How are momentum and reversion long/short strategies dynamically combined in trading?

I'm trying to understand how to combine two strategies dynamically in trading: one mean-reversion and the other momentum.

One way (also the simplest one) of doing this is by scaling/normalizing values from both strategies and simply adding them. However, this doesn't seem to be a very smart way of doing things.

Is there a way (statistics/technical analysis/DSP/etc.) of separating momentum stocks from mean-reverting stocks and applying these strategies separately on those stocks based on whether they are more likely to be trending than mean-reverting? Or maybe some other way of utilizing both strategies in tandem to achieve a higher Sharpe?

-
Summary from our discussion in chat: You have to discern between legitimate momentum and fleeting movement. That's a whole trading strategy, so merely combining alphas will not cut it. – chrisaycock Apr 20 '12 at 16:54
a very successful stat arb traders once told me: mean-reversion is simply the quant modeler's failure to capture (or model) momentum at shorter time-frames. :-) – uday Mar 26 '14 at 18:06
Agree. Over time and with continued research/efforts, I have realised that the apparent profitability of time-series reversion stems from tracking the first differential of a very cyclic market. However, the kind of reversion I had in mind is not necessarily time series reversion but more of cross-sectional reversion and intra-industry play. – Mindstorm Mar 27 '14 at 23:40

2 Answers

Momentum and mean reversion are labels to describe the behavior of a stock relative to the time period under consideration. That means same stock can be a momentum stock at one point in time and mean reverting stock at different point in time. Similarly at same time, a stock can be both a momentum stock and mean reverting stock depending on which time frame one is looking at.

So what helps is an objective criteria to determine when a stock is in momentum state and when it is in mean reversion state relative to your time frame. Detecting this can be as complicated or simple as you want to be. The precision would be the trade off.

Don't know if I can post a link but otherwise following link has one simple study of detecting momentum and mean reversion phases of a stock dynamically and trading them in tandem for lower draw down and higher return compared to individual strategies. Study: Strategy_Diversification

-

If you have a fairly good model of regime separation (of course requiring a good quantitative measure of regime state classifications -- momentum and reverting) and predictive likelihood (using something like a markov state transition matrix)-- one could weight contributions corresponding to next state probabilities. Of course, you will rarely get a specific answer that is implemented because of the secretive nature of the business.

However, there is ample evidence of funds using proprietary strategies, cloaked in vague terminology ( like trend neutral) that is testament to the idea that similar strategies are deployed in practice-)

-