Chuck, my backtesting efforts have been at the macro level. There have been 24 market turns (up or down) in the SPX for data going back to the year 1950 that were followed by moves greater than 20%. The criteria is kind of arbitrary, but I think 20% qualifies as a bear market. So being able to predict when a loss is approaching is more important than missing a gain.
I'm scanning the market for clear signals, so I have not developed a sense for the tool's limitations. For instance, in trying to find the top or bottom of a cycle, smaller cycles with disproportional amplitudes can get detected even though it's past the cut-off frequency of the bandpass filter. This is happening right now. The 80 and 120 day cycles have a combined amplitude of more than 150 SP500 points that shows up as 10 points of noise in the filters for the 3 and 7 year cycles. Similarly, larger cycles get detected in filters for the 80 and 120 cycles. This causes an offset so that techniques for detecting a zero crossing won't work. These imperfections complicate automated backtesting.
Based on the model's long term forecast and unforeseen events of the 2008 crash, I am 100% cash. When banking and finance stocks were under pressure, bans on shorting went into effect. This bear market will be no different in terms of sudden rule changes. It's difficult to formulate a strategy when the rule change. One might be able to justify going short until markets have dropped 20%, trying to get out before an unknown threshold on shorting bans takes place. If those who short have their portfolios blown out of the water by sudden rule changes, then the bottom will be MUCH lower due to less available money. Then I can have a higher percent of ownership of companies in the long run.
About the only thing that would help in my efforts at this time would be a link to 1 minute or 5 minute intraday tick prices. then I could scan for 1 to 10 day cycles.
