个人作品：2020年美国大学生数学建模竞赛（美赛）D题获得F奖。

简介：
In this research, we develop a regression model, grounded in a passing network analysis, to assess the impact of team structure strategies and opposing teams’ counter-strategies on match outcomes. Initially, for Task 1, we compiled a collection of Huskies match statistics for the current season and performed a concise overview of their team dynamics. Subsequently, we constructed a passing network, utilizing pass frequency data, and visually represented the passing network diagrams from three distinct games under the guidance of different coaches. These three diagrams were then employed to describe and comparatively analyze the shifts in the Huskies’ strategic approaches. Following this, we identified recurring patterns within dyadic and triadic network configurations, meticulously counting fifteen different types of these configurations across the aforementioned three matches – reflecting key structural indicators of the passing network. Furthermore, we investigated both temporal and micro-scale characteristics by examining changes in the team’s centroid position over time during the initial match and generating a heat map illustrating the Huskies’ four primary positions throughout the entire season. For Task 2, our regression model was built not only incorporating fundamental data representing both the Huskies’ and their opponents’ capabilities but also extracting six independent variables derived from indicators present within the passing network itself and integrating them into the model. Recognizing that opponents often employ counter-strategies, we also included a product interaction term between opponent data and network structure indicators to account for this influence. Through rigorous training of this regression model, we were able to determine whether each introduced independent variable exerted an influence; specifically, what nature of influence it possessed and to what extent it affected match results. Finally, for Task 3, by incorporating relevant training data into our model framework, we retained ten variables including interaction terms. To validate the models accuracy comprehensively, we utilized Leave-One-Out cross-validation techniques; resulting in a predicted accuracy rate for race results reaching 71.05%. Based on this trained model, we identified effective structural strategies currently utilized by the Huskies—notably demonstrating strong connections among core players. Simultaneously, we provided targeted recommendations for improving team success—emphasizing strategic utilization of triadic player configurations. Moving on to Task 4, we expanded our models application to encompass all teamwork scenarios while introducing the IPOI (Input-Process-Output-Reinput) model. The IPOI model conducts multi-level induction regarding influencing factors and selects assessment indicators from four perspectives: team input, process execution, output generation and reinput feedback – considering elements such as team construction practices alongside operational management strategies as well as feedback mechanisms. We acknowledge that our existing Huskies model represents an initial progression within the broader framework of IPOI; thus adding an evaluation system focusing specifically on input , output ,and reinput aspects – exemplified through modeling a university scientific research team . In conclusion , our developed model demonstrates practicality and reliability when addressing complex problems related to teamwork utilizing network analysis within societal contexts. Keywords: football strategy ,network science ,regression analysis ,IPOI model .