Abstract Resume
Background: In the field of football performance analysis, researchers and coaching staff are always exposed to a “big data” of performance-related match statistics. Hence, it would
be important to identify which ones of them are those that determine the game result (win or loss). The process of this identification can be achieved by modelling relationships between match results
(outcome variables) and performance-related match events and actions (predictor variables), in the means of identifying key performance indicators.
Methods: Data of all the 240 matches of the 2014 season of the Chinese Football Association Super League were collected and analysed. A cumulative logistic regression was run to
modelling relationships of nineteen performance-related match events and actions and one contextual variable (home/away) with the probability of winning. Relationships were evaluated with
magnitude-based inferences and were expressed as effects of a two-standard-deviation increase in the value of each variable on the change in the probability of a team winning a match. Modelling was
performed in four match context of team and opposition end-of-season rank (classified as upper- and lower-ranked teams).
Results: (1) For the upper-ranked teams, (a) when they faced upper-ranked oppositions, an increase of two SDs in the variables of shot on target, ball possession, pass accuracy,
through ball, offside and foul committed could bring a higher winning probability of 13% (± 90% confidence interval: ±15%), 13% (±15%), 16% (±15%), 19% (±16%), 13% (±16%) and 18% (±15%), while the
increase of shot off target, foul drawn and red card would reduce the probability of winning by 21% (±16%), 19% (±16%) and 33% (±25%); (b) when they played against lower-ranked teams, a 2-SD increase
in the values of shot, shot on target, pass, pass accuracy, through ball and tackle could bring 32% (±16%), 30% (±16%), 21% (±16%), 22% (±14%), 13% (±17%) and 17% (±16%) higher probability of winning,
while increase of foul drawn and red card could bring a 21% (±15%) and 31% (±27%) lower probability of winning. (2) For the lower-ranked teams, (a) when facing upper-ranked oppositions, variables that
were positively related to winning include: shot on target (17%; ±15%), corner (16%; ±15%) and foul committed (20%; ±16%), while variables that were negatively related include shot off target (17%;
±16%), pass (21%; ±16%), tackle (17%; ±16%) and red card (43%; ±30%); (b) when playing against lower-ranked teams, variables brought positive effects to winning include: shot on target (18%; ±15%),
foul drawn (14%; ±14%) and offside (18%; ±15%); variables brought negative effects include: shot off target (12%; ±15%), possession (14%; ±15%), cross (17%; ±15%) and foul committed (17%; ±15%).
Conclusions: Results from the modeling could provide information for teams and coaches of different levels to help them carry out effective pre-match training programs, in-match
tactical strategies, and post-matches tactical feedback.
Click here to login