Constructing a time-series momentum strategy involves the volatility-adjusted aggregation of univariate strategies and therefore relies heavily on the e ciency of the volatility estimator and on the quality of the momentum trading signal. Using a dataset with intra-day quotes of 18 assets from May 2017 to May 2019, we investigate these dependencies and their relation to time-series momentum pro tability. Momentum trading signals generated by tting a linear trend on the asset price path maximise the out-of-sample performance in small holding periods while minimising the portfolio turnover, hence dominating the ordinary momentum trading signal in literature, the sign of past returns. Regarding the volatility adjusted aggregation of univariate strategies, the Realized Volatility estimator did not present the best results as it was expected, however the Yang-Zhang range estimator and Garman and Klass Modi ed estimator constitute a good choice for volatility estimation in terms of maximising eficiency (Theorically) and minimising the ex-post portfolio turnover, althought the bias is not minimum.
