STM: An Improved Peak Price Tracking-Based Online Portfolio Selection Algorithm

Price peak tracking algorithms are highly effective in capturing market changes rapidly. However, from a first-order optimization perspective, two main issues arise: the current center point is only updated based on the predicted point from the previous period, and the error radius lacks adaptability. To overcome these challenges, we propose the Short-Term Trend Modified (STM) peak price algorithm. STM utilizes a combination of trend reversal and following vectors to determine the center points and dynamically adjusts the error radius through a sliding window. Experimental evaluations on six real-world datasets highlight the significance of our proposed method and its superior performance compared to existing portfolio selection algorithms in terms of cumulative wealth and Calmar ratio. These results offer valuable insights and significantly enhance the performance of single-trend-based price peak tracking algorithms.