your automation. I don't know if that is available to the API and I'm thinking that will not buy you much even if it is. I am good at looking at the headlines and knowing which ones will move a market. I look at lots of issues with headlines that I know will not move the market and immediately after the news appears I see a speedup in tick count and initial price rise that quickly falls back, so automating that whole process based on the mere existence of a headline and some TA is not going to work well IMO. The news has to be matched up to the company before you can tell how it will impact the issue price and that takes a bit more than some lines of code most of the time. I experimented with keyword/key phrase recognition in stories and made some headway but in the end I am thinking that a knowlegeable operator has to be reading the news and knowing about the company/sector/industry/seasonality, who knows what all else.
  6. Scanning internet forum to measure sentiment is
    a bad idea.
    Someone else could bombard forums automaticaly with bullish or bearish statements to influence the sentiment measure.
  7. I just ran across this today. Kind of fits, but on a much larger/smarter scale.

    Neal Goldman is a math entrepreneur. He works on Wall Street, where numbers rule. But he's focusing his analytic tools on a different realm altogether: the world of words.

    Goldman's startup, Inform Technologies LLC, is a robotic librarian. Every day it combs through thousands of press articles and blog posts in English. It reads them and groups them with related pieces.;_ylu=X3oDMTA5aHJvMDdwBHNlYwN5bmNhdA--
  8. Norm, I have a script that trades really well during high volatility. It picks turning points exactly. Only problem is it that during a trading range, it chops you to death.

    Theoretically, I believe I could trade the entry and exits rules of this system in the hours following the release of major economic numbers - as long as you know when the actual announcement is to be made.

    This may help you with an idea. The only problem is that it wouldn't really work for trading news like terrorist attacks as you can never know when they will occur.

  9. I've looked into doing this, its almost impossible though. The more you look into it the harder it becomes because at first you would just think that you could quantify news as good or bad and look for confirmation in that direction . I mean the phrase "increased guidance" its gotta be good right......stock might go up a dime then tank though (or any number of scenarios).

    Very hard to quantify rapidly intraday and the news released post/pre market is discounted so fast.

    The one area where you would have a chance is with correlation matrix's where for example if company A had good news and moves and your model knows that company B makes a part that goes into company A's product, you could trade the delayed reaction.

    Ex. When logic has news it usually pulls up NTAP and ELX but not always, however when say Apple releases IPOD starts intraday and aapl rips, you might be able to generate a trade off of a component player that supplys them (SYNA for example worked for a while) but even then.......VERY HARD TO DO.
  10. I know a system run by a huge treasury trader that downloads the Bloomberg headlines using the Bloomberg API (client side), so that it would react to economic events a fraction of second faster than human traders. The system just do some very simple text parsing on the economic announcement (since they are basically formulaic), and determine what the direction of the trade. So it avoids any kinds of semantic understanding, AI, linguistics, etc.

    It was an interesting setup, but the system is watched by his clerk all the time, in the event that the system would not interpret the announcement correctly. But the trader was huge enough that a +/- $50k on the event don't make that big of a different to him. And if he is right >= 60-70% of the time, he is happy.
    #10     Jan 17, 2006