A big hedge fund is known for hiring a lot of traders, giving them each a small amount of capital, and then firing them when they lose, say, 5% of their capital. With this rule, they are basically running a strategy momentum portfolio. I know the fund is very successful, with about 140 traders at any time and getting rid of 25% or more each year. They have several billion dollars under management.
I'm not sure, however, this is all a smokescreen. They could make all their money doing some basic strategies, and the 100+ traders coming and going acts as camouflage. But any firm I have known implements a capital rule along similar lines: if you make money, you get more capital, if you lose, you get less. For most traders just starting out, this basically means you will be at 10x next year, or out. This make perfect sense if you think their profits are a function of their alpha plus the specific strategy return, which is often difficult for outsiders to see. Nevertheless, the result is pretty straightforward. It is conceivable that momentum in strategies is strong, and only a hedge fund, with its unlimited, discretionary, and secretive scope, can avail itself to it.
Actual strategies usually are not nearly as homogeneous as those reflected by standard hedge fund indices (converts, distressed, long/short), but much more heterogeneous. Think of them as factor bets within asset classes. Further the factors are not simply macro factors like the market, or size, but also correlations, dispersion, credit spreads, mean reversion. The asset class/factor combinations are many. Thus, this 'strategy momentum' strategy would be difficult to test academically. But if it does work, it gives large, multistrat hedge funds a big advantage. Their only disadvantage is that there is a large amount of operational risk with such comings and goings, and from a PR perspective, it sounds horrible. Rarely does a multistrat ever admit to actually doing this, they merely say they are doing individual analysis all the time. That many (most) asset managers basically allocate capital using the rule, 'did you make money last year?', is perhaps why hedge funds seem to generate positive alpha in contrast to long-only mutual funds (such funds don't have access to near the same scope of strategies).
I remember reading a paper that took several anomalies when they were published and for each year they invested in the top performing anomaly over the past five years or something like that. I would prefer to do something like increase weights to strategies that have done well, decrease weights to strategies doing poorly, but alas that's all he provided.
Serial Correlation in strategies? Asset classes have a strong tendency to trend on very low frequencies so I wouldn't be surprised if strategies exhibited the same.
Could it be something like a bootstrap aggregation across different "random" strategies?
Provides a whole new level for the concept of "regime change". :)
You dont have to hire/fire bodies, you can take virtual or real funds and apply the same rule: allocate more or less resources according to their success. The program is called Evolution something.
Post a Comment