Quantopian's community platform is shutting down. Please read this post for more information and download your code.
Back to Community
How do you know if the mean-reversion is valid in the current market condition?

I have run this mean-reversion algorithm post in the forum.
Although it performs drastically well during the well-known market crashes e.g. ~2009 and ~late 2018, it does not otherwise perform very well, resulting in loss over the simulated period.

So I wonder, if there is an indicator to tell if the mean-reversion is applicable to the current market condition, and apply this idea only when I am convinced of the presence of such condition.

One thin I heard is the auto-regressive analysis, which I did not have time yet to explore in more depth. If anyone has tried similar idea, I wonder if my idea of catching the market dynamics from such an indicator or two is feasible or if such condition does not last for long enough period, too random, flawed by wrong assumptions, etc...

Thank you.

3 responses

btw the code is brought from this post: https://www.quantopian.com/posts/mean-reversion-short-money-in-bad-markets
not my implementation.

@HyungJun Lim,

There is a school of thought that market dynamics is broken down into three phases: trending, cycling and random. Trending is often times associated with momentum, cycling to mean reversion and random to random. If you can figure out how to accurately measure these phases, you would have found what is closest to financial trading holy grail is. Having said that and intuitively, measure of volatility may be one of the keys to accurately model the phases. Low volatility signals ensuing trends while high volatiltiy signals uncertainty and exploited by mean reversion. Hope this guidance helps you.

Some use the Hurst Exponent to determine if a timeseries is trending or mean-reverting. There are several posts on it in the Q forums. Here's one.