The literature contains several researches seeking to analyse and predict the behaviour of 100-meter dash performance through different mathematical models. Although when analysing the historical records in their entirety these simple models fit largely to their behaviour of the historical series, these approximations are not valid when the focus is on the analysis of accumulated times over the past 2 decades. For this reason, this work proposes new alternatives such as polynomial gradient or smooth models capable of explaining with greater accuracy what has happened during the last 20 years for this modality. Therefore, in order to analyse the distribution of competitive marks relative to the top five and bottom five in the ranking over a period of 20 seasons, a total amount of 428 records corresponding to the marks obtained by international level male athletes who conformed the IAAF world ranking in the 100 m race during the 20 indicated seasons were considered for this goal. The main findings of this research conclude with the lack of fitting between the simple approaches (linear or exponential models) and the reported decline in the records -therefore better performance-throughout the analysed period. In return, this work reveals the existence of a tendency towards overall reduction of time records, denoting a positive evolution of the "competitive health" of the discipline. These evolutions, however, seem to be influenced by the position in which athletes qualify, thus showing greater reductions for athletes classified in the bottom five than for those classified in the top five.