Hey Lokked

Burn outs always a biatch - good to see I'm not alone!

I'm writing a program which uses a Neural Net to learn the patterns of related stock's movements to be able to predict the movement of a target stock. Don't laugh, its already been done, with better then 50-50 results.
- 'patterns of related stock movements' - isn't that just the correlation of the target stock in relation to a sample of stocks?
- in the 'ideal world' - a stocks price is meant to reflect all known information - therefore, if the sample stocks have moved in price - the target stock also has already moved and therefore reflects its assessed price. As a result, there is nothing to predict - Unless you are with a massive hedge fund who can move the market...
- related stock movements - each 'related' stock will have, for a better word, relatives which must also be assessed to determine which direction they will move before the movement of the target stock can be assessed. Therefore, you could end up with infinite possible stocks movements leading to a huge mess of numbers...
- not to mention you'd have to look at the 'market' to assess all the betas and the impact of market movements...

/arghhh

Goodluck Lokked - wish you well