Hi Larry,
I don't think you're doing anything wrong. Don's study used an index. Some ETFs (and stocks) do better with different AI options than others. It really depends on volatility and price movement over the period tested.
My suggestion is to use 5 year time periods going back 20 years and test with the various options over each 5 year period.
For example, start with, say, 1990 and test from Jan. 2, 1990 to Dec. 31, 1994. Repeat from Jan. 2, 1995 to Dec. 31, 1999. Continue until you reach the current date.
Then see which options resulted in the best returns. Sometimes you might not see any definitive patterns, in which case you won't learn anything. In other cases you'll see one option gives much better results. Of course past history doesn't guarantee future success, but if one option has consistently outperformed another in the past, chances are it will also do so in the future.
I hope that helps. Let me know if you have any other questions.
Thanks,
Mark.
