I like some of the ideas. A couple of questions/thoughts:
Wouldn't this be a bit more appropriate using log returns instead of linear returns?
I'm wondering if the multiperiod indicators (e.g., 13612) go a long way towards accounting for the trend clarity.
I'm also wondering how much the relative benefit is when allocations are also dependent on volatility? Presumably volatility would be a surrogate for trend clarity.
I guess you're talking about doing a linear regression on the log scale prices? Yeah, the same question also came to my mind. I think it can be a valid alternative and might produce better results for some cases, but I didn't dig any further.
I also feel multi-period indicators should include some information on the quality of a trend even though it's not directly measuring it.
Intuitively it feels that volatility should be relevant to the trend clarity. The paper claims, though, that the findings remained robust controlling for volatility effects. I just switched "Top 1 by 250d Trend Clarity" in https://quantmage.app/grimoire/571cf40db77d47c768d04b1c2b4443a8 to "Bottom 1 by 250d Volatility" and it showed a decent performance, too. So I don't know ^^;
WRT volatility determining allocations, that would be like using inverse volatility with several assets. I'd imagine that the assets with higher volatility (hence lower trend clarity) would drop in allocations on top of any single-asset trend clarity effect, even when trend clarity is robust to volatility effects.
I think combining the two can be helpful. I just tried using an inverse volatility weighting instead of equal weighting for https://quantmage.app/grimoire/35dc85a310ed831d574ee4bb1056fc1f and was able to get a slightly better risk-adjusted return when I used 180-d inverse volatility. The paper's point, AFAIU, is that Trend Clarity captures some info that cannot be explained just by volatility.
I like some of the ideas. A couple of questions/thoughts:
Wouldn't this be a bit more appropriate using log returns instead of linear returns?
I'm wondering if the multiperiod indicators (e.g., 13612) go a long way towards accounting for the trend clarity.
I'm also wondering how much the relative benefit is when allocations are also dependent on volatility? Presumably volatility would be a surrogate for trend clarity.
Appreciate thoughtful comments!
I guess you're talking about doing a linear regression on the log scale prices? Yeah, the same question also came to my mind. I think it can be a valid alternative and might produce better results for some cases, but I didn't dig any further.
I also feel multi-period indicators should include some information on the quality of a trend even though it's not directly measuring it.
Intuitively it feels that volatility should be relevant to the trend clarity. The paper claims, though, that the findings remained robust controlling for volatility effects. I just switched "Top 1 by 250d Trend Clarity" in https://quantmage.app/grimoire/571cf40db77d47c768d04b1c2b4443a8 to "Bottom 1 by 250d Volatility" and it showed a decent performance, too. So I don't know ^^;
WRT volatility determining allocations, that would be like using inverse volatility with several assets. I'd imagine that the assets with higher volatility (hence lower trend clarity) would drop in allocations on top of any single-asset trend clarity effect, even when trend clarity is robust to volatility effects.
I think combining the two can be helpful. I just tried using an inverse volatility weighting instead of equal weighting for https://quantmage.app/grimoire/35dc85a310ed831d574ee4bb1056fc1f and was able to get a slightly better risk-adjusted return when I used 180-d inverse volatility. The paper's point, AFAIU, is that Trend Clarity captures some info that cannot be explained just by volatility.