Sport Science is a rapidly growing field that applies scientific methods to sport settings to improve performance and athlete wellbeing. At the center of this profession is the sport scientist, who is tasked with duties related to performance enhancement programming, testing and profiling, and the monitoring of training load and injury trends (1). With the rapid expansion of technology is sports (2) it is now more important than ever for sport scientists to take ownership of their data by ensuring that the technologies and collection/analysis methods have been critically evaluated (1). Currell and Jeukendrup proposed three factors that contribute to a good performance test: (i) validity; (ii) reliability; and (iii) sensitivity (3). Reliability refers to the reproducibility of the values of a test (4) and can be influenced by variation in performance by the test subject, variation in the test methods, and variation in measurement of testing equipment (5). In a sports scientist's desire to adopt new technologies and assessment methods, the first step of this process should be an evaluation of reliability.

While sports scientists have been quantifying athletes speed (horizontal acceleration and maximal velocity) for quite some time, assessment of horizontal deceleration performance has only recently become more prevalent (6–12). Given the importance of deceleration to the demands of numerous sports, quantifying athletes' ability to reduce velocity and whole-body momentum may have significant implications for health and performance (10). Research has shown that in some field-based sports such as soccer and rugby athletes perform more high-intensity decelerations compared to accelerations, and that these high-intensity decelerative actions may have a profound impact on competition-related muscle damage, fatigue and recovery (7, 13–15). Both American football and lacrosse are multi-directional sports in which athletes are frequently exposed to high intensity deceleration demands in order to generate separation from defenders, perform cutting motions or complete rapid changes in direction (16–20). During maximal horizontal decelerations, the braking steps exhibit a distinct ground reaction force profile characterized by high-impact peak forces and loading rates, in some cases up to six times body mass, which is nearly three times higher than maximal horizontal acceleration (6). These forces and loading rates must be met with appropriate neuromuscular (e.g., braking force attenuation), coordinative, and skill-related qualities to efficiently reduce whole body momentum (6). With potential implications for performance (e.g., generation of space between players, change of direction ability), and health (e.g., injury risk reduction) (10), it is important for sport science practitioners to possess the means to quantify deceleration ability in a reliable and valid fashion.

According to Harper et al., prior to 2020 only a small number of studies have tried to quantify and highlight the importance of horizontal deceleration performance (9, 21–24), likely due to a lack of methods, tests, and technologies to effectively quantify horizontal deceleration. Radar and laser devices have historically been used to assess athlete's horizontal sprint acceleration abilities (25), and more recent research has used this same technology to measure horizontal deceleration. Harper et al. proposed the use of radar-derived instantaneous velocity to calculate different metrics of interest during a novel acceleration to deceleration task, termed the acceleration to deceleration ability assessment (ADA) (8). This test requires athletes to maximally accelerate over 20 m, followed by a rapid and maximal deceleration, and backpedal to the 20-m marker (8). Based on the dimensions of the playing field, recent research has modified this protocol to include a 10-m acceleration instead of a 20-m acceleration (11, 12). While different studies have reported reliability and variability scores in their results, to the authors knowledge, only one study has investigated the inter-, and intra-day reliability of several kinetic and kinematic deceleration metrics derived from a radar device (8). In this study the majority of ADA-derived metrics showed good intra-day reliability and were sufficiently sensitive to detect small-to-moderate worthwhile changes in deceleration performance. Further, only kinetic variables had good inter-day reliability, and were adequately able to detect moderate worthwhile changes in deceleration performance after a single familiarization session. The proposed ADA methods contain two steps that require manual processing of raw data; (i) manually deleting all data recorded before the start of the sprint and following the end of the deceleration phase, and (ii) manually removing unexpected high and low data points on the velocity-time curve that were likely caused by segmental movements (e.g., arms) of the participants while running (8). These steps which contain no clear thresholds for eliminating data points may leave room for methodological variability. Previous studies reporting the reliability of ADA assessments used a single rater to manually process raw data (26), and the introduction of multiple individuals to process the raw data may add methodological variability making it difficult to compare results between studies, different populations or athlete groups, as well as comparing results over time within the same athletes.

Therefore, the primary aim of the study was to investigate the between-rater agreement of four raters for processing and quantifying ADA-derived acceleration and deceleration metrics using radar technology. The secondary aim was to investigate the intra-rater reliability and variability (i.e., within-session) between the two trials each athlete performed for all metrics of interest, and across all raters. Authors speculated possible disagreement between raters based on the subjective nature of the initial data treatment process, which could negatively impact the usage and interpretation of results.

Sport-specific descriptive information presented as means and standard deviations across all raters may be found in Table 2. Inter-rater reliability statistics may be found in Table 3. The inter-rater analysis revealed ICCs ranging from moderate to excellent agreement across all ADA-derived metrics, with CV%'s ranging from 3.5 to 7.8. F-values ranged from 6.07 to 7.8 In line with the secondary aim of the study, Table 4 presents intra-rater (i.e., within-session) reliability and variability statistics across all four raters. Intra-rater ICCs ranged from poor to excellent consistency, while CV%'s ranged from 3.1 to 13.2. F-values ranged from 1.87 to 22.0. SEM values displaying inter-, and intra-rater variability in the same units as the variable of interest may be found in Tables 3, 4, respectively. Figure 1 displays raincloud plots showing boxplots, half-violin plots, and data jitter to visualize agreement between raters for selected deceleration metrics, while Figure 2 visualizes the inter-rater ICCs plus respective confidence intervals.

The primary aim of this study was to investigate the between-rater reliability/agreement for quantifying maximal horizontal deceleration performance in collegiate NCAA Division 1 athletes using radar technology. Four individual raters manually processed raw velocity-time data from an assessment quantifying maximal acceleration and deceleration using suggestions from previous research reports (8). An automated script was used to calculate metrics of interest, which were based on the manually processed velocity-time data. A secondary study aim was to determine the intra-rater (i.e., within-session) reliability and variability of all ADA-derived metrics.

The inter-rater reliability analyses reported ICC's ranging from moderate to excellent agreement, with TTS, DEC Dist, and Avg DEC_E presenting moderate levels of agreement. However, some caution may be advised with regards to TTS and DEC Dist as in other studies, these metrics have been found to produce poor to moderate levels of inter-day reliability (8) and elevated within-session variability scores (12, 21). Additionally, only showing moderate levels of agreement, Avg DEC_E presented with the greatest inter-day CV%, suggesting that between raters up to 7.8% of variability may be seen due to differences in the manual data processing procedures. Avg DEC_E being the average change in velocity between Max Velo and 50% of Max Velo, it is possible that the lack of clear guidelines or thresholds for removal of outliers around the top of the velocity-time curve during manual data processing may be responsible for the reported variability. It seems that this location on the velocity-time curve represents a window in which individual interpretations may negatively influence the consistency between raters. Readers may refer to Figure 1 which presents a visualization of an unprocessed and processed velocity-time curve in which the removal of a single outlier affects the point of Max Velo used to determine Avg DEC_E. More specifically, this figure was modified to only show the acceleration and the deceleration phase, and not the phase of the backpedal. With this in mind, defining the start of the deceleration phase as the first instantaneous velocity point following the athletes’ maximal approach velocity may fail to take into account the athletes' “coasting” or “transitional” phase prior to increasing their rate of deceleration. Future research may explore the utility of other start of deceleration phase detection thresholds. While speculative, jerk being the derivative of acceleration/deceleration with respect to time may offer additional insights into the identification of different kinematic events and phases, similar to research looking at the yank-time signal in vertical jumps (31).

Furthermore, assessing the intra-rater (i.e., within-session, test-retest) reliability and variability analyses, across all four raters, Avg DEC_L showed poor levels of consistency between trials. This suggests the later parts of the deceleration phase may be prone to biological and performance related variability, while the early deceleration phase may be more impacted by rater-related differences in the manual processing procedures. TTS and DEC Dist also presented with ICCs below 0.50, suggesting poor consistency between the two ADA test trials. Given that they reflect the time and distance when decelerating to a stop, the previous two metrics may carry ecological validity regarding the communication with coaches and non-sport science practitioners. However, their questionable intra-rater reliability and variability shown across different studies raises concerns with regards to their utility with athlete populations. While speculative, the low degrees of reliability with regards to DEC Dist may be influenced by the absence of a predetermined stopping point for the deceleration phase. In the ADA, athletes are instructed to rapidly decelerate at a specific location, however the end of the deceleration phase is not provided. Furthermore, while instructed to initiate the deceleration phase as close to the 9.14-m marker as possible, athletes may initiate the deceleration phase prior to or after the 9.14-m marker. To avoid such pacing strategies, previous research has implemented possible solutions, such as comparing athlete's approach sprint times during the ADA to their times during a maximal linear sprint test over the same distance and eliminating trials if the difference between the two is above or below a selected threshold (8). However, in applied settings with large groups of athletes being tested at the same time, these solutions may be challenging to adopt. When looking at the sprint acceleration phase of the ADA (i.e., Avg ACC, Max ACC, Max Velo), Avg ACC presented with excellent levels of consistency across all four raters, while Max Velo showed moderate levels of consistency, and Max ACC showed poor to moderate levels of consistency. In both the inter-rater and intra-rater analyses, Avg ACC, Avg DEC showed greater ICCs compared to Max ACC, Max DEC. This has been shown in previous research suggesting better reliability and sensitivity with taking the absolute value of all raw acceleration and deceleration values and averaging them over the duration of a selected time period (32, 33). Limitations with calculating maximal vs. average metrics should be taken into consideration when interpreting these data.

Additionally, readers should appreciate the differences in the types of variability presented in in this study. Intra-rater variability may be thought of as biological variability influenced by intra-athlete differences in performance between trials, while inter-rater variability may be thought of as technological variability, influenced by the between-rater differences related to the manual and somewhat arbitrary nature of the data processing procedures. While some biological variance is to be expected, technological variability should not exceed the variability induced by biological factors. Figure 3 presents a comparison for metric-specific SEM values, comparing biological and technological variability in our data. Figure 3 shows that biological variability exceeds technological variability for all metrics except for Avg ACC, which while marginally different, showed greater technological variability. Similar to earlier discussions about different locations along the velocity-time curve, the start of each ADA trial may also present a window for individual interpretation of about outlier removal and initiation of the start of the trial. While speculative, our data suggests that manual processing procedures at both the start of each trial, as well as around the top of the velocity-time curve (i.e., start of deceleration) show the largest windows for individual rater interpretation, which could induce unnecessary variability into ADA-derived acceleration and deceleration metrics. Ultimately, this process should become fully automated, based on selected filters or algorithms, rather than the subjectivity of the rater.

To the authors' best knowledge, this is the first study attempting to assess the between-rater agreement for processing and assessing maximal horizontal deceleration performance using radar technology in athletes. Furthermore, authors believe to date, this is the largest sample of NCAA division 1 athletes performing the maximal horizontal deceleration assessment used in this study. However, limitations with the ADA and future avenues of research should still be acknowledged. In our study descriptive statistics for both American football, and Lacrosse athletes suggested that most athletes required between 4.5 and 5 m to maximally decelerate. This is in line with findings by Graham et al. who investigated athletes' deceleration ability in relation to their self-determined limit to accelerate over different prescribed distances (9). In this study, it took athletes 4.94 ± 0.39 m to decelerate to a stop, following a 10-m acceleration during which they achieved 72.2 ± 3.2% of their maximal speed (30-m maximal sprint). It could be worthwhile for sport-science practitioners to further look into deceleration qualities during tasks in which a distinct stopping or turning point is identified. Further, athletes requiring between 4.5 and 5 m to maximally decelerate following a 10-m sprint seems to match up well with the dimensions of the 5-0-5 change of direction deficit test, which could give a more holistic insight into not only horizontal deceleration ability, but also the ability to efficiently turn and re-accelerate. Using radar-, or laser-technology to quantify deceleration and re-acceleration qualities in a task such as the 5-0-5, similar to more recent research (12) may productively add to the existing body of evidence with regards to adding context to deceleration demands and braking strategies during change of direction maneuvers.

Likely the overarching strength of this study is the size and trained nature of the population investigated, as well as the fact that both males and females were studied. The breadth and depth of descriptive information provided in this study may also allow sport science practitioners to use results in comparing or benchmarking athletes. Authors believe this to be the largest sample of trained athletes, both male and female performing the novel, maximal horizontal deceleration task used in this study. Otherwise, limitations pertaining to this study may be identified in the applied nature of the investigation. When working with high-level collegiate athletes, researchers often struggle to control for outside variables such as sleep, nutrition, as well as hydration, amongst others. Future studies may aim to replicate methodologies implementing more rigor with regards to control for outside factors potentially affecting athlete performance. Additionally, future research investigating the intra-rater reliability of deceleration measures may decide to include more than two test trials for each athlete, and further investigate the variation in selected metrics between test-days. Furthermore, future studies may replicate procedures across additional populations, and deceleration assessments, allowing for greater generalization of our findings. Regardless, given the substantial sample size of high-level male and female athletes in studying the reliability of an emerging physical construct of horizontal deceleration performance, the authors believe that this study effectively contributes to a growing body of literature.

This study documented moderate to excellent levels of agreement between four individual raters in quantifying maximal horizontal deceleration performance within the ADA test. Based on our data, it may be speculated that if a foundational understanding and agreement of manual data processing procedures for radar-derived data is given between raters, metrics may be interpreted with moderate to excellent levels of confidence. However, when possible, and when using the Stalker ATS radar technology, authors advise practitioners to use one trained individual to manually process raw data. Ideally, this process should become fully automated, based on selected filters or algorithms, rather than the subjectivity of the rater. Particular caution in the manual data cleaning process may be used for Avg ACC and Avg DEC_E, as based on our data, these metrics present with more variability between raters. This is likely influenced by the processing procedures around the start of the trial and the top of the velocity-time curve (i.e., start of deceleration), which seem to leave a greater window for individual interpretation. The intra-rater analyses revealed poor to excellent consistency between ADA trials, with CV%'s ranging from 3.1% to 13.2%, depending on the respective metric and rater. Caution may be advised with regards to the intra-rater reliability of TTS and DEC Dist across different ADA trials. Findings presented in this study may be of acute interest to sport science practitioners working with multi-directional sport athletes and interested in reliably quantifying maximal horizontal deceleration performance.