Manager finds factor-replication “blunt” and distributional-replication “suspiciously opaque” – advocates mechanical trading instead

Surveys of the hedge fund replication field often divide the various offerings into three distinct classes: factor models, distribution replication and mechanical trading.  We hear a lot about factor models and now about distributional replication, but “mechanical trading” seems to be the neglected sibling in the cloning family.

Now one advocate of the mechanical trading approach, a CTA manager called Conquest Capital Group, has made their case in a paper available here at Eurekahedge.  While a portion of the paper describes the benefits of their particular passive CTA replication fund (a managed futures beta replication strategy with “a fee schedule that is more appropriate for a beta product”), it does make a number of general observations about the field of hedge fund replication.

The authors base their arguments on several papers we have also discussed at AllAboutAlpha.com, namely: the March 2007 Edhec study on replication, Northwater’s recent survey of replication methodologies, Bridgewater’s 2005 observations about hedge fund correlations and Harry Kat’s original papers on distributional replication.

Other examples of the so-called mechanical-trading approach to hedge fund replication (cited by Kat in this paper) include: the Merrill Lynch Equity Volatility Arbitrage Index, the Merrill Lynch FX Clone, and Deutsche Bank’s Currency Return Index.

On Factor Modeling

Conquest is obviously not a fan of factor modeling and decides to pick on Goldman Sachs as it raises concerns about this particular approach:

“One of our first reactions to Goldman’s methodology was simply that it is too slow and static, leaving it especially vulnerable when volatility rises…When volatility rises, however, traders change positions more rapidly and risk factors inferred from past periods may quickly be rendered useless. This is especially true when risk factors are recalculated on a monthly basis. Intuitively, it is hard to see how such a methodology could capture the dynamic high-frequency trading that is common in the industry.”

“The bottom line is that factor-based modeling is at best a blunt tool for replicating specific hedge fund strategy types…”

On Distributional Replication

After unloading on factor modeling, Conquest turns its sights on distributional replication:

“FundCreator can actually be described as a systematic trading strategy, and, in this sense, does not even deserve its own category. It just happens to optimize for the statistical properties of returns other than the mean. From whence does the mean return come, then? Professor Kat somewhat cryptically answers, ‘in an efficient market, in the longer run investors will receive a return in line with that risk they have taken.'”

While Conquest finds Kat’s explanation “cryptic”, it does find Northwater’s subsequent analysis helps clear things up a little:

“A May 2007 study by Northwater Capital Management sheds considerably more light on the question and concludes that, Parameters such as volatility and correlation to a target portfolio are robust with respect to the selection of the reserve portfolio; however the mean return is dependent upon the contents of reserve portfolio. A reserve portfolio with a high Sharpe ratio will tend to produce a replica with a high Sharpe ratio, and a reserve portfolio with a low Sharpe ratio will tend to produce a replica with a low Sharpe ratio.5 In other words, the mean return comes from the underlying reserve asset, while the other explicitly targeted moments of the distribution are a result of Professor Kat’s dynamic hedging process.”

Conquest’s feels that Kat’s lack of disclosure (they call his approach “suspiciously opaque”) means that his approach should not even technically be referred to as replication:

“Since investors presumably care about their mean returns, the selection criteria for the reserve asset are rather crucial, but not public. In this respect, FundCreator does not satisfy one of the main desiderata of replication: that of transparency. Professor Kat does claim transparency in the sense that all we do is trade futures and anyone who wants to see what the portfolio looks like can do so at any point in time,(K&P 2005) but what people normally mean by transparency is the knowing the rules of the strategy.”

Aside from having issues with the methodology’s level of transparency, Conquest also finds the variance in the requisite position sizes to be problematic.  The firm points to an example on the Fund Creator website which includes position sizes that it says would be either too small or too big for any one investor.  Furthermore, Conquest says it is,“skeptical how one might consistently…select a reserve portfolio that is capable of such a high mean return but is capable of generating a negative correlation to a 50/50 stock/bond portfolio, given that the reserve portfolio, which is composed only of futures, cannot be shorted.”

Still, they throw the Kat a bone before concluding that Fund Creator is “not commercially viable” (although we can hear Kat laughing right now since, although he surely disagrees, he has often stated that financial gain is low on his Fund Creator priority list):

“Professor Kat is a very talented statistician, but even if we cast aside all suspicion of overfitting and ignore the transparency issue, the FundCreator methodology strikes us as not commercially viable. Even if the statistical properties do converge to their targets over the longer term, that the sequence of returns is unimportant increases the likelihood that the replication strategy will fail to cushion the returns under adverse conditions.”

“The Conquest Approach”

The rest of the paper focuses on Conquest’s own mechanical-trading replication offering. The firm compares its CTA replication fund with the S&P 500 Managed Futures Index and finds that over the 2005-2007 period, it outperformed by an average of 1% (annualized).  The paper points out that this is roughly equal to the difference between Conquest’s fees (1 and 0) and the typical active CTA (2 and 20) – suggesting a direct causal relationship between lower cost and higher return.  (note: they discount the importance of the 20% performance fee for active managers since the CTA index has been underwater since the beginning of the period of study).

In conclusion, the firm goes to great lengths to point out that this sort of mechanical-trading approach isn’t just for CTAs – pointing to Bridgewater’s 2005 conclusion that a variety of hedge fund indices could be replicated with simple trading rules.

So while the authors dedicate a significant portion of this paper to an analysis of their own proprietary fund, they do raise some useful points about replication in general – and we are pleased to note that Conquest Capital Group also seems to be a regular reader of AllAboutAlpha.com.

Be Sociable, Share!

One Comment

  1. JustTheFacts
    July 20, 2007 at 8:30 am

    Mechanical replication usually refers to creating a systematic version of a discretionary trading process. Trendfollower CTAs ARE systematic. Thus, what Conquest offers is not mechanical replication, it’s actually a trendfollowing system with low fees. It’s not beta at all.


Leave A Reply

← Merrill: Hedge Fund Replication the "Solution" for Hedge Fund Newbies Modified Kat-Palaro HF replication model heralds entry of major new competitor →