Proceedings of the 11th International Symposium on Computer Science in Sport (IACSS 2017)

This study, taking the game results of 3800 matches of the English Premier League (2006/2007–2015/2016 seasons) and their odds offered by 347 lottery corporations worldwide as samples, built indices like Odds Adjustment Difference (OAD), Odds Difference Level, as well as Concordance Rate (CR), Draw Rate (DR), Reversal Rate (RR) between instant odds and game results, and applied those indices to the analysis of basic characteristics of European odds (also known as decimal odds). In addition, this study also put forward Odds Difference Coefficient (ODC), and compared it with the Competitive Balance Entropy (CBE) of the English Premier League, leading to the following conclusions: (1) The relationship between odds and the winning percentage followed a curve similar to “L” with an obvious inflection point; (2) In the English Premier League, the CR, DR and RR between instant odds and game results were 53.6, 25.8 and 20.6% respectively. The OAD between instant odds and low initial odds was the smallest, followed by that between instant odds and odds on draw, and that between instant odds and high initial odds; (3) The difference between low and high initial odds had a significant influence on the CR, DR and RR; to be more specific, when the odds difference increased, the CR would be significantly increased while the DR and RR were significantly reduced. The CR, DR and RR of the English Premier League were relatively stable in the recent 10 years; (4) The CBE of the English Premier League had a moderate negative correlation with its initial ODC; (5) Gamblers can refer to the OAD, and instant odds on CR, DR and RR, as well as ODC and CBE when betting on football matches, and some formulas and models are used by lottery corporations for analysis.

This study takes match outcomes of European eight leagues, China Super League (CSL) and Japan J1 League (J1) in 20 seasons (1996–2015) as samples, and proposes UNIANOVA (multi-factors: home or away game, results of the previous games, combination of the two) to explore the Hot Match Effect in professional football leagues. The results show that (1) “Home advantage” phenomenon is significant in professional football leagues; (2) Hot Match Effect exist in professional football leagues, and is obvious in Home matches; (3) the interaction effect between the two factors (home or away, results of the previous games) is significant in terms of winning percentage.

Internal forward models are used to explain motor prediction processes in motor control and learning e.g. predicting an upcoming miss in a throwing task before the knowledge of results is available. In this study we used artificial neural networks (ANN) to model such movement outcome prediction processes. Additionally, we varied the inputs of four different multilayer perceptrons (MLP) with respect to the quantity and the reliability (quality) of input parameters to account for perceptual noise. The results show that ANNs are able to learn the non-linear input-output mapping underlying the throwing task even with few input variables (velocity and angle at ball release). Results improve when providing additional information about the ball flight (prediction accuracy increases from RMSE = 7.9 mm to RMSE = 3.9 mm). However, when a model is provided with noisy inputs only, model training and prediction suffers substantially (RMSE = 53.8 mm). Yet, additional reliable information about the ball flight (in addition to noisy velocity and angle) leads to very high model prediction accuracy again (RMSE = 4.1 mm). In a nutshell, ANNs can be used to model internal forward model predictions, but the availability of reliable input information is essential at least to some extent.

In this paper, we propose an algorithmic approach for a motion analysis framework to automatically recognize local muscular endurance (LME) exercises and to count their repetitions using a wrist-worn inertial sensor. LME exercises are prescribed for cardiovascular disease rehabilitation. As a technical solution, we propose activity recognition based on machine learning. We developed an algorithm to automatically segment the captured data from all participants. Relevant time and frequency domain features were extracted using a sliding window technique. Principal component analysis (PCA) was applied for dimensionality reduction of the extracted features. We trained 15 binary classifiers using support vector machine (SVM) to recognize individual LME exercises, achieving overall accuracy of more than 98%. We applied grid search technique to obtain the optimal SVM hyperplane parameters. The learning curves (mean ± stdev) for each model is investigated to verify that the models were not over-fitted and performed well on any new test data. Also, we devised a method to count the repetitions of the upper body exercises.

This study tests correlation entropy, \(K_2\), as a measure of stability for gait analysis. An average of 13 strides from 10 participants in each combination of one footwear (barefoot vs shod) condition and one gait mode (walking vs running) were collected during treadmill walking and running. Sagittal plane ankle, knee and hip angular displacement and velocity data were used for analysis. Two-way repeated measures ANOVA showed a main effect for gait mode (\(p=.03\)) – running had lower \(K_2\) than walking, indicating higher stability. Although the sample of strides and participants was small, we speculate that the greater inertia for running helped stabilize movement control, making the running coordination pattern more resilient against small stride-to-stride perturbations.

Aim of the study was to identify possible predictors influencing the variability of individual short-term heart rate (HR) responses to submaximal interval exercise using a probabilistic model. Short-term HR responses to the change of load bouts obtained in a twelve-week training intervention were analyzed. Questionnaires gathered preceding sport activity, sleep, nutrition, health and mood prior to each training session. Additionally, time of the day and number of interval was included in calculation. Multiple regression method was used to identify predictors for start heart rate (HR), steady state HR, and for the slope of the HR curve. Especially the number of the interval, physical and mental health, and negative mood were influencing these responses. The start heart rate was identified as predictor in five of eight response parameter. Time was a factor highly varying between participants. Future research need to validate the results in a wider sample and integrate more parameters in the analysis.

In professional football clubs, information from many different sources has to be joined for informed decisions on team management. A central club management information system (CMIS) could help to support the employees at the club with information that they need for their daily decisions. This paper analyses the information demand and lists requirements for an optimal solution for the problem. With a beta version of a CMIS-software, field tests are conducted to improve the understanding of the needs and habits of employees at the club. This information will run into the development of the next version of this software. This iterative process aims at a final solution that meets the needs of football clubs.

According to the official rule of water polo, player substitution is allowed whenever the coach needs without restriction of times. Thus, information of total playing time for each player will be effective in real-time for the water polo coach as one of decisive information for player substitution. The purpose of present study is to develop a system which enables water polo coach to grasp the playing time of each player by using a tablet device during the match in real-time. A relational database software FileMaker Pro12 and mobile template FileMaker Go were used to develop playing time calculating system for water polo, and an iPad-mini was used as inputting apparatus. An operator with iPad-mini observe the match and input official playing time information and the cap number of player who participating on the water polo field at every substitution time. Total playing time for each player are showed on the same screen as inputting on iPad, which enables the coach to know objective information concerning the opportunity of player substitution.

Monocular 3D reconstruction is an important problem. We solve the problem of reconstructing a 3D ball or shuttle position from a single-view television video. The contextual constraint is vital in this paper, which is implemented by a confirming point. The confirming point represents all the contextual cues, such as human pose, stroke technique, and shadow on the ground. The confirming point also tells us where the 3D point is. The confirming point decision is made by a human operator. Thus, the proposed method is a mixture of computer vision and human intelligence. Moreover, we propose a new air-ball friction model. This model provides a more accurate result because the aerodynamic drag force cannot be ignored in ball game.

Gathered kinematic data usually requires post-processing in order to handle noise. There a three different approaches frequently used: local regression & moving average algorithms, and Butterworth filters. In order to examine the most appropriate post-processing approach and its optimal settings to human upper limb movements, we examined how far the approaches were able to reproduce a simulated movement signal with overlaid noise. We overlaid a simulated movement signal (movement amplitude 80 cm) with normal distributed noise (standard deviation of 0.5 cm). The resulting signal was post-processed with local regression and moving average algorithms as well as Butterworth filters with different settings (spans/orders). The deviation from the original simulated signal in four kinematic parameters (path length, maximum velocity, relative activity, and spectral arc length) was calculated and checked for a minimum. The unprocessed noisy signal showed absolute mean deviations of 54.78% ± 12.16% in the four kinematic parameters. The local regression algorithm revealed the best performance at a span of 420 ms with an absolute mean deviation of 2.00% ± 0.86%. For spans between 280–690 ms the local regression algorithm still revealed deviations below 5%. Based on our results we suggest a local regression algorithm with a span of 420 ms for smoothing noisy kinematic data in upper limb performance, e.g., activities of daily living. This suggestion applies to kinematic data of human movements.

With modern cycling computers it is possible to provide cyclists with complex feedback during rides. If the feedback is course-dependent, it is necessary to know the riders current position on the course. Different approaches to estimate the position on the course from common GPS and speed sensors were compared: the direct distance measure derived from the number of rotations of the wheel, GPS coordinates projected onto the course trajectory, and a Kalman filter incorporating speed as well as GPS measurements. To quantify the accuracy of the different methods, an experiment was conducted on a race track where a fixed point on the course was tagged during the ride. The Kalman filter approach was able to overcome certain shortcomings of the other two approaches and achieved a mean error of \({-0.13}\,{\text {m}}\) and a root mean square error of \({0.97}\,{\text {m}}\).

Goal-directed reaching movements in three-dimensions are important for our interaction with the environment. Instrumented setups displaying virtual targets for rehabilitation training of reaching movements often provide limited depth cues, which may affect movements. This work aims to quantify effects of limited depth cues on reaching movements. We developed a virtual environment for assessing three-dimensional reaching movements that allows different depth cues to be enabled or disabled. By imposing a fixed spatial tolerance around targets for speed-accuracy trade-off, completion time normalized with the straight-line distance to the target was used to measure reaching performance. In the present study, 8 (control) subjects using a typical monitor setup applied in rehabilitation were compared to 7 subjects using a head-mounted display and receiving additional depth cues, namely, hard-referenced objects of known size, motion parallax due to tracked head-movements, and stereopsis. Control subjects required on average 9.88 s/m straight-line distance, while subjects using the head-mounted display required only 3.15 s/m straight-line distance. Additionally, movement trajectories of control subjects showed a different pattern, indicating a lack of reliable depth information. Thus, state-of-the-art rehabilitation setups are challenging already for healthy subjects. This challenge can be reduced by improving the provided visual depth cues.

Virtual Reality (VR) has become common practice in the field of sports, but autonomous virtual environment (VE) systems, especially in fast reacting sports, are rare. The current study demonstrates the development of an autonomous character (AC) in karate kumite, which performs attacks against a freely moving, real athlete. The development of the AC consists of four steps: selection of relevant karate techniques, development of a decision system, creation of an animated model of the AC, and the evaluation. A Cave Automatic Virtual Environment (CAVE) and a Head Mounted Display (HMD) were chosen for the VE. The evaluation of the AC in the VEs was conducted by expert interviews (n = 6). The results reveal a feeling of comfort for all athletes in VR which underpins a high degree of realism in the VEs. Moreover, the HMDs are seen as more suitable than CAVEs for presenting a karate specific environment. Based on these results the developed AC seems applicable for anticipation research and training in karate kumite. The discussion includes further possible improvements for the AC as well as future directions for further investigations and training programs using the AC. Moreover, the procedure of the AC’s creation can be transferred to other sports.

In this paper, cross-sectional surveys of students’ access to, use of and attitude towards computer, the Internet, mobile technologies and eLearning are described covering a period of 15 years. A sample of 1,106 students (age: M = 21.5 years, SD = 2.9) of sport science participated in the study. They completed a questionnaire addressing the topics mentioned above. Main findings are that almost every student has access to the computer, mobile ICT, and the Internet. There are differences between study programs favoring students attending study programs that combine sport and computer science. Furthermore, attitudes have changed over the past 15 years. Finally, gender differences concerning the use of and attitudes to the computer and the Internet but not eLearning still exist. Important limitations of the studies are the sequential cross-sectional design and sample distortions.

The overlap between the workspaces of humans and robots has been increasing in the last decades. Situations in which humans interact with robots are becoming more frequent. The uni- and bidirectional interactions of humans and robots when learning movements have so far not been adequately investigated. The presented studies investigate the unidirectional interaction of humans and robots in different settings. Results present first indications for an efficient and effective interaction configuration of a bidirectional interaction between human and robot for the learning of movements and to combine their advantages of humans and robots.

Muscle fibers composition is determined by the specific genetic characteristics to a large extent and it is connected with various physical traits of athletes. Establishment of genetic markers for sport prediction is an important task of great scientific and practical importance. We aimed to create a relevant model for prediction of the athlete muscle fibers structure based on a complex analysis of known genetic factors. The model included 14 single nucleotide polymorphisms and was tested for 55 subjects. Based on the performed ROC analysis, the model accuracy measured as area under the receiver operating characteristic curve (AUC) was 81% for professional athletes and 73% for non-athletes. The obtained results demonstrate that the proposed models can be used for sport testing.