Accuracy of pattern detection methods in the performance of golf putting

Decision making influences movement variability and performance of high-level female football players in an elastic resistance task

Introduction

The inclusion of sport-specific constraints in resistance training promotes the development of player abilities in an integrated way, which maximises the effectiveness of player adaptations induced by training. Considering that perceptual-cognitive abilities play a fundamental role in football, decision making could be introduced to enhance the cognitive similarity of resistance tasks to sport actions. However, it is unknown how decision making as a constraint could affect the player during an elastic resistance task. Therefore, the aim of this study was to investigate the effects of decision making of high-level female football players on movement variability and performance during an elastic band resistance task.

Methods

Twenty-three high-level female football players performed the elastic resistance task with a ball, both as attackers and as defenders without decision making (NDM) and with decision making (DM). The movement variability was quantified using the sample entropy derived from the acceleration recorded with an accelerometer placed at the lower back of each player. The passing accuracy of the attacker was quantified using a scoring scale.

Results

Results revealed that adding decision making to an elastic resistance task increased the movement variability of the defender but did not affect the movement variability of the attacker. In contrast, the passing accuracy of the attacker was reduced. Overall, the attacker had a higher movement variability compared to the defender.

Discussion

These findings suggest that decision making, as a football-specific constraint, can enhance the potential of an elastic resistance task in training. This is due to the fact that it reduces control and regularity of movement for the defensive role player and increases technical difficulty for the attacking role player. Furthermore, these effects are beneficial, as they can promote the adaptive processes necessary to optimise the performance of the players.

A sense of distance and movement characteristics of golfers tested without visual feedback of outcomes: Is a putt that feels subjectively good also physically good?

For motor tasks that require fine-tuning, such as adjusting the applied force according to the distance to the target, as required for driving and putting in golf, it is important to develop a high degree of sensitivity of one's movement-produced feedback. In previous research focusing on mental representation in golf, this ability to control distance has been called “a sense of distance”. In particular, this study focused on three skills: motor control of the putter head, perception of the impact force, and prediction of the ball's travel distance. However, the relationship between the motor control of the putter head and the error of the outcome estimation is yet to be clarified. The purpose of the present study is two-fold: first, to clarify whether kinematic variation of putter head is correlated with error in estimating the outcome and, second, to quantitatively evaluate the performer's sensations of good and poor performance generated by the ball's impact, for a comparison of the kinematics and impact force of the putter head based on their assessment. Twelve professionals and 12 intermediate amateurs played two distance targets (at 2.4 and 4.8 m) without visual feedback of the outcomes. The kinematics of the putter head, impact force, final ball position, outcome estimation, and subjective assessment were measured. Our results show that the variability in the peak velocity was moderately correlated with the error of the outcome estimation in amateurs' 4.8-m putting task. In addition, amateurs estimated undershoots (overshoots) when they provided worse (better) evaluations. However, the amateurs' trials that were rated as better putts were actually overshoots. These results suggest that the subjectively “good putt” of amateurs was not physically good, and the amateurs putted hard to compensate for the risk of undershoots. However, no remarkable feature of the professional's sense of distance was found. This suggests that professional golfers' sensation is not significantly different from the outcomes that can be physically detected.

Skill Differences in a Discrete Motor Task Emerging From the Environmental Perception Phase

Because of the challenges associated with measuring human perception and strategy, the process of human performance from perception to motion to results is not fully understood. Therefore, this study clarifies the phase at which errors occur and how differences in skill level manifest in a motor task requiring an accurate environmental perception and fine movement control. We assigned a golf putting task and comprehensively examined various errors committed in five phases of execution. Twelve tour professionals and twelve intermediate amateur golfers performed the putting task on two surface conditions: flat and a 0.4-degree incline. The participants were instructed to describe the topographical characteristics of the green before starting the trials on each surface (environmental perception phase). Before each attempt, the participants used the reflective markers to indicate their aim point from which the ball would be launched (decision-making phase). We measured the clubface angle and impact velocity to highlight the pre-motion and motion errors (pre-motion and motion phase). In addition, mistakes in the final ball position were analyzed as result errors (post-performance phase). Our results showed that more than half of the amateurs committed visual–somatosensory errors in the perception phase. Moreover, their aiming angles in the decision-making phase differed significantly from the professionals, with no significant differences between slope conditions. In addition, alignment errors, as reported in previous studies, occurred in the pre-motion phase regardless of skill level (i.e., increased in the 0.4-degree condition). In the motion phase, the intermediate-level amateurs could not adjust their clubhead velocity control to the appropriate level, and the clubhead velocity and clubface angle control were less reproducible than those of the professionals. To understand the amateur result errors in those who misperceived the slopes, we checked the individual results focusing on the final ball position. We found that most of these participants had poor performance, especially in the 0.4-degree condition. Our results suggest that the amateurs’ pre-motion and strategy errors depended on their visual–somatosensory errors.

Victory Prediction of Ladies Professional Golf Association Players: Influential Factors and Comparison of Prediction Models

This study aims to identify the most accurate prediction model for the possibility of victory from the annual average data of 25 seasons (1993–2017) of the Ladies Professional Golf Association (LPGA), and to determine the importance of the predicting factors. The four prediction models considered in this study were a decision tree, discriminant analysis, logistic regression, and artificial neural network analysis. The mean difference in the classification accuracy of these models was analyzed using SPSS 22.0 software (IBM Corp., Armonk, NY, USA) and the one-way analysis of variance (ANOVA). When the prediction was based on technical variables, the most important predicting variables for determining victory were greens in regulation (GIR) and putting average (PA) in all four prediction models. When the prediction was based on the output of the technical variables, the most important predicting variable for determining victory was birdies in all four prediction models. When the prediction was based on the season outcome, the most important predicting variables for determining victory were the top 10 finish% (T10) and official money. A significant mean difference in classification accuracy was observed while performing the one-way ANOVA, and the least significant difference post-hoc test showed that artificial neural network analysis exhibited higher accuracy than the other models, especially, for larger data sizes. From the results of this study, it can be inferred that the player who wants to win the LPGA should aim to increase GIR, reduce PA, and improve driving distance and accuracy through training to increase the birdies chance at each hole, which can lead to lower average strokes and increased possibility of being within T10.

The Influence of Functional Flywheel Resistance Training on Movement Variability and Movement Velocity in Elite Rugby Players

The aim of this study was to identify the changes in movement variability and movement velocity during a six-week training period using a resistance horizontal forward-backward task without (NOBALL) or with (BALL) the constraint of catching and throwing a rugby ball in the forward phase. Eleven elite male rugby union players (mean ± SD: age 25.5 ± 2.0 years, height 1.83 ± 0.06 m, body mass 95 ± 18 kg, rugby practice 14 ± 3 years) performed eight repetitions of NOBALL and BALL conditions once a week in a rotational flywheel device. Velocity was recorded by an attached rotary encoder while acceleration data were used to calculate sample entropy (SampEn), multiscale entropy, and the complexity index. SampEn showed no significant decrease for NOBALL (ES = -0.64 ± 1.02) and significant decrease for BALL (ES = -1.71 ± 1.16; p < 0.007) conditions. Additionally, movement velocity showed a significant increase for NOBALL (ES = 1.02 ± 1.05; p < 0.047) and significant increase for BALL (ES = 1.25 ± 1.08; p < 0.025) between weeks 1 and 6. The complexity index showed higher levels of complexity in the BALL condition, specifically in the first three weeks. Movement velocity and complex dynamics were adapted to the constraints of the task after a four-week training period. Entropy measures seem a promising processing signal technique to identify when these exercise tasks should be changed.

Practice Motions Performed During Preperformance Preparation Drive the Actual Motion of Golf Putting

Of the various types of preperformance preparatory behavior that are acquired during motor learning, the effect of a practice motion performed just prior to execution of an actual motion is not yet fully understood. Thus, the present study employed a golf putting task to investigate how a practice motion in the preparation phase would affect the accuracy of motor control in the execution phase and how proficiency would influence this relationship. To examine the impacts on kinematics and final ball position, the velocities of practice strokes made by tour professional and amateur golfers were experimentally manipulated in the following three conditions: the equal condition, which presented a target that was at the same distance during the practice strokes and the actual stroke; the confusing condition, which had two different distances during the practice and actual strokes; and the no condition, which did not include a practice stroke. The results, based on final ball position, indicated that practice strokes in the equal condition were linked with the highest accuracy levels during the actual stroke in both professionals and amateurs. In the confusing condition, regardless of skill level, the velocity of the actual stroke was influenced by a faster or slower stroke during the pre-shot phase. These relationships between the practice and actual strokes imply that the golfers effectively utilized kinesthetic information obtained during the practice strokes as a reference for the actual stroke. Furthermore, the differences in proficiency level indicated that the club head velocity of amateurs in the no condition was significantly faster than in the equal condition. Therefore, the present results imply that the role of a practice stroke may differ between professionals and amateurs.

Structure of force variability during squats performed with an inertial flywheel device under stable versus unstable surfaces

The use of unstable surfaces during resistance training has demonstrated a maintenance or reduction on force production. However, the use of unstable surface on force variability has not been assessed using non-linear methods that may be better suited to detect changes in movement variability throughout a given movement. Consequently, this study compared the use of stable vs unstable surfaces on force variability during bilateral squats performed with an inertial flywheel device (Eccoteck, Byomedic System SCP, Spain). Twenty healthy men (mean ± SD: age 22.9 ± 2.9 years, height 1.81 ± 0.7 m, body mass 76.4 ± 7.6 kg and 1RM back squat 110.9 ± 19.7 kg) with a minimum of four years in resistance training performed six sets of six repetitions of squats at maximal concentric effort with one minute rest between sets. Force output on the vertical axes was measured using a strain gauge and the results were processed using non-linear sample entropy (SampEn). Results showed no differences for any of the dependent variables between stable and unstable conditions. SampEn showed no differences between conditions (chi-squared = 0.048 P = 0.827), while Force_mean and SampEn presented a small correlation (r = 0.184; p < 0.01). No changes in entropy were found over the course of the series. Together, these results suggest that the structure of force variability between stable and unstable surfaces are similar. This lack of difference between surfaces may be due to postural and anticipatory adjustments. Consequently, by introducing unstable surfaces to the flywheel bilateral squat exercise, practitioners may not observe changes in Force_mean and force variability when compared to stable surface training suggesting that increased training volumes or intensity may be required during unstable environments to cause a desired training stimulus.

Motor control of practice and actual strokes by professional and amateur golfers differ but feature a distance-dependent control strategy

We explored how practice and actual putting strokes differed between professionals and high-level golf amateurs, and how practice strokes reflected subtle differences in putting distances. We analysed swing amplitude, impact velocity, and acceleration profile of the club-head. The acceleration profiles showed that the motor control pattern of the practice stroke differed from that of the actual stroke. To clarify the effects of different putting distances on the practice stroke and to analyse how much the actual stroke could be explained by the practice stroke, we conducted individual regression analyses. The practice strokes of all participants could be divided into three strategies and five types by the coefficient of determination and the slope. This implies that the purpose of the practice stroke varied among golfers. Most golfers used the individual velocity criteria in their practice strokes, which resulted in different putting distances based on their criteria. Unexpectedly, we found no significant difference in skill level between professionals and high-level amateurs. The results of this study imply that the practice stroke does not duplicate the actual stroke, even for professional golfers with excellent skills. However, most high-level golfers adopted distance-dependent control strategies for slightly different putting distances.

Machine and deep learning for sport-specific movement recognition: a systematic review of model development and performance

Objective assessment of an athlete's performance is of importance in elite sports to facilitate detailed analysis. The implementation of automated detection and recognition of sport-specific movements overcomes the limitations associated with manual performance analysis methods. The object of this study was to systematically review the literature on machine and deep learning for sport-specific movement recognition using inertial measurement unit (IMU) and, or computer vision data inputs. A search of multiple databases was undertaken. Included studies must have investigated a sport-specific movement and analysed via machine or deep learning methods for model development. A total of 52 studies met the inclusion and exclusion criteria. Data pre-processing, processing, model development and evaluation methods varied across the studies. Model development for movement recognition were predominantly undertaken using supervised classification approaches. A kernel form of the Support Vector Machine algorithm was used in 53% of IMU and 50% of vision-based studies. Twelve studies used a deep learning method as a form of Convolutional Neural Network algorithm and one study also adopted a Long Short Term Memory architecture in their model. The adaptation of experimental set-up, data pre-processing, and model development methods are best considered in relation to the characteristics of the targeted sports movement(s).

Resolution of low-velocity control in golf putting differentiates professionals from amateurs

It is difficult for humans to apply small amounts of force precisely during motor control. However, experts who have undergone extended training are thought to be able to control low-velocity movement with precision. We investigated the resolution of motor control in golf putting. A total of 10 professional and 10 high-level amateur golfers participated. Putting distances were 0.6-3.3 m, in increments of 0.3 m. We measured the impact velocity and the club-face angle at impact, and the acceleration profile of the downswing. The professionals showed significantly smaller coefficients of variation with respect to impact velocity and smaller root mean square errors in relation to acceleration profiles than did the amateurs. To examine the resolution of motor control for impact velocity, we investigated intra-participant differences in the impact velocity of the club head at two adjacent distances. We found that professionals had higher velocity precision when putting small distance intervals than did amateurs. That is, professionals had higher resolution of low-velocity control than did high-level amateurs. Our results suggest that outstanding performance at a task involves the ability to recognise small distinctions and to produce appropriate movements.

Modeling temporal sequences of cognitive state changes based on a combination of EEG-engagement, EEG-workload, and heart rate metrics

The objective of this study was to investigate the feasibility of physiological metrics such as ECG-derived heart rate and EEG-derived cognitive workload and engagement as potential predictors of performance on different training tasks. An unsupervised approach based on self-organizing neural network (NN) was utilized to model cognitive state changes over time. The feature vector comprised EEG-engagement, EEG-workload, and heart rate metrics, all self-normalized to account for individual differences. During the competitive training process, a linear topology was developed where the feature vectors similar to each other activated the same NN nodes. The NN model was trained and auto-validated on combat marksmanship training data from 51 participants that were required to make "deadly force decisions" in challenging combat scenarios. The trained NN model was cross validated using 10-fold cross-validation. It was also validated on a golf study in which additional 22 participants were asked to complete 10 sessions of 10 putts each. Temporal sequences of the activated nodes for both studies followed the same pattern of changes, demonstrating the generalization capabilities of the approach. Most node transition changes were local, but important events typically caused significant changes in the physiological metrics, as evidenced by larger state changes. This was investigated by calculating a transition score as the sum of subsequent state transitions between the activated NN nodes. Correlation analysis demonstrated statistically significant correlations between the transition scores and subjects' performances in both studies. This paper explored the hypothesis that temporal sequences of physiological changes comprise the discriminative patterns for performance prediction. These physiological markers could be utilized in future training improvement systems (e.g., through neurofeedback), and applied across a variety of training environments.