Luca Ferrari, Gianluca Bochicchio, Alberto Bottari, Francesco Lucertini, Silvia Pogliaghi
{"title":"Predicting Future Performance in Powerlifting: A Machine Learning Approach.","authors":"Luca Ferrari, Gianluca Bochicchio, Alberto Bottari, Francesco Lucertini, Silvia Pogliaghi","doi":"10.1186/s40798-025-00903-z","DOIUrl":null,"url":null,"abstract":"<p><strong>Background: </strong>Powerlifting is a discipline in which athletes aim to lift the maximum weight in 3 exercises: Squat, Bench Press, and Deadlift. Since the introduction of \"Classic\" powerlifting by the International Powerlifting Federation (IPF) in 2012, there has been an increase in popularity, athlete participation, and attention from sports science research. Previous studies have examined factors influencing the long-term longitudinal adaptation of upper- and lower-body strength, but no one used this information to develop predictive models of future classic powerlifting performances, especially considering the different age, sex, and weight categories, with the final aim of tailoring the medium- and long-term training goals. This study aims to develop and validate a machine learning-based linear regression model to predict single-lift and overall performance in classic powerlifters. The model considered variables such as sex, age, weight, initial strength levels, and competition history. The study also seeks to provide European normative powerlifting performance data across different categories to assist in talent identification and optimization of training.</p><p><strong>Results: </strong>The final dataset included 54,064 observations from 8,907 unique lifters. Normative data differed between sex, age categories, and initial strength level (p < 0.001). The predictive model demonstrated high predictive accuracy (Root mean Square of Error 10.41 to 19.4; R<sup>2</sup> 0.90 to 0.94), with no differences between mean values (p 0.733 to 0.930), extremely large correlations (r 0.95 to 0.97), and no significant bias (z-score - 1.78 to - 0.64) between predicted and actual performance values across all lifts.</p><p><strong>Conclusions: </strong>The developed machine learning model provides valid and accurate predictions of individual powerlifting performance, by accounting for various individual characteristics. The model can assist coaches and athletes in setting realistic training goals and monitoring progress. Moreover, normative data for each lift and total performance were provided, stratified by sex, age, weight category, and initial strength levels, offering valuable benchmarks for athletes and coaches.</p>","PeriodicalId":21788,"journal":{"name":"Sports Medicine - Open","volume":"11 1","pages":"112"},"PeriodicalIF":5.9000,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Sports Medicine - Open","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.1186/s40798-025-00903-z","RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"SPORT SCIENCES","Score":null,"Total":0}
引用次数: 0
Abstract
Background: Powerlifting is a discipline in which athletes aim to lift the maximum weight in 3 exercises: Squat, Bench Press, and Deadlift. Since the introduction of "Classic" powerlifting by the International Powerlifting Federation (IPF) in 2012, there has been an increase in popularity, athlete participation, and attention from sports science research. Previous studies have examined factors influencing the long-term longitudinal adaptation of upper- and lower-body strength, but no one used this information to develop predictive models of future classic powerlifting performances, especially considering the different age, sex, and weight categories, with the final aim of tailoring the medium- and long-term training goals. This study aims to develop and validate a machine learning-based linear regression model to predict single-lift and overall performance in classic powerlifters. The model considered variables such as sex, age, weight, initial strength levels, and competition history. The study also seeks to provide European normative powerlifting performance data across different categories to assist in talent identification and optimization of training.
Results: The final dataset included 54,064 observations from 8,907 unique lifters. Normative data differed between sex, age categories, and initial strength level (p < 0.001). The predictive model demonstrated high predictive accuracy (Root mean Square of Error 10.41 to 19.4; R2 0.90 to 0.94), with no differences between mean values (p 0.733 to 0.930), extremely large correlations (r 0.95 to 0.97), and no significant bias (z-score - 1.78 to - 0.64) between predicted and actual performance values across all lifts.
Conclusions: The developed machine learning model provides valid and accurate predictions of individual powerlifting performance, by accounting for various individual characteristics. The model can assist coaches and athletes in setting realistic training goals and monitoring progress. Moreover, normative data for each lift and total performance were provided, stratified by sex, age, weight category, and initial strength levels, offering valuable benchmarks for athletes and coaches.