A Multiple Rolling Turning Point Detection Method

Detecting time series turning points is crucial in the financial field where series are characterized by several changes in their trajectories. This paper proposes a multiple rolling test of hypothesis of a regression model slope change where the entry and exit windows contain more than one observation, thus contributing to a significant reduction of false signals and the corresponding probability of wrong decisions. To give evidence of the procedure's performance in predicting turning points, we consider – as a preliminary analysis – a set of twenty stocks selected from the EURO STOXX 50 Index, covering the historical period 2010 – 2021. The model is run with different values of the main parameters, providing additional information in investment decision making.