Development of an artificial intelligence-enabled web application and chatbot for measurement and evaluation of sports performance: Validity, reliability and sensitivity of sportsmetric application

Although laboratory-based gold standard devices are preferred for assessing vertical jump performance, access to these devices has several limitations. The development of technology can enable current assessments to be performed with artificial intelligence, and users can interpret test results using artificial intelligence. This research aimed to develop an artificial intelligence-supported web application (SporMetrik) to assess jumping performance and to train a chatbot model to interpret the results. The research was conducted in three sessions based on the relational screening model, and a total of 101 athletes were included in the study. The validity, reliability, and precision of the SporMetrik web application were evaluated simultaneously in the hands-on waist countermovement jump test with the Smartjump contact mat, the EZEjump contact mat, the Myjump 2 mobile application, and the AI-supported Myjump Lab application. For statistical analyses, intraclass correlation coefficient analysis (ICC), coefficient of variation analysis (CV%), paired samples T-test analysis, independent samples T-test analysis, Pearson correlation coefficient analysis, linear regression coefficient analysis, and Bland-Altman analysis were applied. The results revealed that the SporMetrik web application has excellent reliability (ICC = 0.90 – 0.96, CV% = 2.29 – 3.35), and no statistically significant difference was found between the test means of SporMetrik with the three measurement tools (p = 0.06 – 0.87), except Myjump Lab (p = 0.01). In addition, the SporMetrik web application had similar means within itself in all three sessions (p = 0.05 – 0.67). Validity analyses revealed that there was a perfect positive correlation between SporMetrik and other measurement tools, and the results obtained from SporMetrik predicted the results of other measurement tools at a very high level (r = 0.92 – 0.95, p = 0.01; r2 = 0.84 – 0.89, p = 0.01). Average bias analyses showed that SporMetrik had a bias between 0.35 cm and 5.37 cm with other measurement tools. As a result, the SporMetrik web application can be a valid, reliable, and sensitive alternative for evaluating vertical jump performance. In addition, field experts can assess the results with the chatbot feature in the web application and receive AI-supported performance improvement suggestions.