What is Digg?Sponsored by Sony Pictures
Watch a scene from 2012, in theaters November 13 view!
whowillsurvive2012.com - Get ready for the biggest event in history - the end of time. How will you survive? 2012- opening 11/13
23 Comments
- CaptUnderpants, on 10/12/2007, -0/+5Gotta Digg you up for using the word 'automatical'
- JRumph, on 10/12/2007, -0/+3This is already being done. I've met with a few managers that are implementing a system very similar to this. They pretty much let the computer make its own decisions, but they are still watching it pretty closely.
- halik, on 10/12/2007, -0/+3 One of the guys that I interned with at a very prominent investment bank did some reasearch into this topic while finishing his MBA:
Apparently there is a delay of a couple of minutes between when the news come out and when the stock price is affected - he wrote an app that would scour thru rss feeds, looking for abviously positive or negative words and ticker symbols and then he'd track the movement of the stock price. From what he was saying, it seemed pretty reliable. - smith, on 10/12/2007, -0/+3SkyNet should consider a stock split.
- kjigz, on 10/12/2007, -0/+2http://en.wikipedia.org/wiki/Technical_analysis#Neural_networks
- stauffere, on 10/12/2007, -0/+2@halik,
Neural networks are a method of distributed parallel processing and would actually be well suited to this task. It isn't science fiction if it actually exists :) - pukeAndCry, on 10/12/2007, -0/+2I'm currently doing research in this area. I am using Neural networks and RSS feeds to process news. I'm also analyzing regular quote data (price, volume, etc...) as inputs to the network. Initially I was using information in Harvard's General Inquirer project (google it) to look for words in news articles categorized as positive / negative and attempted to score the news article like that. This, of course, didn't work well as I had no sense of context. I could map words but I didn't know what the article meant. This was because a word like "moderate" may be categorized as positive but the phrase "the earnings decline was moderate" is certainly negative. I've started looking into natural language processing to fix this but it looks like Reuters may have done some of the work for me. I was hoping to have more success with natural language processing than most because my domain is limited to financial news, I'm not trying to process the English language.
- datastorageguy, on 10/12/2007, -2/+4There will come a day when humans don't even participate in market trades, but rather rely solely upon computers and computer models to do the trading for them.
Headline 2054 : In today's news SkyNet shares have risen to an extraordinary share price of $1,239,345,312,456,232.00. - yzerizef, on 10/12/2007, -0/+2I'm assuming that they'd be using neural networks to develop a dynamic algorithm to mimic investor reactions to the news. I think it actually might make trading more interesting. If a hedge fund, or individual investor for that matter, can "beat" the algorithm, they have the opportunity to bring in long-term profits until it can correct itself. Anyone have any input on how this affects EMH?
- pukeAndCry, on 10/12/2007, -0/+2Does anyone have more information about what Reuters is doing with the tags?
- sissipher, on 10/12/2007, -0/+1I have an extra few computers lying around and I would love to do this for a class project. If anyone can at least point me in the right direction to get started, that would be fantastic. Thanks in advance.
- yzerizef, on 10/12/2007, -0/+1Jrumph -
I agree that it might support a semi-strong form, but how accurate do you think it could be? I think it might actually make trading more volatile and create more opportunities to generate alpha. Investor reactions are fairly difficult to project, but if the market is controlled strictly on news and there's an algorithm behind the movements, it might actually make it easier to make profits based on public news.
Halik -
How are neural networks sci-fi? I'm not disputing the claim, just wondering if there's something I don't know about. - ediPG, on 10/12/2007, -2/+3As far as I know http://www.vast.com index millions of classifieds per day and that process is automatical for any format, site or markup. Tagging is very primitive way of classification there are much more sophisticated techniques that allow parametrized search and tracking.
- jstandard, on 10/12/2007, -1/+2DEWEY DEFEATS TRUMAN
- JRumph, on 10/12/2007, -0/+1I agree with you. It all depends on how accurate this thing is. On one extreme it can be 100% accurate and the entire market will adjust perfectly the second information comes out. That will make the market perfectly efficient and alpha will be impossible. On the other extreme, the model will be unreliable, leading to volatility and imperfect price swings. That will lead to the potential to generate alpha assuming someone can "game" the system and figure out how to profit from the algorithm (possibly finding some arbitrage hidden in there for someone who can act fast).
The truth is that it will be somewhere in between. I think it certainly has the potential to make the market more efficient and alpha more difficult to obtain, but I don't foresee any possible way to eliminate alpha altogether. If nothing else, there is potential alpha in what manager has the "better" algorithm and can act more accurately before the other guy.
As systems like these get more mainstream I look forward to meeting with more managers who implement them just to get their take on it all. - JRumph, on 10/12/2007, -0/+1@PecanHead
Haha, I wish I was a nerd. I once had a very attractive mid 30's co-worker tell me that if she could change one thing in her life it would be that she was nicer to the nerds in high school. She said all the "cool" kids (like her husband) ended up doing nothing with their lives like their 15 minutes were used up in high school, whereas all the nerds seemed like they were really successful and were doing great things. From that point on I told her she could call me a nerd anytime. - JRumph, on 10/12/2007, -1/+1Depending on how technical you want me to get, it will support the EMH, especially the weak-form (which rejects technical analysis and is already supported through historical data). Semi-strong form EMH, which deals with fundamental analysis relating to public information will also be supported. This is important because semi-strong EMH is the only form of EMH that isn't completely supported by historical data. Strong-form EMH, which relates to private information a la insider information won't really be affected by this.
All in all, if this works, it will create a much more efficient market and will make alpha even more difficult to obtain. Depending on the results, this is just one more argument for buying nothing but passively managed index funds and taking the market return as opposed to seeking out alpha via active management.
I can get more in-depth but most people just get bored with it (ie my girlfriend). - Dysus, on 10/12/2007, -0/+0This is an example of a company that provides the equipment and software:
http://www.imagination-engines.com/default.htm - halik, on 10/12/2007, -1/+1Well what I've gathered by the o/p use of "neural networks" he meants adapting systems, which hardly will be the case when dealing with real money. I could be wrong, but any autonomous adaption requires certain amount of trial and error, which is not something you'd want your algorithm doing with someone elses money.
By saying 'scifi' I didn't mean that the technology isn't used; I'm saying that emulating what an equity trading desk thinks is scifi. Wired had a nice writeup how adapting algorithms are nowdays used for all kinds of random stuff, including CGI animating of walking. - yzerizef, on 10/12/2007, -1/+0It's pretty close to that now. Although I highly doubt anyone would ever pay nearly that much for one share of a company. :-)
- tehpoutine, on 10/12/2007, -2/+0Wow, you mean to tell me computers are taking INPUT and making OUTPUT? what will these crazed mad scientists think of next.
- halik, on 10/12/2007, -4/+1Neural networks? This isn't scifi channel...
Vast majority of autonomous trading right now is for processing orders (timing) and they use price history, statistics and technical analysis. I would imagine they'd use the news feeds as another input for their models, rather than data you'd exclusively use to make decisions. - PecanHead, on 10/12/2007, -3/+0Shut up, you ***** nerds.
jk
© Digg Inc. 2009
— Content created and posted by Digg users is