Traders' Roundtable

How do we determine that profit-taking is truly superior to letting profits run?

Kind of a crude method follows:

I'd say just compare the number of trades in 2 identical systems (one that takes profits; and one that does not). If we're getting better results taking profits and placing something like 300 more trades over a 20-year sample, then we're onto something. So, we're onto something!

Perhaps you are onto something, but perhaps your clients don't want you to be onto that particular something?! A lot of people view CTAs in general and trendfollowers in particular as portfolio insurance, and the insurance policy will have a much worse pay out if you use targets. In addition, how do you decide what is superior, by just looking at the first two moments or by looking at the higher ones? Is that before or after you have adjusted for the fact that you have added, at least, one extra parameter to your system? And isn't it true that the system would perform even better if we exclude the really shitty trendfollowing currency pair CAD/USD? What I'm trying to say is that "better" is very subjective and that there are a lot of ways of fooling oneself. Targets may be great for some people, but for others they make no sense at all (for me, for example). The only thing we know for sure is that they will cut the right hand tail (trade-by-trade) and increase the level of data-mining. Sorry if I'm just repeating myself

SimJimons wrote:A lot of people view CTAs in general and trendfollowers in particular as portfolio insurance, and the insurance policy will have a much worse pay out if you use targets.

This is an important point I think. Which is - the customer wants to optimize his or her overall risk/reward. Even if the customer is just you. Just optimizing parts of your portfolio is likely to lead to suboptimal outcomes.

For example let's say you optimize allocation between futures, stocks, bonds and cash. Then you optimize each component. Perhaps you also optimize your stocks at a macro level into country and then optimize each country. Maybe within countries you then optimize into market segments and value/growth. Finally you optimize your individual stocks. You can easily end up with four levels of optimization and a lot of money left behind on the table.

This happens in real life. I just finished reading "quantitative equity portfolio management" by Chincarini et al. The books seems to take it as axiomatic that the fund manager will try and optimize performance relative to the benchmark ie the "Sharpe Ratio' would be (return less return of the benchmark)/(standard deviation of performance less benchmark performance). In this world, a constant return of 10% would be seen as *less* desirable than a return of 10% that hugged a highly volatile benchmark.

This is not at all likely to be what the customers want!

Hello,
My 1st post, I hope I post this right.

Good points on the age old problem of profit targets and trend following.



Profit target testing assumes all markets behave the same way. A long held belief for back testers. Generally, I agree. Micro wise, smaller Markets more prone to rogue moves. To tackle a complex task like profit taking, I think the lower hanging fruit is looking at liquidity of each market and writing the code to express it's properties. Ie lumber futures vs. Eurodollar int rate contract