|
Hi all,
I would like to hear your thoughts on how you would approach a simple question: does this event / signal have some kind of predictive power on prices?
I get asked similar questions a lot in my day to day and have a simple approach to answer that:
1. Given a signal, I compute the dates in which it breaches an intuitive threshold (it can be below / above a certain threshold) 2. From that dates, look at the subsequent returns of prices (1 day to 1 month for example), and see how they are distributed (average, sharpe, hit ratio, ...)
One possible improvement would be comparing the return distribution vs the one of the entire period, to see if the two have a different average using Anova, but Anova needs assumptions on the normality of the returns, same volatility, ... and also my observations can be limited, so I don't have strong confidence in the results.
I am conscious that this isn't proper research, it doesn't answer the questions of what threshold to use, how to trade the signal or how to optimize a portfolio around it, but I'm curious to hear your thoughts.
Do you have any ideas to share? |
|
|
|
|
It depends on the distribution of your signal.
Does it normally spike and then go back down? In that case, your approach probably makes sense.
Is it a random value from a fixed distribution? Employ a correlation with price with price lead under several time horizons.
Does it follow a random walk pattern like a stock price typically does? In that case, I would recommend looking at dates on which the CHANGE in the signal (rather than the signal itself) is above a certain threshold and then doing something similar to what you describe.. |
|
|