The assessment of multi-person group collaboration has garnered increasing attention in recent years. However, it remains uncertain whether haptic information can be effectively utilized to measure teamwork behavior. This study seeks to evaluate teamwork competency within four-person groups and differentiate the contributions of individual members through a haptic collaborative task. To achieve this, we propose a paradigm in which four crews collaboratively manipulate a simulated boat to row along a target curve in a shared haptic-enabled virtual environment. We define eight features related to boat trajectory and synchronization among the four crews' paddling movements, which serve as indicators of teamwork competency. These features are then integrated into a comprehensive feature, and its correlation with self-reported teamwork competency is analyzed. The results demonstrate a strong positive correlation (r>0.8) between the comprehensive feature and teamwork competency. Additionally, we extract two kinesthetic features that represent the paddling movement preferences of each crew member, enabling us to distinguish their contributions within the group. These two features of the crews with the highest and the lowest contribution in each group were significantly different. This work demonstrates the feasibility of kinesthetic features in evaluating teamwork behavior during multi-person haptic collaboration tasks.