CAPS5StarIndex and CAPS1StarIndex: The meaning of their anti-correlated relationship?
The CAPS5StarIndex and CAPS1StarIndex have been "live" for the last 7 months, since November 1, 2006. Although this is certainly too brief a time to definitively judge their performance, both indices are doing quite well so far. The CAPS5StarIndex had, as of market close on May 31, a cumulative score of +1491.01 and an average score per pick of +2.63. In the same vein, the CAPS1StarIndex had a cumulative score of +1616.56 and an average score per pick of +4.24. That's fairly substantial outperformance of the market on both counts, and as I discuss in their respective blogs, only a fraction of this performance is accounted for by ended picks, meaning that the active picks continue to rack up outperformance.
For a full discussion of recent performance, check out the respective index blogs:
That makes for an encouraging state of affairs, as this means that so far, the 5-Star stocks with the most player ratings are, on average, outperforming the S&P 500, and the 1-Star stocks with the most player ratings are, on average, underperforming the S&P 500. This is exactly the result one would hope for. Indeed, both indices have obtained All-Star status in the player rankings, meaning that blind strategies of choosing those stocks most preferred by the CAPS community to outperform and those least preferred by the community to underperform winds up doing better than the vast majority of individual players. It's a nice example of the "wisdom of crowds" at work.
But that's just a sidebar. I don't really want to focus on the performance statistics at this early stage of the game. What has been really catching my attention for some time is the way that these two indices move in relationship to one another. I first speculated on this in the CAPS5StarIndex blog back in January.
So that you can see the type of relationship that I’m talking about, here's a chart of the score for the CAPS5StarIndex and CAPS1StarIndex over the last few months:
What should be immediately obvious is that although the general trend is upward for both indices, the fluctuations within that trend are in direct opposition to one another. When the CAPS5StarIndex moves up, the CAPS1StarIndex tends to move down, and vice versa. Indeed, it looks as though if you averaged the two lines, you’d get a straight line with an upward slope. This apparent anti-correlation means that you could actually predict one index’s behavior from the other’s, albeit in an inverse direction. (I’d love to be able to export the day-by-day score stats for a given player. Then I could do time series analyses to pull out the trend for these two indices and find the true correlation absent the trend. But it sure looks as though there would be a strong negative correlation.)
What this means exactly, I’m still grappling with. One early possibility, that I discussed in an earlier post (it was also a subject of discussion in the blog for CAPS5StarIndex, which I linked to earlier), was that the CAPS ratings tend to favor small-cap stocks, so that it would be possible for all stock rating levels to simultaneously outperform (or underperform) the S&P 500 benchmark. Although this looked likely earlier in the life of these two indices, it no longer appears to be the case. Indeed, both indices are currently outperforming the S&P 500 by a substantial margin. Since small cap stocks as a class can't simultaneously out- and under-perform the S&P 500, the star ratings do appear to carry additional information.
A more plausible possibility is that the 5-Star and 1-Star rated stocks tend to be more volatile than the S&P 500 benchmark (that is, they have high betas). Hence, on days when the market moves up, the 5-Star stocks would tend to go up even more, which means they outperform and improve the CAPS5StarIndex’s score. On such days, the 1-Star stocks would also move up more, which means they decrease the CAPS1StarIndex’s score. By the same token, on days when the market moves down, the both the 5-Star and 1-Star stocks would move down even more, which would tend to increase the CAPS1StarIndex score and decrease the CAPS5StarIndex score. This would explain the anti-correlation of the indices, and may even explain some (or all) of the outperformance of the CAPS5StarIndex in an upward-trending market. That is, there is a benefit of volatility if the general direction of the trend is upward, because you’ll have more up days than down. The converse would be true as well, however; in a downward trending market, volatility would cause a more rapid decline. A pure volatility explanation would generally predict that the CAPS1StarIndex will do better (score-wise) than the CAPS5StarIndex in a bear market, while the CAPS5Star should do better in a bull market. However, the CAPS1StarIndex has been doing somewhat better, on average, than the CAPS5StarIndex, and with lower turnover, despite a bullish market. (Interestingly, one would predict that the CAPS1StarIndex will compare even more favorably in a bearish market, which corresponds to my informal observation that it tends to do best on down days.) I think this is an indication that the CAPS system is doing a better job of ferreting out likely underperformers than outperformers in general. Until we see how the indices perform in a sustained downward trending market, however, I wouldn’t be confident in saying that the outperformance of the 5-Star stocks cannot be attributed merely to higher volatility. (It should be simple enough to test this volatility idea by finding the betas on all stocks in each index and calculating the average. If it’s higher than 1, then they are generally more volatile than the S&P 500. I don’t know off-hand of a good screening tool that includes beta, however. If you can point me toward one, please do so.)
Of course, there may be other possibilities as well, and I’d be happy to hear your ideas on the subject.
To my thinking, at least, it is looking more and more likely that the 5-Star and 1-Star stock rankings are beginning to provide some potentially useful information. The trick is going to be figuring out how to combine that with other metrics in order to better identify those stocks most poised for profitable positions. I certainly wouldn’t recommend investing based on these indices, in part because the transaction costs (commissions, fees, and short-term tax consequences) would be large, and in part because their history is still far too brief.