When selecting a correlation between strategies and a time period, how does it correlate them? If for example I utilize the default 252 window, does it take an average of each 252 period (assuming a 10 year backtest) for its final correlation value?
Rename
To answer the question in topic title, it works as explained in the WL Help > Power Pack > MetaCorrelations. It's just an Excel correlation (i.e. the familiar built-in Corr indicator) on the last bar of the MetaStrategy backtest using the specified lookback period.
Is there a way to add in correlation values on a rolling basis or by lookback period across the entire backtest? For example if I used 252, I would see each years correlation values instead of only the last 252 days.
Hmm I'd consider the visualizer to be feature complete. What you're asking for sounds like quite a bit of work while rolling values don't seem to be in high demand in this regard. Besides, it would be not very intuitive to have such feature without some visualization or a clear explanation. Otherwise it might simply confuse its users.
QUOTE:
correlation values on a rolling basis ...
Usually correlations are about the strategies and not about the data. Either the strategies correlate or they don't. This correlation shouldn't be a function of time (or data).
I would find it highly unusual if two strategies correlated in one moment in time and not another.
You're making an interesting point, plus the visualization of a potential strategy correlation dynamics (that may be a natural next step with original poster's idea) is far beyond this visualizer's design.
Your Response
Post
Edit Post
Login is required