Behavioral and cortical dynamics underlying superior accuracy in short-distance passes

Improved pass accuracy is a prominent determinant of success in football. It demands an effective interaction of complex behavioral and cortical dynamics. Exploring differences in the ability to sustain an accurate pass behavior in a stable setting and the associated cortical dynamics at different expertise levels may provide an insight into skilled strategies contributing to superior accuracy in football. The aim of this study is to compare trial-to-trial variability of pass biomechanics and the corresponding cortical dynamics during short-distance passes between novices and experienced football players. Thirty participants (15 novices, 15 football players) performed 90 short-distance passes. The intertrial variability of pass biomechanics (foot acceleration, range of hip flexion, knee flexion and foot rotation) was assessed by means of multiscale entropy. The task-related cortical dynamics were analyzed via source-derived event-related spectral perturbations. Experienced players demonstrated higher accuracy and overall lower entropy values across multiple time scales which was significant for hip flexion. The electroencephalography data revealed group differences in parieto-occipital alpha desynchronization and frontal theta synchronization in successive phases of passes. The current findings suggest that experienced football players may show a skilled ability to recruit and retain pass biomechanics promoting higher accuracy, whereas novices may show an explorative behavior with higher spatial variability. This difference may be associated with distinctive visuospatial and attentional strategies acquired with expertise in football. Our study provides an insight into expertise-specific behavioral and cortical dynamics of superior accuracy in football and a basis for its prospective investigation in enriched contexts.


Introduction
Football is one of the most played sports in the world and sporting success is undoubtedly the core endeavor of every team [1,2].The variables designating the result of a football game have been studied extensively and pass accuracy was reported to be a prominent determinant of success [3][4][5].Either in terms of opportune passes to a teammate or successful kicks toward the goal, kicking with precision increases the likelihood of winning and is underlined as a technical component contributing to tactical skills [6,7].Especially accurate short-distance passes retain ball possession during a game, provide different scoring opportunities and increase the number of shots on target [8][9][10].Boosting this sport-specific skill has therefore drawn the attention of recent studies [11][12][13][14].
The superior skills of athletes rely on years of practice and consequently achieved strategies operating at different domains [15].Among these, higher-level strategies are described as neural features such as efficient cognitive and sensorimotor processes which thrive specific to sport type [16][17][18] and enhance sport performance in competitive athletes [15,16].To date, studies have manifested distinctive brain dynamics associated with superior performance in experts, which points at the significance of higher-level strategies in sport-specific performance [18].Especially for target-directed tasks, EEG studies have asserted attentional and sensory processes as higher-level strategies linked with superior performance at higher skill levels [19][20][21].In football, passing a ball accurately toward a target requires the successful translation of visuospatial and proprioceptive information into a coordinated motor output with high cortical demands [22,23].Two recent EEG studies have analyzed cortical dynamics with different foci and settings and provided a preliminary insight ino these demands [24,25].Palucci-Vieira et al [24].have explored long-distance kicks and reported the influence of frontal theta and posterior alpha activity on ball velocity and radial error, respectively, in skilled players.In a repeated-measures design, Piskin et al [25].have shown the reliable activity of alpha and theta oscillations in posterior and frontal cortices during short-distance passes executed by novices as accurately as possible.With these findings, both studies remarked on the significance of attentional and sensory processes which are indexed by characteristic task-related responses of theta and alpha oscillations.However, it is still unknown if and how these dynamics may fluctuate with different levels of expertise and facilitate an improved passing behavior.
The behavioral underpinnings of an accurate pass and kick have been analyzed with manifold variables such as muscular activity of the kicking leg [26], different kicking techniques [27] and biomechanics of different body segments [28,29].Although these variables may provide important descriptive dynamics associated with accurate passes, they may lack information about higher-level processes promoting advanced passing behavior in football.However, the ability to produce an accurate behavior consistently may shed light on skilled neural mechanisms such as predicting the sensory consequences of a movement correctly (feedforward loops), integrating predicted sensory information to actual state (state estimation), adjusting the input congruently based on sensorimotor outcomes (feedback loops) [15,16,30] and recruiting attentional strategies to optimize these sensorimotor processes [31].Regarding this, movement variability is a commonly studied behavioral endpoint in sports contexts to understand skilled performance [32][33][34][35][36][37].While increased intertrial variability may be advantageous in terms of adapting efficiently to dynamic environments with a rich movement repertoire [36,37], decreased intertrial variability may show the ability to predict and sustain congruent movement parameters facilitated by the aforementioned mechanisms, which may diminish noise and enhance accuracy in stable settings [30,36,38,39].In this regard, sample entropy (SampEn) is a non-linear measure to analyze movement variability and its relation to performance in human movement [40][41][42].It appraises the predictability of a time series and produces lower values if a certain pattern is detected consistently, i.e. if the series depicts less variability.Although SampEn provides information about spatiotemporal fluctuations during a movement, it is limited to a single time scale.Its modification of multiscale entropy (MSE) conceives regularity not only in the original series, but also in its derivatives at longer time scales which are achieved by means of coarse-graining [43].Herewith, it integrates the complexity of a system across different temporal resolutions and reveals dynamics operating not only at lower, but also at higher physiological levels.Longer time scales may show how the complexity of fast behavioral dynamics evolve into the ability to maintain a gross motor behavior over an extended period of time [44,45].The MSE profile of trial-to-trial pass biomechanics at different expertise levels may therefore help to understand differences in the ability not only to pursue accurate fast dynamics within a pass, but also to pursue an accurate gross behavior consistently [46].Differences, particularly in the latter, may propose the influence of distinctive neural processes such as sport-specific higher-level strategies developed over time [19][20][21]42,45].Furthermore, the fusion of MSE and EEG analyses may be a complementary approach to understand and describe skilled higher-level strategies in the light of the behavioral context, which may provide a contextual guide for interventions, aiming to facilitate or challenge these strategies in different levels of expertise and advance short-distance pass accuracy in football [47].
Therefore, in this study, we aim to compare trial-to-trial movement variability at multiple time scales between novices and experienced football players.Furthermore, we aim to compare pass-related cortical dynamics in order to confirm and describe differences in neural mechanisms suggested by behavioral findings.To quantify and compare intertrial variability in the accelerative and spatial characteristics of passes, we will compute MSE for foot acceleration, hip and knee flexion and foot external rotation.We will investigate associated cortical dynamics as measured by event-related spectral perturbations (ERSP) based on our previous study [25].We hypothesize to find overall higher entropy values for pass biomechanics in novices [48][49][50] and observe group differences in pass-related posterior alpha and frontal theta oscillations [25].

Participants
Fifteen healthy novices (7 female, mean age: 26.87 ± 3.54 years) and 15 healthy amateur football players (6 female, mean age: 22.40 ± 3.62 years) participated in the study.Their demographics are presented in Table 1.The inclusion criteria for novices were defined as: i) 18-35 years old, ii) right-dominant in lower extremity, iii) not playing/ having played football [21,51] iv) not playing/ having played any other sport on a regular basis in the last year which requires running, cutting and pivoting [52], and v) not having any neurological diseases, previous orthopedic injuries and being on psychotropic medication.Football players were recruited based on the criteria: i) 18-35 years old, ii) right-dominant in lower extremity, iii) actively playing football for at least 10 years, iv) training at least twice and playing one game per week, v) not having any neurological diseases, previous orthopedic injuries and being on psychotropic medication.The laterality of the lower extremity was ascertained based on the Lateral Preference Inventory [53].The participation in sports and activity level were examined using custom questionnaires and the Marx Activity Scale [54].All participants had normal or corrected vision during the experiments.

Experimental procedure
Upon completing preparations, the participants executed shortdistance passes to a rectangular target (10×15 cm) placed at 3 m [25].A FIFA size 5 ball was placed on a standard location perpendicular to the target.The participants positioned their left foot neutrally in alignment with the ball, and their right foot externally rotated behind the ball.They were instructed to kick the ball with the inside of their right foot as accurately as possible [25,55].Prior to the experiment, they performed trial passes to assign the optimal position for their left foot and familiarize themselves with the task.Once they confirmed the location of their left foot for optimal performance, it was marked to maintain it standardized throughout all passes (The experimental setting is presented in the supporting file).The number of trial passes did not exceed 10 repetitions per participant, and subsequently, 6 blocks of 15 experimental repetitions were completed.Meanwhile, the biomechanics of passes and cortical activity were recorded, as well as the total number of hits and misses to calculate accuracy rate in percentage.At the end of the experiment, the participants reported a perceived stress level on a visual analog scale (VAS) in order to account for the possible influence of stress on accuracy [56].

Sample and multiscale entropy analysis
Three-dimensional (x, y, z) biomechanics of the lower extremities were recorded for each block using an inertial measurement unit (IMU) system (myoMOTION, Noraxon, USA) at a sampling rate of 200 Hz.IMU systems were shown to produce accurate results for the quantification of kicking biomechanics [57].Six wearable sensors were attached bilaterally on the dorsal surfaces of the feet, on the anteromedial tibial surfaces of the shanks and the lateral lower quadrants of the upper thighs, whereas the pelvis sensor was placed on the sacral surface of the pelvis [58].The digitized signal of foot acceleration in the x-axis [23], the anatomical motion angles for hip flexion, knee flexion and foot external rotation were exported using myoRESEARCH (version 3.14, Noraxon, USA).The exported data were further processed in Matlab (version R2020b, The Math Works, USA).SampEn was computed on the discontinuous segments of the original time series in order to profile variability on a trial-basis for foot acceleration (SampEnAcc), hip flexion (SampEnHF), knee flexion (Sam-pEnKF) and foot rotation (SampEnFR) [59].The onset of passes was determined through minimizing a cost function over possible linear change points in the acceleration data [60].Acceleration signal in the x-axis was rectified and smoothed using a Gaussian-weighted moving average filter at a window length of 20 data points [61].Subsequently, the data points corresponding to an abrupt change in the mean signal were detected as pass onsets using ischange function in Matlab.Based on pass onsets, the time series of foot acceleration, hip and knee flexion, and foot rotation were parsed into segments of 500 data points, which comprised the sequences from pass onset to repositioning of the right foot (2500 ms).MSE was then calculated on the sparse data over 20 scales with the embedding dimension of m = 2 and tolerance margin of r = 0.15 [45], predicated on the work by Costa et.al [43].Although the adopted r threshold varies between 0.1 and 0.5 in the MSE analysis of biological signals [62,63], no substantial differences were found in entropy with varying r values [59].SampEn computation over multiple scales was shown to produce accurate and precise results for a minimum number of 10,000 successive data points in the original series [64] and 100 successive data points at the shortest scale [65].In our study, the original sparse data consisted of 45,000 data points (90 trials x 500 data points), which exceeded the recommended minimum far better following coarse-graining for all 20 scales [42].Additionally, an overall complexity index (ComInd) was calculated as the area under SampEn vs. time scale curve using numerical integration by trapezoidal method.In case of remittent complexity along multiple time scales, a ComInd can provide a measure regarding total amount of complexity [45].

EEG recording and analysis
Cortical activity was recorded using 65 active electrodes (actiCap, Brain Products, Germany) and a wireless amplifier (LiveAmp64, Brain Products, Germany) [66,67].Active electrodes are shown to outperform passive electrodes especially at higher impedance levels [68].The international 10-20 system was accustomed to positioning electrodes, whereby AFz and FCz were set as the ground and reference electrodes, respectively [69].Upon lowering impedance to 25 kΩ by applying high-viscosity electrolyte-gel, the signal was recorded at a sampling rate of 500 Hz using BrainVision Recorder (Brain Products, Germany).A 3D acceleration sensor (Brain Products, Germany) attached posteriorly to the lateral malleolus and connected directly to the amplifier was used to record the acceleration signal of the foot.
The recorded signal was processed offline in Matlab (version R2020b, The Math Works, USA) using the EEGLAB toolbox (version 14.1.2b)[70].The Cleanline plugin [71] was used to remove sinusoidal line noise and the signal was band-pass filtered at the cut-off frequencies of 3 and 30 Hz using a basic finite impulse response filter.The detection and removal of noisy channels was based on deviation criterion, i.e. channels whose robust z-scores of robust standard deviation were higher than 5, were removed (n NOVICE = 0.27 ± 0.59, n EXPERT = 0.67 ± 1.59) [72].The data were subsequently re-referenced to a common average and downsampled to 256 Hz.
The epoching was based on pass onsets detected using the principles of linear computational cost [60].The acceleration signal was rectified and smoothed by means of a Gaussian-weighted moving average filter using a window length of 1000 data points [61].Abrupt changes in the signal were subsequently located using the ischange function in Matlab.The data was epoched to 0-2500 ms from pass onset to focus on post-onset cortical dynamics described in Piskin et al [25].and decomposed into maximally independent components (IC) using an adaptive mixture independent component analysis [73].The sources of the decomposed components were estimated based on a standardized four-shell spherical head model (BESA, Germany) implemented in the DIPFIT plugin [74].Functional brain components were retained based on their source location, activity and residual variance (≤15 %) [75] and clustered using k-means algorithm.To avoid circular inference in the statistical analysis, clustering was based on only dipole locations [76].Three optimization algorithms were computed to approximate the optimal number of clusters [77][78][79].In line with our hypothesis, two clusters assigned to posterior and frontal regions were considered for further analysis in order to focus on previously described cortical dynamics associated with kicking [25].
Following baseline correction for the time window of − 2500 to − 500 ms, the ERSP matrices were retained for the frequency ranges of 3-20 Hz using the study_ersp function in the EEGLAB toolbox.Due to possible contamination caused by muscular activity, frequencies higher than 20 Hz were sidelined [80].This resulted in a 50 ×200 matrix with the x-axis corresponding to time and the y-axis to frequency.For a synchronized analysis of cortical and behavioral dynamics, the time dimension of the ERSP matrix was expanded to 500 pixels via bicubic interpolation using imresize function in Matlab [80].Talairach coordinates were used to identify approximate cortical structures represented by ICs for the interpretation of ERSPs [81,82].

Statistical analyses
Due to lack of studies providing comparable EEG findings, we have calculated a sample size in G*Power [83] based on the behavioral findings reporting differences with a large effect in accuracy performance and complexity between novices and experts [21,84,85].This indicated a minimum number of 14 participants per group for a power of 0.80 at an alpha level of 0.05.
Behavioral endpoints were reported as mean ± standard deviation (SD).Visual inspection of histograms revealed no heteroscedasticity.SampEn of the original time series and their surrogates were compared using an independent t-test to eliminate the possible role of random noise on results [86].Group differences were analyzed by means of independent t-test (ttest2 function).A further analysis based on magnitude-based inferences (MBI) and precision of estimation was also performed for a more detailed inspection of observed differences [42,87,88].A custom-made spreadsheet [89] was used to quantify differences based on standardized thresholds of 0.2, 0.6 and 1.2, for small, moderate and large effects, respectively [90].Direction of the differences (trivial, substantially positive and substantially negative) was determined mechanistically using 90 % confidence intervals (CI), whereby the overlap of CI with the thresholds for the smallest positive and negative effects was deemed trivial.Furthermore, the likelihood of differences was appraised by means of percentage scale, with intervals: 0-0.5 %, most unlikely; 0.5-5 %, very unlikely; 5-25 %, unlikely; 25-75 %, possibly; 75-95 %, likely; 95-99.5 %, very likely and 99.5-100 %, most likely [42,88,90,91].In order to examine the influence of joint intertrial variability on accuracy rate, Pearson's correlation coefficients were computed for accuracy rate and SampEnAcc, SampEnHF, SampEnKF and SampEnFR.
Statistical analyses of the EEG data were computed in Matlab.Group differences in time-frequency analyses were inspected individually for both clusters via statistical parametric mapping [92,93].ERSP values were compared pixel-wise between the two groups to map p-values in a 2D fashion (x-axis: time, y-axis: frequency) using an independent t-test (ttest2 function).Due to simultaneous statistical inferences [94], false discovery rate corrections were applied according to Benjamini and Hochberg [95].All statistical analyses were performed at a significance D. Piskin et al. level of 0.05.

Behavioral endpoints
Perceived level of stress was not significantly different between groups (t(28) = 0.64, p>0.05).The mean values of endpoints, group differences and their qualitative inferences are presented in Table 2.The analysis of accurate and missed passes revealed a significantly higher accuracy rate in experienced players (t(28) = − 5.71, p<0.01) with a very likely (98.9 %) large effect (Fig. 1).There were no significant differences in either SampEnAcc at any scale or in its overall ComInd.MBI suggested a higher trend in experts at initial scales with a possible (60.1 %) small effect.This trend everted as of scale 3 and showed a lower fashion (Fig. 2).SampEnHF was not significantly different between the two groups at scale 1.However, at scales 3-5 and 9-20 differences gained significance and resulted in significantly higher ComInd in novices (t(28) = 2.09, p<0.05).Furthermore, correlational analyses revealed a negative significant correlation between accuracy rate and SampEnHF for all participants (r = − 0.54, p = 0.002).SampEnKF and SampEnFR were not significantly different at scale 1, however MBI suggested higher values in novices with a possible (66.4 %) and a likely (79.5 %) small effect, respectively.Although differences increased at higher scales, they did not gain significance.There were neither significant differences in the ComInd of SampEnKF and SampEnFR nor correlations with accuracy rate.

Cortical clusters
The number of clusters suggested by the optimization algorithms was ).The right parieto-occipital cluster representing 100 % of novices and 93 % of experts and the frontal cluster representing 80 % of novices and 87 % of experts were assigned to ERSP analysis [25].

Event-related spectral perturbations
The pass-related frequency modulations depicted an alpha desynchronization (8-13 Hz) in the right parieto-occipital cluster subsequent to ball contact at approximately 1500 ms in novices (Fig. 3).Experienced players demonstrated a similar desynchronization.However, it was more pronounced and had an onset prior to ball contact, showing significant differences compared to novices, particularly at 1000-1250 and 2000-2250 ms.
In the frontal cluster, a theta synchronization could be observed in both groups which was significantly stronger in football players at ball contact between 1000 and 1250 ms (Fig. 4).There were significant differences between two groups especially at 500 -750 and 1150-1300 ms.

Discussion
The current study explored differences between novices and experienced football players observed in intertrial biomechanical variability across multiple time scales and the associated cortical dynamics while performing short-distance passes.Behavioral analyses elicited higher pass accuracy and overall lower entropy values, thus, reduced intertrial variability for passing biomechanics in experienced football players, which was significant for hip flexion, especially at longer time scales.Intertrial variability for hip flexion also correlated significantly with accuracy in both groups.Further, the comparison of associated cortical dynamics as measured by ERSPs showed group differences in the right parieto-occipital alpha desynchronization and frontal theta synchronization in the successive phases of passes.Our findings may propose an advanced ability to consistently produce a passing behavior yielding higher accuracy in experienced football players, which may be promoted by modifications observed in the pass-related activity of parietooccipital and frontal cortices.

Multiscale entropy profile of pass biomechanics
The analysis of MSE demonstrated significantly lower values for hip flexion in experienced football players which is indicative of less intertrial variability in spatial characteristics of passes.The cognitive frameworks explain learned movement patterns as the development of generalized motor programs with a dyadic structure: a constant part, which can be described as a spatial, task-related pattern, and a variant part, which constitutes dynamic parameters (speed, strength) to be adjusted congruently in a given condition while executing this pattern [96,97].From this perspective, good performance can be described as the ability to reproduce a certain motor behavior consistently in a stable context, as variability may reflect inadequate consolidation of spatial patterns and/or the inability to recruit congruent parameters while executing the movement [36,38,39].In our study, lower behavioral variability observed in experienced players may show their ability to pursue a reinforced spatial pattern and the skilled employment of sensorimotor processes (feedforward and feedback loops, state estimation), which may minimize error and produce higher accuracy in the given conditions.The findings of Button et al. [48], Schorer et al [49].and Fleisig et al [50].are in line with this remark, reporting lower intertrial variability in skilled athletes while performing basketball throws, handball throws and baseball pitches, respectively.
The analysis of spatial variability in different joints revealed significantly lower entropy values at multiple time scales and a lower overall complexity for hip flexion in experienced players (Fig. 2).Several studies highlight the impact of the hip flexion component in a proximalto-distal motion sequence for an efficient pass [55,98,99].As postulated by these studies, experienced players may have an advanced control of hip flexion during the swing phase to generate optimal positive and negative angular velocities consistently, which may explain significant differences found in the current study [100,101].The entropy values portrayed increasing group differences at longer time scales with higher values in novices (Fig. 2).In both groups, short scale dynamics for hip flexion were less variable from trial to trial, whereas coarse movement showed higher variability at longer time scales.Larger differences at longer scales yielded by this monotonic increase indicate even higher predictability of the coarse motor behavior in experienced players, which may posit the influence of higher-level strategies [45].Given that significant differences were found only for SampEnHF, the hip flexion component in a pass sequence may be more distinctive of skilled pass dynamics associated with these strategies.The significant negative correlation found for hip flexion and accuracy rate may also highlight its role as an important marker of advanced pass accuracy.Interestingly, SampEnAcc disclosed an antipodal trend between novice and experienced players as of the third scale (Fig. 2).This may be indicative of accelerative noise detectable at fine scales and being smoothed upon coarse-graining procedure at subsequent scales [102].Experienced football players may tend to kick at higher speeds as a result of accustomed momentum, which may be a learned strategy and is directly linked with higher acceleration of the foot [103,104].However, prioritizing accuracy is known to demand a trade-off between speed and accuracy and result in reduced speeds [105,106].Although this was not the main interest of the current study and therefore not analyzed specifically, the attempts of higher speeds at rare trials may have led to noise observable at fine scales.Upon its removal from the third scale on, the genuine consistency in acceleration may have produced lower entropy values.The non-significant higher trend seen at coarse scales in novices may be due to the changeable trial of different speeds caused by lower accuracy.Overall, based on our behavioral findings, expertise in football may promote the skilled recruitment and sustainment of biomechanical characteristics yielding higher accuracy in short-distance passes, and this may be more evident at coarse motor behavior suggesting the influence of neural processes.

Event-related spectral perturbations
The EEG findings of our study suggest the broad involvement of posterior and frontal cortices in the execution of a short-distance pass, characterized by distinct patterns of alpha desynchronization and theta synchronization, respectively (Figs. 3, 4).Comparable ERSPs were also detected in our previous study [25] and interpreted as reliable cortical dynamics associated with target-directed short passes.Novices and experienced players demonstrated significant group differences in these pass-related patterns, which may reflect distinctive neural strategies developed over time.
In experienced football players, an alpha desynchronization in the right parieto-occipital cluster occurred already in the swing phase, while its onset was subsequent to ball contact in novices.The parieto-occipital areas are mentioned to contribute to visuospatial and attentional networks, to online control of limb movements towards a target and to spatial perception [107][108][109][110]. Especially, the right parieto-occipital areas may predominate in the maintenance of allocated attention to spatial locations [111], which may explain the rightward shift of ICs in this cluster, as observed in our previous findings [25].Given that desynchronization of alpha oscillations is a reliable indicator of cortical activation [112], this pattern may indicate the recruitment of visuospatial processes already in the swing-phase in experienced players, whereas novices depict a delay.Palucci-Vieira et al [24].and Borra et al [113].have reported an alpha desynchronization in the preparation period of a kicking and reaching task respectively, which may reinforce its attribution to the goal-directed visuomotor nature of a pass and propose a skilled strategy of visual anticipation simultaneous with the movement in experienced players [114].
Following ball contact, the desynchronization of the alpha band was more pronounced in experienced players, yielding significant differences at 2000-2250 ms.Based on the aforementioned relevance of parieto-occipital alpha oscillations in visuomotor processes, this difference may expound augmented visual processing in experienced players following ball contact.Simonet et al [115].found an alpha desynchronization in posterior regions following a cricket shot, which was significantly stronger in expert players and attributed to trajectory anticipation.The stronger desynchronization observed in experienced players may similarly suggest a more effective performance tracking based on trajectory anticipation and developed through expertise [116,117].Overall, our findings in the pass-related parieto-occipital alpha perturbations posit that processing visuospatial information during and after ball contact may be a part of skilled strategies to facilitate accuracy through anticipation and performance tracking during and after ball contact in a pass.
The frontal ERSPs were characterized by a theta synchronization, demonstrating differences between novices and experienced players, especially at ball contact (Fig. 4).Fronto-central theta activity pertains to attentional processes originating in the prefrontal and anterior cingulate cortices [118][119][120].A theta increase was also found during other target-directed tasks, such as golf putts, shooting and basketball throws, and linked to focused attention before release [19][20][21].Based on this attribute of frontal theta activity, our findings may highlight increasing attentional demands, particularly at ball contact in a target-directed pass.The higher synchronization observed in experienced players may suggest their improved ability to maximize attention upon releasing the ball, which might consequently enhance motor accuracy.Palucci-Vieira et al [24].have also reported an increase in theta power from approach to ball contact, which was higher for on-target kicks.Indeed, converged attention is given as a cognitive skill promoted by interceptive sports [121] and our findings may acknowledge this skill to be distinctive specifically at ball contact in terms of superior accuracy in target-directed passes.
The findings of the present study collate differences in the ability to sustain a consistent pass behavior from trial to trial and in the associated cortical activity between novices and experienced football players.Higher intertrial variability observed for hip flexion in novices may stand for an explorative behavior [46], whereas expertise in football may bring in the skill to pursue spatial characteristics consistently, which may reduce noise in a stable condition [36,38,39].Larger differences observed at longer time scales may propose the influence of higher-level strategies on this skill [45].Differences observed in parieto-occipital alpha and frontal theta perturbations may affirm distinctive visuospatial and attentional processes developed as sport-specific strategies.To the best of our knowledge, our study provides preliminary evidence about expertise-related behavioral and cortical dynamics which may contribute to superior pass accuracy.Our findings may provide a basis for the prospective investigation of pass accuracy in enriched contexts, for instance, with different target distances and in an unpredicted environment.Understanding expertise-related modifications not only in a stable, but also in a dynamic environment may provide a physiological measure for intervention studies, aiming to promote pass accuracy in different football-specific situations.

Methodological limitations and outlook
The methodological limitations of the current study should be attended to while interpreting and translating its findings into practice.Starting with the sample, both cohorts consisted of female and male participants which may introduce a gender-based bias in the findings.Performance in football including the accuracy aspect may differ between female and male players and prospective studies should consider exploring gender-based differences [122,123].Furthermore, short-distance passes may provide a preliminary insight into how expertise may modulate behavioral and cortical dynamics associated with pass accuracy.However, in real situations long-distance passes are also frequently used [124], which calls on components such as speed and strength [125].Furthermore, in the real world, passes and kicks are executed as a product of rapid decision-making mechanisms in an unpredictable environment [126], which was ignored in our setting.The simplicity of the task could also explain non-significant differences observed for foot acceleration, knee flexion and foot rotation.Exploring if and how a longer pass distance and an unpredictable environment may influence variability in pass biomechanics and the associated cortical dynamics may help to understand more complex, situation-specific strategies and utilize them to improve accuracy and enrich the skill repertoire of football players.Our study focused on intertrial variability to investigate expertise-specific capability to recruit and pursue biomechanics yielding higher accuracy in a stable setting, however, exploring intratrial variability in the execution of a pass could also provide important insights about motor acuity, which may facilitate an efficient ball impact [127,128].
The known inherent limitations of EEG such as approximate source localization [129] should underline the imprecision in the pronounced cortical structures.Furthermore, EEG analyses were restricted to parieto-occipital and frontal clusters.Although the activity of these cortices may reflect the predominant demands of a target-directed pass [24,25], other regions may also provide knowledge about important cortical processes involved in the execution of a pass.A higher sample size may enable a higher proportion of contribution in the ignored clusters.Finally, the statistical challenge due to unequal distribution of ICs per participant should also be mentioned as a known, but still unsolved problem in EEG analysis.

Conclusion
The aim of the current study was to explore differences in intertrial behavioral variability and cortical dynamics associated with shortdistance passes between novices and experienced football players.Our results revealed lower entropy values for hip flexion, proposing a reinforced passing behavior in experienced players.The EEG data showed a more pronounced parieto-occipital alpha desynchronization and a frontal theta synchronization in successive phases of a pass, which may show the influence of visuospatial and attentional strategies on motor consistency associated with higher accuracy.Our findings extend the knowledge regarding the physiological underpinnings of superior pass accuracy in football and provide a basis for its prospective investigation in enriched contexts.

Fig. 2 .
Fig. 2. Comparison of entropy values at multiple scales (left column) and the complexity index (right column) for the acceleration of the foot and the motion ranges of hip flexion, knee flexion and foot external rotation during short-distance passes performed by novices and experienced football players.

Fig. 3 .
Fig. 3.The scalp map (right and the dipole locations (right bottom) of the right parieto-occipital cluster.The alpha desynchronization commences earlier and occurs more pronounced following ball contact in experienced football players (left second), compared to novices (left top), yielding significant differences in the statistical map (left third).The synchronized acceleration data exhibits the approximated ball contact time (left bottom).

Fig. 4 .
Fig. 4. The scalp map (right top) and the dipole locations (right bottom) of the frontal cluster.The novices show a weaker theta synchronization at ball contact (left top), whereas it is more pronounced in experienced football players (left second).The statistical map (left third) shows significant differences for this synchronization.The synchronized acceleration data exhibits the approximated ball contact time (left bottom).

Table 1
Demographics of the participants reported in mean ± standard deviation.

Table 2
Behavioral endpoints for novices and experienced players with their qualitative inferences.