Features that set apart the sports experts from the non-experts have been a topic of considerable research efforts during the last years. In this regard, it has been proposed that physical (e.g., anthropometric), physiological, psychological, sociological, technical, and tactical factors discriminate between athletes of different levels of expertise (Burgess and Naughton, 2010; Sarmento et al., 2018; Piggott et al., 2020). Most emphasis in research has been placed on studying physical aspects and biomotor abilities (such as speed, strength, power, agility, and endurance), although recent studies suggest that the predictive validity of these factors regarding performance and talent identification is limited, especially when they are looked at in isolation (Wagner et al., 2016, Wagner et al., 2019, 2020; Bennett et al., 2018; Bergkamp et al., 2019). This might be due to the fact that these particular predictors are not representative enough with respect to the contextual constraints of on-field behavior (Bergkamp et al., 2019). To overcome this problem, multidisciplinary approaches have been suggested to diagnose future performance and talent. Particularly, the need to focus more on the psychological and perceptual-cognitive components of athletes has been highlighted in recent studies (Bangsbo, 2015; Woods et al., 2016; Bennett et al., 2018; Bergkamp et al., 2019; Sherwood et al., 2019; Piggott et al., 2020). Indeed, mounting evidence suggests that perceptual-cognitive skills such as decision making constitute a performance-discriminating component in team-sports (Mann et al., 2007; Travassos et al., 2013; Silva et al., 2020; Ashford et al., 2021).

It is against this background that the development and evaluation of tests to diagnose perceptual-cognitive skills and expertise in athletes has gained more and more attention in the last years. A commonly used approach to study sport-specific decision making is the so called temporal occlusion paradigm (Jones and Miles, 1978). Essentially, studying decision making with this approach involves presenting video clips of selected game sequences on a screen, and subjects watch these clips while usually being in a sitting or standing position (Travassos et al., 2013 for details). After the end of the video sequences, subjects are mostly required to verbalize their intended response for the game situation in question, or to verbalize their generated options (Johnson and Raab, 2003; Raab and Johnson, 2007; Raab and Laborde, 2011). Decision making in real life evolves from a complex and uncertain context (especially in team-sports), requiring athletes to constantly process information while acting under time and information constraints (Travassos et al., 2012; Kinrade et al., 2015). Against this background, it appears problematic to not consider the specific environment in which the players actually perform, and particularly to neglect the impact of sensorimotor interactions in decision making (Burk et al., 2014; Raab, 2014). Recent examinations in netball (Bruce et al., 2012), and soccer (van Maarseveen et al., 2018a) clearly suggest that the decision-making performance in perceptual-cognitive tests (using verbal reports, button press, or micro-movements) differs from actual real-world decision-making performances, thus hampering the ecological validity of findings (Araújo et al., 2007; Ashford et al., 2021). Taken together, these studies highlight that assessing verbal or micro-movement responses (Travassos et al., 2013) might be not sufficient to predict on-field performance, let alone to detect talents. Notwithstanding, there is also evidence in perceptual-cognitive research which showed that uncoupled perception-action responses, given either verbally or by keystroke, are similar (Farrow et al., 2005) or even more accurate (Ranganathan and Carlton, 2007) than coupled perception-action responses, requiring sport-specific action responses. It was also found that when observers are static in computer-based experiments, the motor regions of the brain are still linked to the perceptual information picked up (Aglioti et al., 2008). In team-handball (Huesmann et al., 2021), compared the anticipation performances of advanced, intermediate, and novice team-handball goalkeepers in a perception-action artificial (verbal responses) and simulated (motor responses) condition with. The authors revealed overall superior performances (higher prediction accuracies) in the artificial, verbal response condition, outlining that the evidence regarding the necessity of the involvement of motor components seems mixed. However, when capturing expert performance in decision making, expertise effects are most pronounced when the participants actually performed actions under in situ task constraints (Travassos et al., 2013) in realistic test paradigms under field conditions (Mann et al., 2007).

Since the discrepancy between decision making in decoupled vs. coupled processes of perception and action in task designs has only recently become known, there are only a few studies assessing decision making in near-game test conditions with requirements to perform a sport action (Travassos et al., 2013). One notable example is the study by Magnaguagno and Hossner (2020) in team-handball, which investigated decision making in a performer environment by using virtual reality. Specifically (Magnaguagno and Hossner, 2020), presented 1 vs. 1 video sequences between a defending teammate and a left or right back attacker, respectively, ending in either a successful or a lost defense action of the teammate. The participants were put into an assisting role as a defender next to the 1 vs. 1 situation. Depending on the teammates’ either weak or strong defending behavior, participants had to decide (based on their anticipatory performance) whether to move sideways for tackling the attacker (in case the weak teammate has lost his duel), or whether to stay in the passive position (in case the strong teammate successfully defended the attacker). The authors found expertise effects in response correctness, showing that expert players responded more appropriately on a lost 1 vs. 1 duel of the respective defending teammate than the near-expert players. However, response correctness simply based on the decision whether to stay passive or to tackle. Further options for defense actions, e.g., provoked by additional varying attack sequences, were not regarded. Also, the respective time of responses (decision time) was not recorded, even though decision time is thought to be an important metric of decision making (Vaeyens et al., 2007; Raab and Laborde, 2011; Ratcliff et al., 2016; Seale-Carlisle et al., 2019).

In the present study, we investigated whether decision-making in a near-game performer environment would differ between expert and near-expert athletes. To this end, we used a team-handball specific sensorimotor decision-making task with varying attack sequences based on the temporal occlusion approach (see Jones and Miles, 1978 for details). We have previously shown that this test setup is sufficiently reliable to study decision making (Hinz et al., 2021), and by comparing experts vs. near-experts we now undertake the next step to discover the potential usefulness of this test to distinguish expert vs. non-expert performance. This test setup involves both domain-specific motor responses (as compared to, for example, button press) and the opportunity to choose among various response options (as compared to “either-or decision making”). We also recorded decision times which allowed us to study whether expert and near-expert players would initiate a different defense action (e.g., a “proactive” behavior like tackling vs. a “passive” behavior like blocking) in response to identical visual information, and whether there are differences in the accompanying decision times. In order to use representative task constraints (Travassos et al., 2013), we also investigated the decision-making performances in right- as well as left-hander attack sequences, due to handedness advantages in favor of left-handers in sport (Hagemann, 2009; Loffing et al., 2015).

Carrying out between-group comparisons with multiple choices for responses entails difficulties in estimating a priori effect sizes. Therefore, our sample size recruitment complied with sub-sample sizes from earlier between group investigations in this field (e.g., Raab and Johnson, 2007; Zoudji et al., 2010; Bruce et al., 2012).

The sample of participants consisted of 44 male team-handball players (M_age = 19.11 years; SD = 6.56 years). Two teams (n = 22; M_age = 17.59 years, SD = 3.67 years) were recruited of a professional youth academy of a first league team-handball club of the German Handball Federation. All players competed in the highest possible league within their age category. Players of these teams performed a minimum of 14 h training per week with one competition match at weekends. All athletes practiced team-handball for at least 8 years. Based on the recommendations of Swann et al. (2015) how to classify expertise level in sports science, players of the two teams can be considered as elite level players. The players of the other two teams (n = 22; M_age = 20.71, SD = 8.54) were recruited from non-professional, local league teams within their age categories. All athletes performed 4 h of training per week with one competition match at weekends, and players practiced team-handball between 2 and 22 years. According to the definition of Room (2010), who defines a player, “who takes part in sport for pleasure, as distinct from a paid professional” as amateur player, the athletes of these two teams can be considered as amateur level players. Differences in age between both groups were not significant (p = 0.952).

The experiment was conducted during the first half of the championship season 2020/21, in October and November. At that time, all teams had a normal weekly training and match schedule, without being affected or restricted by any local or federal COVID-19 regulations. During the test, participants were instructed to perform with a maximum effort. Injuries lead to exclusion of the study. Prior to their participation, all participants and legal guardians were informed about the purpose, risks, and benefits of the study. All participants had to give a written informed consent before the first test day. Participants were not identifiable from the test results. The study protocol was approved by the local ethics committee from the Otto von Guericke University Magdeburg and met the requirements of the Declaration of Helsinki and its later amendments.

All tests were conducted on the contact plate system SpeedCourt^® (Q12 PRO mobile, GlobalSpeed, Hemsbach, Germany) which enables sport-specific motor responses to temporally occluded videos. As a basis for profound interpretations of envisaged results, we used the test setup from a previous study (Hinz et al., 2021), which was introduced and checked for basic psychometric properties (reliability), using four team-handball specific attack actions for intra- (cross-sectional) and inter-session (longitudinal) agreement of the motor response choices and times. Significant Cohen’s (0.44–0.54) and Fleiss’ (0.33–0.51) kappa statistics (Fleiss and Cohen, 1973) revealed moderate agreement level of motor responses. Please refer to this paper for detailed explanations regarding test construction and item analyses. In the study at hand, we used the identical test setup and video stimuli, along with the same mapping of the four choice responses (forward/tackling response; sideways left or right movements; blocking/passive behavior) to the contact plates on the SpeedCourt^®.

The experimental scenario consists of Breakthrough, Standing throw, Jump throw, and Pass videos, which were temporally occluded within a general time frame of ball was passed to attacker (t₆) and obvious end of attack (t₀), with time intervals of 200 ms (t₆ = −1200 ms, t₅ = −1000 ms, t₄ = −800 ms, t₃ = −600 ms, t₂ = −400 ms, t₁ = −200 ms, t₀ = 0 ms). The duration of each video clip was not longer than 2 s (stopping at t₀). Dummy trial videos, showing too ambiguous attack actions for an appropriate defense response, were included in the test scenario, aiming at avoiding expectation effects in response behavior (Anderson, 1983). Due to handedness advantages in favor of left-handers in sport (Hagemann, 2009; Loffing et al., 2015), all video clips were mirrored.

The videos were sized 1280 × 720 pixels (width × height) and the test scenario was implemented by using Lazarus (Version 2.0.10) software. In total, 112 right- and left-handed attack video clips were presented to the subjects during the measurement procedures.

The test procedure always started on the marked 7 m line on the central contact plate of the SpeedCourt^®. In this starting position, a 3 s countdown appeared on the screen, followed by a video stimulus showing an attack action. The aim was to respond as intuitive and as realistic as possible after the video presentation ended. Subjects were then returned to the starting position to prepare for the next countdown. Subjects were instructed that the motor response to a video stimulus should replicate their first intuition for a defense response that came to their mind while watching the video. Too early or unintended given responses were marked for later exclusion. No information about decision performance or the remaining number of videos was provided to the subjects.

In relation to the Bayesian integration framework (Vilares and Kording, 2011; Gredin et al., 2020), all subjects received the same team-handball specific instructions (stable priors) about the attacker’s action tendencies, the defense tactics, and the match status. Stable context priors via action tendencies were provided, meaning that the center back player in the video can be considered as an allrounder or playmaker, being able to put high pressure on the defense through a variety of long and near range standing and jump throws, strong one-on-one actions, and high-quality passing. Tactical priors were also supplied to the subjects. More specifically, they were instructed to put themselves in the position of the central block defender in a classic man-to-man defense without teammates, or other opponents than the attacker in the video/situation. The match status was pre-specified as the 50th minute of play (of 60 min in total) and the game score was tied.

Following the instructions, subjects performed 10 familiarization trials, showing a selection of occluded attack actions in randomized order. After familiarization, the test started with a block of right-hander video stimuli followed by a left-hander block, interspersed by a short break of approximately 5 min. Within each block, the videos were presented in quasi-randomized order, starting at occlusion t₆ (fewest information) and ending at t₀ (full information) videos. The test duration was approximately 35 min.

Significant between-group differences of the response frequency distributions are shown in Figure 1. Significant effects of expertise were present in the left-handed Breakthrough and Pass only. Full illustrations of response frequency and distribution in all attacks, as well as individual Chi-square statistics, are provided in Supplementary Figures 1, 2 and Supplementary Tables 1, 2.

Irrespective of group, visual inspection of frequency distributions shows dynamically changing response patterns over occlusion points, most likely due to the varying amount of visual information provided about the attacker’s action. As indicated by significant results of Fisher’s combination, the elite players responded significantly more often with forward/tackling movements in Breakthrough [χ²(14) = 25.06, p = 0.033, OR = 2.76, CI = 1.54–4.95] and Pass [χ²(14) = 37.19, p = 0.001, OR = 3.00, CI = 1.78–5.04]. On the contrary, amateur players instead showed a more frequent use of passive/blocking in Pass [χ²(14) = 28.28, p = 0.013, OR = 2.70, CI = 1.64–4.46] (see Supplementary Table 2). Of note, especially regarding single occlusion points with comparably few visual information, elite players use more frequently forward/tackling in Breakthrough [t₃: χ²(1) = 5.20, p = 0.023, φ = −0.36]. In Jump throw, elite players use more frequently forward/tackling in t₄ [χ2(1) = 4.64, p = 0.031, φ = −0.36], but switching to more frequent passive/blocking responses at t₃ [χ2(1) = 4.50, p = 0.034, φ = 0.35]. Another significant between-group difference was observed in Jump throw, where amateur players responded more often with a sideways right move [t₄: χ2(1) = 3.90, p = 0.048, φ = 0.33] (see Supplementary Table 2).

As expected, based on our previous study (Hinz et al., 2021), faster decision times occurred with increasing visual and kinematic information of the attacker in both groups (Figure 2). This is evidenced by significant results for the main effect “occlusion” in the repeated measures ANOVA in right-handed [Breakthrough: F(6, 90) = 4.42, p < 0.001; Jump throw: F(6, 78) = 10.34, p < 0.001; Standing throw: F(6, 96) = 9.52, p < 0.001; Pass: F(6, 96) = 6.51, p < 0.001] and left-handed [Breakthrough: F(6, 84) = 27.48, p < 0.001; Jump throw: F(6, 96) = 32.51, p < 0.001; Standing throw: F(6, 120) = 18.67, p < 0.001; Pass: F(6, 114) = 17.29, p < 0.001] actions. Between-group comparisons of the main effect “expertise,” however, revealed no significant effects. A significant group-by-occlusion [F(6, 96) = 2.33, p < 0.038] interaction was detected in the right-handed Standing throw. A detailed overview of ANOVA statistics is presented in Supplementary Table 3.

Considering the need of motor responses in expert decision-making research (Travassos et al., 2013), the current study compared the decision-making behavior between elite and amateur team-handball players, by using sport-specific motor responses in representative near-game test situations. To do so, we measured the frequencies of selected responses, which were given as a team-handball specific defense action on occluded video sequences showing varying attack actions. Additionally, we were interested in the decision time each player required to initialize the respective response selected.

Regarding response frequency, we identified significant effects of expertise in the left-handed Breakthrough and Pass, where elite players demonstrated an overall significant preference to respond with forward/tackling movements on both attack actions. The amateur players, however, preferred to stay rather passive or blocking in Pass. More specifically, preferences for forward/tackling response or passive/blocking responses by the elite players in single occluded time points were also found (at t₄ in left-handed Breakthrough and left-handed Jump throw, and t₃ in left-handed Jump throw). Interestingly, the amateur players demonstrated significant more frequent sideways right responses in the left-handed Jump throw attack. Taken together, the differing frequencies of selected responses from both player groups suggest an expertise-dependent decision-making behavior. Expert effects in our study align with previous motor experiments in decision-making research in team-handball (Raab, 2003; Magnaguagno and Hossner, 2020), and extend them by new insights into how elite and amateur players differ in their tactical understanding of defending when limited visual information is provided.

Regarding decision time, we detected a reciprocal decrease of the time for a decision with increasing kinematic information in the presented attack actions (as found in Hinz et al., 2021). Despite the overall significance of this effect in all of the attacks, an expertise effect in decision time did not appear at all. To the best of our best knowledge, comparable decision (or response) time data from related motor experiments in team-handball using near-game environments do not exist (Bonnet et al., 2020). Previous non-motor experiments (Raab and Laborde, 2011) using offense sequences found expert players to decide better and faster, however, a similar study investigating the influence on mood on decision making found no expertise effect in decision time (Laborde and Raab, 2013). Likewise, the decision time data in our study was also not able to discriminate between groups. The comparability to both of the mentioned studies is to be seen with care, due to the specificity of the offense or defense situations the players were tested in, and due to the methodological aspect of sensorimotor responses in our test instead of verbal reports as responses.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the Local Ethics Committee from the Otto von Guericke University Magdeburg. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.

MT, KM, and MH: conceptualization. NL, MT, and MH: methodology and validation. MH: software, investigation, resources, data curation, and writing – original draft preparation. NL and MH: formal analysis and visualization. MT, HW, NA, JT-C, and NL: writing – review and editing. MT and HW: supervision. MT and KM: project administration and funding acquisition. All authors have read and agreed to the published version of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.