Reversion to Mean

My current view on the markets is grounded on years of research, going back more than 50 years and much of it based around development of trading programs created by Artificial Intelligence work. This was in the 1980’s and 1990’s using all the data available, over 100 years in some markets. What we found in that research was quite revealing, especially with what we have seen in the stock market since Bernanke’s QE2 funny money stimulus which started in 2011.

First maybe a little review of our findings using AI in our research:

  1. Commodity markets were the easiest for the AI to write trading rules and to apply successfully, markets like oil, wheat, copper, etc. Markets where supply / demand was key and markets where there were many players that interacted on both sides. These markets worked well with AI.
  2. One caveat on commodity market rule writing can be noted however, markets where a few were in control, like cattle where there are only a few meat packers, were prone to squeeze situations with little valid supply / demand information.
  3. Currency market were the next step up in difficulty. Here supply and demand has less relevance, as one is basically working with government interaction, with some background monetary foundation.
  4. The next step up in difficulty on this scale, is stock markets. Here there really is little supply / demand relevance. All the players want the market to go higher and until the players take it too far, it just keeps going higher. While we could get the AI programs to write rules for short-term settings to write sell signals, on a long-term basis most of the programs would never sell, as they found that if they rode out even huge losses, sometimes 60 to 80 percent, the market would always come back. Obviously where one is in terms of age is a big factor as this could be problem if one retires at the beginning of a downturn and dies before the next rally.
  5. So as I have aged the relevance of Reversion to Mean programs has become of more interest. This type of non-AI Program simply sells a portion of a portfolio as the the market becomes more and more overbought, selling a larger percentage of the portfolio as as the stretching escalates. Obiously a long term perpective is key, this is not a five year program, this is more like a 15 to 20 year program.
  6. What we are seeing in the current market and what we have shown in an adaptation of Reversion to Mean for both Hedging and Speculation was illustrated on this site last week.

Leave a Reply

Your email address will not be published. Required fields are marked *

eight − seven =