In the academic literature, I found that momentum returns are negatively skewed (e.g. Daniel and Moskowitz, 2002). As far as I understand, this usually happens when the "past losers" rebound causing heavy losses on the short leg of the strategy.
Nonetheless, I noticed that many CTA/trend-following funds are convex (and have returns similar to the payoff of a straddle: they don't make money when prices don't move much, but they make a lot of money in tail events). This causes the distribution of returns to be positively skewed.
This is confusing. Are momentum returns positively or negatively skewed? Do time-series momentum and cross-sectional momentum have different skewness characteristics?
Thank you in advance