Monday, May 27, 2013

Predicting Future Stock Performance using Historical Data

For the lack of a better word, signal bias; the one thing that everyone is looking for which appears to be a recurring truth, or is it? I decided to put it to the test today with a stock picking game that I decided to make. I've gone into Yahoo finance and picked up historical stock prices for over 5000 companies and decided to work on the premise that if the price of a company is growing now, what is the probability of a company growing in the next time frame in the future?

My motivation for this experiment was to determine how would it be possible to select stocks that given growth in a certain time frame would continue to grow in the future. It is a naive question to ask, but every stock that has gone through a meteoric rise in share price over many years would appear to show growth over an initial period of time and then continue to grow. Those examples that would illustrate this would be companies like Apple, Google and Amazon. What company would be the next to follow in their pattern and how would one be able to recognize it?

Initial testing has been dismal and yet humbling as I have run the most simple test of filtering for stocks that have increased by 25%, looked at the chart and told the computer either a thumbs up or down signal and then calculated my average performance using the next 6 months of historical data. The results say that I'm in the market average based on that strategy. It just goes to say that I am not (yet?) a very good technical trader or the concept of technical trading is not very useful.

I figure that if one is going to be putting money into investments into the long term, that they should really dive into how things work and optimize to see if this is really the right place to be putting my capital or be doing other things. I may release a game based on this principle later on, but for now, this experiment was very educational.

Time to start reading what the quant people are doing and what does the academic field look like in this area.