I'm not looking at the wholemarket right now. The first thing one does is train the network, and for this model I am looking at 5 years of data and running walk three forward tests, about 6 months apart. Understand that the net can be used for developing trading strategies independently of predictive analysis. Predicting 1 to 10 days out for the close, change or % change requires that one select the data elements or indicators that are predictive for that stock. Then you forward test to see if it is in fact predictive. Since there are over 800 indicators that I could use to determine what is predicitve just for the net alone, and the same number, plus all the predicive runs that can be used for just TA, what may be predictive for one stock/sector/subindustries may be wholly worthless for another.
The net will be in testing for all this month before anything is ever posted for it. If you would like me to run a forward test on a few stocks (please don't send me several hundred, gang) I will do that, but to give an example, I was testing NOVL and it said to get out at 11.75 today, which I dutifully did...the move up past 12.25 was not predictable - which didn't make for any fun watching that after the exit...
At any rate, if you do want me to look at a stock, let me kow what indicators (Open, low, volume, stochastic, exponetially weighted MA, whatever) seem to coorelate with positive or negative movement and I will load it.
To make this manageable and lottery-like, I'll give Judy one pick and three other picks from the first three posters (one each).
lastshadow |
