Thanks for the feedback…

I am aware of the arguments against using the Sharpe Ratio when evaluating CTAs but they don’t hold water with me. You might want to check out this interesting study:

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=943777

I have also studied most all the other measures and they too lack any predictive ability (on their own). One of the stats packages I use computes over 100 performance statistics on each manager! At first that seems great until you realize you’re now wasting your time x100 rather than x1! These statistics suffer from horrible multicollinearity.

I also don’t like the vague discussions about “understanding how a manager trades”. I want to be able to cut every aspect of a manager and his performance down to a number. The benefit of this is that you can then go back in history and create hard and fast mechanical, non subjective, non hindsight biased rules based on how you evaluate CTAs and test those rules to see if they work! I know from experience with my trading systems that most hunches and opinions about what should work don’t! This is why you need a way to test your ideas.

I am aware of far too many advisers and consultants who tell lovely, well polished stories about the depth of their due diligence and experience when evaluating CTAs, but if you look at ALL their historical recommendations often times their suggestions and timing are sub-par (at best).

There is a loophole in the regulation that allows CTA “recommenders” not to be required to produce a disclosure document that shows every single recommendation they have ever put out, and the exact outcomes. As a result, you often tend to see selective memory or “cherry picking” taking place when you talk to them.

If I were working with a consultant I would demand a list of every single recommendation they ever put out (and some kind of proof to validate it). This would include the suggested entry and exit dates. Then I would compute a composite performance table based on this data.

For example, one of the brokerage firms I work with is Attain Capital Management. They are a reasonably sharp group (above average from my perspective), but I don’t like how they present CTAs. The use a flag system that changes all the time. How hard is it in hindsight to put many flags next to the CTAs who recently did well? I think they should be required to keep a copy of EVERY iteration of their flag system so I can look back and see if they recommended good CTAs BEFORE they did good or AFTER. I would also create a composite track record of every CTA they ever recommended, and the precise entry and exit dates. It is my assumption that if they did this that their track record would show a net loss! (I don’t know this for sure, it is just an opinion). Part of my opinion stems from the fact that my own CTA was successful during the time I worked with them*, yet almost every one of their clients lost money trading with me! I have to honestly ask if they are good at timing me, or other CTAs.

These are all things to considered….

Dean

*PAST PERFORMANCE IS NOT NECESSARILY INDICATIVE OF FUTURE RESULTS.

What investors need to do is look around the curve and is gain a better understanding of their investments. I believe the best time to enter a trend-following CTA is during a drawdown. The drawdown suggests that commodity prices have been fairly stable and the CTA recently has not had many opportunities to make a profit. However, that will inevitably change as we live on an unstable planet filled with unstable people. Supply and demand will not stay in balance forever and as that slips, prices will shift creating a more profitable environment for the CTA.

What I like to do is look at the CTA and try to determine if this manager has a profitable strategy. If I believe the answer to be a yes, then I will look at the investment environment and enter when I believe good periods of returns are coming. Once invested, I usually like to stay with them knowing overtime their edge will pay out.

When selecting a CTA, past performance could be used as a guidance of what to expect in a given market environment (i.e. how this CTA behaved during a bull market, during a bear market, during a low/high volatility market, etc.). Past performance is also useful to ask a fund manager “what happened back then” (for either strong returns or poor returns).

However, this “expected behavior” based on the past behavior of a CTA has to be put into context: what was the size of the CTA by then from now (if the CTA had much less AUM by then, it might not be as reactive today), were the markets traded by then the same today, did the CTA change its models, what about their margin to equity then and now, etc.

In order to invest in CTAs, just as in any other investment vehicle, one should understand precisely how the manager trades, how does he control risk, etc. and of course look at the operations. Understanding how models work (even at the high level – no need of the secret sauce) is mandatory in order to understand what to expect. But this is far more work (and required expertise & experience) than just sorting out a database and throwing a mean-variance optimization at it…