Stock prediction using Weather, Twitter, News & Stock sites...
http://173.203.58.79/scatter/sc.processtext.php?x=0
While waiting for a few final things on my start-up I mentioned in an earlier posting, I wrote a simple stock prediction program in less than a day...This program was created in 20 hours. It pulls in tweets, weather forecasts, stock market websites, current and breaking news, as well as looks as the trends of the stocks that it finds over time, to create the following collection of stocks. The stocks are rated based on all the factors previously mentioned. I am constantly tweaking each parameter, everyday, and the results seem to be getting better, and better. I will be releasing the source code over the next few days once I clean it up. If you have any suggestions on how I could make the predictions better, feel free to contact me at the email address below.
The entire project is written in PHP using MongoDb and memcache.
Here is the link: http://173.203.58.79/scatter/sc.processtext.php
"BY VISITING OUR WEB SITE, YOU ACCEPT THESE TERMS, AS WELL AS OUR PRIVACY POLICY WITHOUT LIMITATION OR QUALIFICATION."
I mean, how could I read the damn notice without "accepting" it? Also, helpfully put BELOW the useful part, which you cannot "unsee" if you don't want to accept the license...
http://173.203.58.79/scatter/sc.transactions.php
Not a very useful key. Maybe I'm missing something. At first, I thought maybe the colors were the prediction: green means up, red means down, but they just seem to reflect the percentage of "success".
What are the actual predictions? Is there a time frame for them? What do you mean by "success"?
The success is a percentage - the higher it is, the more ideal of a buy it is. The lower, the more chance it should be shorted.
The 16 hour predictions are more "short term advice" where the > 48 hours is more of a long term state.
It does not act on data older than 48 hours as things change to often.
>Please Do Not Submit Ideas, Suggestions, or Other Content to Us Unless Your Intent is to Give Us That Information Unconditionally
One can only assume, the creator has had some problems before with previous sites, or has had extensive legal dealings in offline world.
someone should write a "how to write a trading algorithm" tutorial.
ernie chan's blog is also a decent starting point: epchan.blogspot.com
It'd just be nice to have a clear, concise guide to point to every time something like this comes up.
There's two parts: there is the strategy itself. And then there's everything that surrounds it: what a generic strategy looks like, how to test it, how to operate it, how to build risk models, etc. There's probably more work to the second part than the first.