Hi, Ok, I realize this is probably a very sensitive subject, but I thought I would ask anyway. Has anyone written a software to scan a news source (what is a good news source to scan?), and if so, what sort of algorithms are you using? Does anyone autotrade on news scans? For example, MER is getting hit today inspite of a strong market. This could be due to the TDAmeritrade and Etrade merger news. How would the scanner rank the news in relation to MER? nitro
Latent Semantic Analysis. We are using it to filter the news. It is not important what source of the news you use, it is important, however, to have lots of it. We use about 250 different WEB crawlers (all free) to collect the news. http://lsa.colorado.edu/papers/dp1.LSAintro.pdf
I know someone that trades 200 million notional, and is only in the market for 30 seconds of each day. Liquidity providers shun him. There are only 2 or 3 times a week, when key reports are released. If you want to develop something pm me.
This might be of interest. http://newsroom.progress.com/phoenix.zhtml?c=86919&p=NewsArticle&id=1013542
Media stories stated hedge funds were doing automated news trading a few years back. Now Dow Jones offers it to the public. A few months ago I hapened to be looking at a stock when buyback news came out. The stock went nuts in milliseconds. Not to be discouraging, but you could work on a system for a long time only to discover the big boys like Goldman have already squeezed all of the "edge" out of this idea - because it's been out there for a while. Peace, Wilbur
That is a defeatest attitude imo. Every day I see $$$ sitting there to be taken, inspite of other people claiming it cannot be done. Also, there is more than one use of news to trigger, or turn off, a system on a given asset with news. I may not be trying to beat others at news, just not run over by them. nitro
Your program will have to be really super-robust to make much money there. I have back-tested buying into volume break-outs on news, assuming a latency of about 20 minutes, and the results have not been encouraging. Trying to shoot for smaller latency will likely result in not being able to distinguish the noise from genuine stock-moving news.