One of my curiosities is accuracy and in particular what's a good accuracy?
For example, if I roll a 6 sided die and I can correctly guess the number 30% of the time, that would be pretty good (if it were just random, we'd expect about 16.7%). Conversely, if I were to call underperform on a bunch of leveraged bear ETFs, it should come as no suproise if my accuracy for those picks is high. At a bear mimimum, a"good accuracy" should be one which is better than chance guessing. That will depend upon what's being selected.
I remember reading a study by Blackstar a while back (see here) that had some interesting findings (two are noteworthy for present discussion):
39% of all stocks had negative returns.
64% of all stocks underperformed the Russell 3000.
The latter statistic indicates that (if the future is like past) that I can obtain a higher accuracy just by down thumbing stocks at random (whether or my score will improve is a different matter entirely). So it seems as though people who focus on outperform picks in CAPS may find it more difficult to get a higher accuracy than those with a good portion of underperform picks. (Score is an entirely different matter).
So I'd like to suggest that what counts as a "good" accuracy will at least depend on whether or not I make outperform or underperform picks.
So I decided to look at my own pics out of curiosity.
Total pics: 422.
Total with positive score: 242.
Total Accuracy: 57% (which is slightly different from what CAPS says... makes me wonder how they calculate it.)
Outperform Picks: 310
Outperform with positive score: 158
Outperform Accuracy: 51%
Underperform Picks: 112
Underperform with positive score: 84
Underperform Accuracy: 75%
If guessing at random implies that I should guess about 36%/64% of Outperform/Underperforms correctly then 51%/75% doesn't seem too bad.
But it does indicate, even aside from leveraged ETFs, that the CAPS rating system seems to favor underperform picks (assuming the Blackstar stats continue into the future).