Trend following equity curve
- Thread starter phil_galt
- Start date
I'm asking some questions for educational purposes, so do not interpret the following as an attack on you or your ideas.
Could you clarify your 1st point? I don't see how an equity trader can trade "multiple markets" (multiple products) very effectively, unless he specifically looks for uncorrelated items with returns that have no bearing on each other.
And as for market selection, why is his selection incorrect? My only thinking is that he should test the strategy on more data.
When you select a portfolio of stocks to test a specific strategy on, you need to be careful that the strategy is not capturing a phenomina specific to that set of stocks.
Its the difference between looking for a particular behavior specific to a set of stocks versus a behavior that appears infrequently across all stocks.
Interesting. When I develop strategies, I try to get stylized facts for a specific stock or set of stocks, and then move out and search through the space of all stocks to see if I can get some easy, low-hanging fruit by simply reparameterizing my strategy.
I mean, in terms of just looking for a generic strategy, couldn't the guy just test it on SPY first, and then move out his strategy on everything that happens to be correlated?
I guess it depends on what type of strategy you're developing. Your stylized facts approach would work for volatility based models; I know that a volatility breakout model, for example, would likely be effective on products that exhibit the "stylized facts of volatility", i.e. periodic bouts of high vola, followed by periodic bouts of low vola.
To your second point, yes. Several of my strategies start on the SPY and the concept is then applied to high beta stocks. The trick is to not be overly specific with rules, or, simply not use rules that do not have a generalized/fundamental basis to them.