Predictive models reduce talent development costs in female gymnastics

Area of Science:

• Sports Science
• Biomechanical Engineering
• Talent Identification

Background:

• Optimizing talent identification in female artistic gymnastics is crucial for athlete development and resource allocation.
• Current selection methods rely on coach assessment, which may not be fully objective or predictive.
• High dropout rates in elite gymnastics suggest potential issues with initial talent selection procedures.

Purpose of the Study:

• To compare the effectiveness of linear and non-linear predictive models in identifying talented female gymnasts.
• To assess the potential of these models to optimize selection procedures and reduce talent development costs.
• To investigate the impact of predictive modeling on the dropout rates of elite female gymnasts.

Main Methods:

• A retrospective study analyzing data from 243 female elite gymnasts 5 years post-talent selection.
• Comparison of coach classification accuracy with linear (discriminant analysis) and non-linear (Kohonen feature maps, multilayer perceptron) predictive models.
• Evaluation of model performance based on correct classification rates and potential cost reductions.

Main Results:

• Coaches correctly classified 51.9% of participants.
• Discriminant analysis improved classification accuracy to 71.6%.
• Non-linear models showed higher accuracy: Kohonen feature maps at 73.7% and multilayer perceptron at 79.8%.
• Implementing predictive models could reduce the selected athlete pool by 33.3%, lowering costs.
• The multilayer perceptron model demonstrated the highest predictive accuracy.

Conclusions:

• Advanced predictive models, particularly multilayer perceptron, significantly enhance talent identification accuracy in female gymnastics compared to traditional coach assessments.
• Utilizing these statistical models can optimize the selection process, reduce financial expenditure, and prevent the exclusion of promising athletes.
• The findings suggest that integrating sophisticated predictive analytics into talent identification can lead to more efficient and effective development programs in elite sports.