Basketball has seen an increase in the number of players who perform multiple tactical roles. Therefore, the aim of the study was twofold: (i) to define a method to characterize basketball players as versatile or specialists, based on 13 game-related statistics; (ii) to evaluate versatile-specialist tendencies in a professional national league. A predictive model was proposed using the Automated Machine Learning (AutoML) of the H2O framework. The model was tested using data from nine seasons (2008–2017) from the Brazilian national league (NBL), encompassing 1497 players' observations, achieving an accuracy of 70.81%. We classified players as versatile or specialist and observed the following: (i) the number of versatile players has grown over the nine seasons period (from 25.16% to 47.85%), with Small Forward and Power Forward players presenting the fastest growth in versatility; (ii) NBL teams had similar proportions of versatile and specialist players; (iii) for the best players in the NBL (All-Star game players), there was a trend toward a higher number of versatile players (58.33%) compared to specialist ones. In conclusion, the method was effective in indicating the players' degree of versatility and demonstrated a tendency of increasing versatility over the analyzed seasons. In practice, it may support the assessment of player's profile and contribute for coaches' strategic decisions.

The aims of the present study were to systematize a set of concatenated space creation dynamics in basketball and to use it to analyze teams’ offenses. Space creation dynamics concatenation concept was investigated and classes were defined. Modeling resulted in independent and dependent concatenations’ classes. Agreement was at least 0.85 for all space creation dynamics classes. Afterwards, space creation dynamics classes were applied to match analysis. Analytic procedures encompassed both the frequency and types of space creation dynamics in ball possessions. The four best teams on the 2013–2014 NBA season were evaluated. The Thunder had more ball possessions with a single space creation dynamics than expected, while the Spurs had longer sequences than expected with four or more dependent concatenations. Interestingly, the Heat and the Thunder had the highest number of dependent concatenated space creation dynamics two steps ahead of the outcome. Additionally, the Heat was the team with greatest diversity of space creation dynamics and the highest frequency of dependent concatenations one step before the outcome. Results indicated that the analytical framework efficiently discriminated teams’ strategies. The concatenation concept helped elucidating how teams approach defensive disruption, with more straight or progressive space creation and concatenated space creation dynamics performed one or two steps before the outcome.

An important task of a basketball coach is to transfer information between game performance and team preparation. Therefore, the goals of this study were twofold: i) to define a framework encompassing the steps of team strategy, training practices, and game performance – the Team Learning Cycle (TLC); ii) to test TLC’s support for evaluation of team preparation-competition coherence with a junior basketball team. Team plays were assigned as an independent variable, systematically measured along the TLC. Frequency, diversity, and efficiency (points per possession) of plays performed in a game were compared both with alternatives of plays in the team strategy and emphasis during practices. TLC was implemented in a customized software for improving data acquisition reliability. We used a cluster analysis to group team plays according to similarities of offensive features and we applied Bayesian methods to compute the posterior distributions of the parameters describing minutes planned for team plays and training variables. Plays proportions were compared between practices and game. The 25 plays variations were grouped in seven clusters suggesting strategic diversity. Training presented significant tendencies towards offense phase, with opposition (emphasis on tactics) and situational practices (games and competition) – p(robability) > .90. The seven clusters of plays had a large variance in their training volume. The most frequently performed plays in the game were not those most trained but they had the most points per possession in the game. Results evidence the TLC may help coaches interpret the ongoing learning process of the team, improving team’s preparation.

The aim of this work was to define a model of volleyball drills’ structure. A set of parameters has been designed and tested for: i) pertinence and accuracy; ii) criteria reliability; iii) practical application. Expert judges evaluated model’s pertinence, accuracy and criteria reliability. A sample of 50 drills was assessed for drills’ analysis. The model demonstrated pertinence and accuracy, with complete agreement among experts. Criteria were reliable (Kappa test results > 0.8). Analysis indicated significant differences in the frequency of model’s parameters (graph topology), for instance among attributes (basic complex - 30%) with manifold drills (46,7%) in the technical domain (100%). The model contributes with theoretical support for a coach’s key task of designing practice contents.

No abstract