Abstract
Momentum is one of the largest and most pervasive market anomalies. However, despite a high mean and Sharpe ratio, momentum suffers from large negative skewness that comes from momentum crash periods. These crashes occur in times of both market stress and market rebound and thus variables that capture these episodes, can be used as momentum predictors. Once momentum prediction has been proved, the predictors can be applied to momentum risk management. I introduce two new momentum predictors and show their predictability in single and multiple regression models in the presence of other predictors that have been used before. I then introduce a new method of momentum risk management that has a lower transaction cost than existing methods, both in terms of turnover and price impact.