Mean-Reversion Study (Part 3)

The last two posts have detailed a mean-reversion study I did by making use of a price oscillator. Today I want to wrap up with some discussion.

As I pointed out, I would not conclude mean reversion from these data. If it’s mean-reverting in one direction and trend-following in the other then something else is taking place. The simplest answer seems to be volatility.

I would also make a couple observations to satisfy the curve-fitting crowd. I used 23 as the period for Osc, for MFE, and for MAE. To ensure what I saw here was not fluke, I should repeat the study for other periods (e.g. 25, 21, 27, 19, 29, 17, etc.). Similarly, I arbitrarily divided the data into 20 groups. I think it would be useful to divide into 10 groups, five, and maybe four to see if the observations are consistent. If they are then we have something. If not then it’s chaos.

Finally, given that most data were clustered around the extremes of Osc, it is tempting to use only the highest and lowest groups for the analysis. With 20 groups this is 36% of the data. With 10 or four groups this would be 44% or 68% of the data, respectively.