The scientific community has identified four principal parts on approximation algorithm’s logical structure (tuning parameter, generating initial solution, searching and generating neighbor solutions, stopping algorithm execution) and has performed improvements for solving problems such as sorting, forecasting, classification, clustering, constraint satisfaction, decision, optimization. A reviewing of state of art for algorithms Tabu Search (TS) and Threshold Accepting (TA) indicates that in most cases it has performed improvement on logical part for tuning parameter. A reflection about it shows that up to now there has not been an analysis about the other logical parts: initial solution and searching. Study cases for TS and TA over one dimension Bin Packing problem were configured for discovering knowledge of these parts on algorithm performance in terms of quality and time. The results for all configurations indicated that generate a deterministic initial solution and neighbor solutions from this using one method for searching permit to algorithms TA and TS perform better on very short time (reaching solutions with best, same quality and worst with a very minimal difference) than intensify the searching with several methods. It would be of interest if this found knowledge is like other algorithms for other problems, which could contribute to building self-adaptive algorithms that do not waste effort to give the best solution to complex problems.