Background It was not possible to detect the common problem of insincere grip effort in grip strength evaluation until now. The usually used JAMAR dynamometer has low sensitivity and specificity in distinguishing between maximal and submaximal effort. The manugraphy system may give additional information to the dynamometer measurements used to assess grip force, as it also measures the load distribution of the hand while it grips a cylinder. Until now, the data of load distribution evaluation were analyzed by comparing discrete variables (e.g., load values of a defined area). From another point of view, the results of manugraphy measurements form a pattern. Analyzing patterns is a typical domain of machine learning. Methods We used data from several studies that assessed load distribution with maximal and submaximal effort. They consisted of 2016 total observations, including 324 patterns of submaximal effort. The rest were from grips with maximal effort. After preparation and feature selection, XGBoost machine learning was used for classification of the patterns. Findings After applying machine learning to the given data, we were able to predict submaximal grip effort based on the inherent pattern with a sensitivity of 94% and a specificity of 100%. Interpretation Using techniques from applied predictive modeling, submaximal effort in grip strength testing could be detected with high accuracy through load distribution analysis. Machine learning is a suitable method for recognizing altered grip patterns.