summary:
In this paper, we establish a regression model based on the passing network to evaluate the influence of team structure strategy and opponents’ counter-strategy on the match results. Fortask1,wefirstlistsomeHuskiesmatchstatisticsforthisseasonandanalyzetheteamin brief. Secondly, we construct a passingnetwork based on the number of passes and visualizes the passing network diagram of three games under three different coaches. We use these three diagrams to describe and compare the changes in Huskies’ strategies. After that, we identify network patterns of dyadic and triadic configurations and count 15 kinds of these two configurations in the above three matches, reflecting the structural indicators of the passing network. We also explore time scale and micro scale by giving the change of the team’s centroid over time in the first match and the Huskies’ 4 positions heat map over the season. For task 2, we construct the regression model not only introducing the basic data representing Huskies’ and opponents’ ability, but also extracting six independent variables from the indicators of the passing network into the model. Considering opponents’ counter-strategies, we also introduce the product interaction term between opponents’ data and network structure indicators. Through the training of regression model, we can judge whether the independent variables introduced have influence, what kind of influence and how much influence the independent variables introduced have on the result of the match. For task 3, by bringing in data for training, the model leaves 10 variables including interactionterms. Inordertoverifytheaccuracyofthemodel,weuseLeaveOneOutcrossvalidation, andthepredictedaccuracyoftheraceresultreached71.05%. Then,basedonthetrainedmodel, we point out the effective structural strategies Huskies currently have, such as the strong connection between the core players. Meanwhile, we also give specific advice for Huskies team to improve team success, such as the emphasis on triadic configurations among players. Fortask4,weextendthemodelappliedtohuskiestoallteamworkscenariosandintroduce the IPOI model. The IPOI model conducts multi-level induction of influencing factors and selection of assessment indicators from the four aspects of team input, process, output and reinput, taking into account team construction, operation, management, feedback and other aspects. WeconsiderthattheexistingHuskiemodelisprogresspartofIPOImodel,andweadd the evaluation system of input, output and reinput part, taking the university scientific research team modeling as an example. Insummary,ourmodelispracticalandreliableforhandlingnetwork-basedteamworkproblems in society.
Keywords: football strategy, network science, regression analysis, IPOI model.