A Secretary Bird Optimizer with Good Point Set Initialization and Trend-Aware Search for Benchmark and Portfolio Optimization

Accepted population displacements provide directional information that can be reused in subsequent derivative-free search. This paper develops the Good Point Set and Trend-Aware Secretary Bird Optimization Algorithm (GT-SBOA), combining a dimension-admissible number-theoretic initialization with an auxiliary search direction formed from individual and population-center motion. The auxiliary update is applied to the worse-performing half of the population in every complete sweep, with all objective calls charged to a common evaluation budget. Evaluation covers 29 CEC2017 functions, a two-component ablation, directional-weight sensitivity, and five OR-Library portfolio instances. In the comparison at D=30, GT-SBOA has a mean rank of 2.4138 across eleven algorithms, with 14 lowest mean errors and a significant win/tie/loss record of 16/4/9 against SBOA. The four-method comparison places it first at D=50 and third at D=100, showing that the aggregate advantage is dimension-dependent. The portfolio evaluation retains native initialization in the complete-method comparison and uses GD and IGD to assess frontier accuracy and coverage, with a common-initialization control against the parent algorithm. These comparisons delimit the conditions under which displacement-guided updates offer useful search information within a fixed evaluation budget.