Momentum investing is the idea of buying stocks that are strong and not buying (or shorting) stocks that are weak. Like it's almost opposite cousin, value investing (buying stocks that are undervalued and not buying stocks that are overvalued), momentum investing shows extremely strong long term returns, but also goes through long periods of underperformance. In a new paper "Two Centuries of Price Return Momentum" by Chris Geczy and Mikhail Samonov, the authors look at momentum going back into the 1800's and find significant periods of time when momentum investing is painful.
If momentum investing works so well over the long term, why doesn't everyone do it? The answer is what Cliff Asness refers to as time dilation. An academic can do research on momentum and find strong long term returns and what look like insignificant period of underperformance or bad performance and not be worried at all. In the real world, even a short period where something is not working can create tremendous pressure to sell out or change course. So while momentum is an exploitable market anomaly, in the real world one cannot put all their eggs in the momentum basket.
The key to making momentum "work" is to combine a momentum stock strategy with other types of strategies that have different risk/return characteristics. Some examples might be intermarket analysis, value investing, or even momentum with volatility filters to limit drawdowns.
Momentum investing is a very powerful anomaly that investors should exploit. However, historically momentum has had a number of periods or underperformance or bad performance, to mitigate this it should be combined with other, non correlated, methodologies.