Performance Analysis has become an integral part of the coaching process within elite wheelchair basketball, assisting staff with the delivery of augmented feedback (Fliess-Douer et al. 2016). Whilst previous attempts to explore the key determinants of success within wheelchair basketball have identified that stronger teams accumulate a greater number of assists, turnovers, free-throw and field goal shooting attempts (Gómez et al. 2015; Molik et al. 2009), the use of secondary box score data in such studies has been questioned for both its validity and reliability (Ziv et al. 2010). The purpose of this study, therefore, was to explore the key determinants of team success within elite men’s wheelchair basketball using a valid and reliable wheelchair basketball data collection system using primary data collected from match notation (see Francis et al., 2015). Following University ethical approval, footage from 31 men’s games at the 2015 European Wheelchair Basketball Championships was coded using a developed template in SportsCode (V10, SportsTec Inc.) that included 108 action variables grouped into 19 categorical variables: Time, Home Team, Away Team, Offensive Unit (3.0/3.5), Offensive Unit (4.0/4.5), Defensive Unit (3.0/3.5), Defensive Unit (4.0/4.5), Match Status, Start of Possession, Man Out Offence, Shot Taken, Shot Point, Shot Outcome, Shot Location, Shot Clock Remaining, End of Possession, Defensive System, Defensive Outcome and Possession. The template’s reliability had been assessed by Francis et al. (2015) (inter-observer reliability: 0-5% error; intra-observer reliability 0-5% error). The data was subjected to a two-stage statistical analysis procedure in R (R Core Team 2015). Stage 1: Chi-squared tests highlighted 15 categorical variables that were significantly (p<0.05) associated with final game outcome (winning versus losing). The category with the highest level of statistical significance was Match Status (p<0.001). Stage 2: The multicollinearity between explanatory categories were explored. Categories that demonstrated perfect collinearity were removed. Using a 70% sample of the data (4,288 possessions), a forward and backwards stepwise elimination approach was used to build a final model, which included seven categories comprising of 37 action variables: Match Status, Defensive Unit (3.0/3.5), Offensive Unit (3.0/3.5), Offensive Unit (4.0/4.5), Stage, Defensive System and Start of Possession. When tested against the remaining 30% data set an area under the curve value of 0.749 was achieved which suggests the model has ‘fair’ predictive qualities. The final model indicates the importance of maintaining a winning state throughout the game, selecting a unit which predominately comprises of three point players and countering when the defence are pressing. Coaches, players and support staff can utilise the findings from the study to assist with the planning of offensive and defensive game strategies by identifying areas for development within training sessions, supporting selection and line-up combinations and informing the decision-making process of coaches and players during performances.
|Publication status||Published - Mar 2017|