Sports coaches today have access to a growing amount of information that describes the performance of their players. Methods such as data mining have become increasingly useful tools to deal with the analytical demands of these high volumes of data. In this paper, we present a sports data mining approach using a combination of sequential association rule mining and clustering to extract useful information from a database of more than 400 high level beach volleyball games gathered at FIVB events in the years from 2013 to 2016 for both men and women. We regard each rally as a sequence of transactions including the tactical behaviours of the players. Use cases of our approach are shown by its application on the aggregated data for both genders and by analyzing the sequential patterns of a single player. Results indicate that sequential rule mining in conjunction with clustering can be a useful tool to reveal interesting patterns in beach volleyball performance data.
O. Ueberschär, D. Fleckenstein, F. Warschun, N. Walter, J. C. Wüstenfeld, B. Wolfarth and M. W. Hoppe
Hypogravity treadmills have become a popular training tool in distance running and triathlon. Counter-intuitively, tibial acceleration load is not attenuated by hypogravity unloading during running, while, equally surprisingly, leaps become flatter instead of higher. To explain these effects from a biomechanical perspective, Polet, Schroeder, and Bertram (2017) recently developed an energetic model for hypogravity running and validated it with recreational athletes at a constant jogging speed. The present study was conducted to refine that model for competitive athletes at relevant running speeds of 12–22 km h−1 and gravity levels of 100 %, 80 % and 60 %. Based on new experimental data on 15 well-trained runners in treadmill tests until volitional exhaustion, the enhanced semi-empirical model well describes energy expenditure and the observed biomechanical effects of hypogravity running. Remarkably, anaerobic contributions led to an increase in energy cost per meter for speeds above 16–18 km h−1 (p < 0.001), irrespective of hypogravity unloading. Moreover, some converging trends were observed that might reflect general adaptations in running motor control for optimization of efficiency. In essence, the outcome of this research might help sports scientists and practitioners to design running programs for specific training stimuli, e.g. conditioning of anaerobic energy metabolism.
Even several years after total hip (THR) and total knee replacement (TKR) surgery patients frequently show deficient gait patterns leading to overloads and relieving postures on the contralateral side or in the spine. Gait training is, in these cases, an essential part of rehabilitation. The aim of this study was to compare different feedback methods during gait training after THR and TKR focusing, in particular, on auditory feedback via sonification. A total of 240 patients after THR and TKR were tested in a pre-post-test design during a 3-week rehabilitation period. Even though sonification did not show, statistically, a clear advantage over other feedback methods, it was well accepted by the patients and seemed to significantly change gait pattern during training. A sudden absence of sonification during training led to a rapid relapse into previous movement patterns, which highlights its effectiveness in breaking highly automated gait patterns. A frequent use of sonification during and after rehabilitation could, hence, reduce overloading after THR and TKR. This may soon be viable, since new technologies, such as inertial measurement units, allow for wearable joint angle measurement devices. Back to normal gait with sonification seems possible.
Katharina Petri, Steffen Masik, Marco Danneberg, Peter Emmermacher and Kerstin Witte
We conducted a virtual reality (VR) training with ten sessions, performed by fifteen young karate athletes, who responded to attacks of a virtual opponent to improve their response behavior and their decision-making. The control groups continued with their normal training. Results of the Friedman tests with subsequent Dunn-Bonferroni post-hoc-tests and estimation of effect sizes showed that the karate specific response behavior (measured by a movement analysis) improved significantly due to the training. The parameters time for response (as the time for the attack initiation) and response quality improved with large effect sizes for the intervention groups, whereas the control groups demonstrated improvements with only small effect sizes. The unspecific response behavior (analyzed by two forms of the reaction test of the Vienna test system) did not show any significant changes. Paired t-tests revealed an improvement in attack recognition. While in the pretests, the intervention groups responded to late movement stages of the attack (execution of the main phase), they responded to early movement stages (reduction of distance and preparing steps) in the posttests. Furthermore, Friedman-tests and bivariate correlation analysis showed that the intervention groups were highly motivated to perform the VR training because of the new and safe learning conditions.
A. Edelmann-Nusser, A. Raschke, A. Bentz, S. Montenbruck, J. Edelmann-Nusser and M. Lames
Three inertial measurement unit (IMU) based tennis sensor systems from BABOLAT (PURE DRIVE PLAY, POP) and HEAD (Tennis Sensor) and a camera-based system (PlaySight) were tested with respect to the question whether the information about the number of strokes by swing type and spin type in training exercises and/or matches and the average as well as the maximum speed of the service per session are reliable. Subsequently, the question whether the mechanical properties of the BABOLAT PURE DRIVE PLAY racket are the same as the mechanical properties of the BABOLAT PURE DRIVE racket without IMU was addressed.
For swing types in standard exercises the results are acceptable for forehand groundstrokes, backhand groundstrokes and services but not for volleys. In a match environment we find inacceptably high errors (>10%) for the number of strokes for forehand and completely inacceptable levels for volley. The wrist-based IMU of BABOLAT POP has not reached an acceptable accuracy at all. For spin types the results are acceptable. The large variances in service speed assessment between devices make it doubtful whether any of them may be used for the control of training processes aiming at increasing the average service speed The mechanical properties of the BABOLAT rackets with and without IMU are quite the same.
Sotirios Drikos, Ioannis Ntzoufras and Nikolaos Apostolidis
In volleyball, due to the sequential structure of the game, each outcome results from events that follow consistent consecutive patterns: pass–set–attack–outcome, serve–outcome and block–dig–set–counter attack–outcome. There are three possible outcomes: point won, point lost, and rally continuation. With the aim of quantifying the importance of volleyball skills, data of world champions of the male International Volleyball Federation tournaments for three age categories (Youth, Juniors and Men) were used to construct a transition matrix between subsequent moves and skills within the game. A Dirichlet-Multinomial Bayesian model was used to estimate the transition probabilities between the subsequent moves along with the marginal probability of success of each skill in the complex. The prior distribution of each transition probabilities between moves/skills was elicited to incorporate experts' opinion. For the final evaluation of the skills a simple Monte Carlo scheme was applied to obtain a random sample from the posterior distribution. The findings of the study indicate that the relative importance of volleyball skills is robust across world champions of different age categories. Slight variations are observed on specific skills. A new index (Quantile Mid-range Ratio) is proposed for highlighting skills that are valuable for team’s gameplay.
The increasing availability of data from sports events has led to many new directions of research, and sports analytics can play a role in making better decisions both within a club and at the level of an individual player. The ability to objectively evaluate individual players in team sports is one aspect that may enable better decision making, but such evaluations are not straightforward to obtain. One class of ratings for individual players in team sports, known as plus-minus ratings, attempt to distribute credit for the performance of a team onto the players of that team. Such ratings have a long history, going back at least to the 1950s, but in recent years research on advanced versions of plus-minus ratings has increased noticeably. This paper presents a comprehensive review of contributions to plus-minus ratings in later years, pointing out some key developments and showing the richness of the mathematical models developed. One conclusion is that the literature on plus-minus ratings is quite fragmented, but that awareness of past contributions to the field should allow researchers to focus on some of the many open research questions related to the evaluation of individual players in team sports.
The purpose of the present study was to demonstrate an inductive approach for dynamically modelling sport-related injuries with a probabilistic graphical model. Dynamic Bayesian Network (DBN), a well-known machine learning method, was employed to illustrate how sport practitioners could utilize a simulatory environment to augment the training management process. 23 University of Iowa female student-athletes (from 3 undisclosed teams) were regularly monitored with common athlete monitoring technologies, throughout the 2016 competitive season, as a part of their routine health and well-being surveillance. The presented work investigated the ability of these technologies to model injury occurrences in a dynamic, temporal dimension. To verify validity, DBN model accuracy was compared with the performance of its static counterpart. After 3 rounds of 5-fold cross-validation, resultant DBN mean accuracy surpassed naïve baseline threshold whereas static Bayesian network did not achieve baseline accuracy. Conclusive DBN suggested subjectively-reported stress two days prior, subjective internal perceived exertions one day prior, direct current potential and sympathetic tone the day of, as the most impactful towards injury manifestation.
P. Browne, S. Morgan, J. Bahnisch and S. Robertson
In netball, analysis of the movement of players and the ball across different court locations can provide information about trends otherwise hidden. This study aimed to develop a method to discover latent passing patterns in women’s netball. Data for both pass location and playing position were collected from centre passes during selected games in the 2016 Trans-Tasman Netball Championship season and 2017 Australian National Netball League. A motif analysis was used to characterise passing-sequence observations. This revealed that the most frequent, sequential passing style from a centre pass was the “ABCD” motif in an alphabetical system, or in a positional system “Centre–Goal Attack–Wing Attack–Goal Shooter” and rarely was the ball passed back to the player it was received from. An association rule mining was used to identify frequent ball movement sequences from a centre pass play. The most confident rule flowed down the right-hand side of the court, however seven of the ten most confident rules demonstrated a preference for ball movement down the left-hand side of the court. These results can offer objective insight into passing sequences, and potentially inform team strategy and tactics. This method can also be generalised to other invasion sports.
K. Talattinis, G. Kyriakides, E. Kapantai and G. Stephanides
Realizing the significant effect that misprediction has on many real-world problems, our paper is focused on the way these costs could affect the sports sector in terms of soccer outcome predictions. In our experimental analysis, we consider the potential influence of a cost-sensitive approach rather than traditional machine-learning methods. Although the measurement of prediction accuracy is a very important part of the validation of each model, we also study its economic significance. As a performance metric for our models, the Sharpe ratio metric is calculated and analyzed. Seeking to improve Sharpe ratio value, a genetic algorithm is applied. The empirical study and evaluation procedure of the paper are primarily based on English Premier League’s games, simple historical data and well-known bookmakers’ markets odds. Our research confirms that it is worthwhile to employ cost-sensitive methods for the successful predictions of soccer results and better investment opportunities.