Human emotions can't be modeled, no. But, can a change in attitude be detected before it becomes a full-fledged "exhuberance" or panic? Yes, if the emotions and their impacts build over a period of days or weeks.
My models appear to detect increased danger of a fall in the market. That's what they seem best at. I rarely outperform the market in bullish phases, but the methods help me dodge the dips (such as the 1998 20% "correction" and the recent 13% dip). Historically, the models avoid all the bear markets since the early '70s.
I've worked very hard studying this, for many thousands of hours. It's not the ideal thing for everyone, but I think I can give people warnings when a drop seems more likely based on the historical patterns I've discovered. I try to write something on SiliconInvestor every day. I update the web site predictions for the next day every day, normally by 6:00 Eastern time, and I try to add commentary frequently too.
Yes, the neural network methods are somewhat similar to what I do. A neural network would search for patterns "automatically", whereas I've had to laboriously experiment with whatever formulas I came up with. Fortunately, over 13 years I came up with enough ideas to find a few that work!
But--what works this year may not work next year, so my labors are never complete. Alas!
Kevin Farnham |
