Weightlifting requires high levels of muscular strength and barbell handling skills to lift maximal barbell loads in the snatch and the clean and jerk (1, 2). Therefore, a major goal of weightlifting training is to improve the necessary physical abilities using competition lifts (i.e., snatch, clean and jerk) and specific assistant exercises (i.e., power snatch, hang clean, snatch pull) (3). During a lift, the barbell and weightlifter form the barbell-lifter system (4). Based on the barbell lifter system, the analysis of barbell kinematics can be used to assess the lifter’s technical mastery and physical performance measures (i.e., barbell power output) (5). In this context, the barbell trajectory, the vertical barbell velocity, and vertical acceleration are among the most important kinematic parameters (1, 6, 7). Determining the effect of training on changes in barbell kinematics of the individual weightlifter is of major interest for coaches during training and competition. To distinguish a “real” training effect from potential “noise” (i.e., variability) during lifts, the typical magnitude of variability of the lifts needs to be known. For this purpose, an intra-session test-retest experiment can be used to estimate the variability of barbell kinematics in weightlifting exercises (8). In a complex movement task like weightlifting, externally measured variability during repeated lifts is the sum (i.e., total variability) of the variability of motor performance itself, the error of the used measurement system (e.g., video recordings, accelerometers) and the variability related to the used test-condition (9). The typical magnitude of the total variability of barbell kinematics during specific exercises (e.g., snatch, clean and jerk) can be used as a reference point to determine “real” intra-individual changes of barbell kinematics in training and competition (10, 11).

The barbell kinematics in weightlifting are most frequently analyzed based on two-dimensional video recordings (12). As recently pointed out by Nagao und Yamashita (13), the video-based marker-less automatic two-dimensional analysis of barbell distance and velocity highly agrees with data from a three-dimensional infrared motion capture system [intraclass correlation coefficient (ICC): 0.971–0.999; systematic bias: −0.001–0.001 m, −0.034–−0.005 m·s⁻¹]. In addition, it has been reported that a video-based automatic marker-less two-dimensional barbell tracking software has a high test-retest reliability for barbell distance, velocity, and acceleration if the same videos where analyzed twice [ICC: 0.98–0.99; standard error of measurement (SEM): 0.005 m, 0.01 m·s⁻¹, 0.18 m·s⁻²] (12, 14). Further, the mentioned SEMs associated with the automatic video-analysis of the barbell kinematics are constant and do not change in relation to the barbell load used or exercise performed. Therefore, for the same exercise, changes in total variability during repeated lifts or during a single lift (time series data) are mainly caused by changes in the variability of the motor performance.

Until now, only limited amounts of studies have analyzed the variability of barbell kinematics in weightlifting. Furthermore, the analysis of the variability (i.e., reliability) of biomechanical parameters in weightlifting focused on discrete values rather than on time series data. Therefore, this study aimed to assess the intra-session variability of time series barbell kinematics in weightlifting via reliability-metrics (i.e., SEM, ICC) during maximal and submaximal loads in the snatch and the clean and jerk. We hypothesized that barbell load (submaximal vs. maximal) and lifting phase would affect the variability of the competitive lifts (15, 16).

In terms of practical use, total variability should be expressed as a metric of the measured scale (i.e., SEM) to have a direct interpretation of the amount of precision of individual test scores (11). Therefore, we focus on the SEM in the written presentation of the results.

From the results, it can be concluded that lifting phase (Tables 1–3) and barbell load (submaximal vs. maximal, Table 4) influence SEM_IPI, ICC_IPI, and consequently SRD_IPI in the snatch, the clean, and the jerk.

In general, the barbell load and lifting phase less substantially affect the SEM_IPI of the horizontal barbell trajectory in the snatch and the jerk, as opposed to the clean (Table 4). However, for the vertical barbell trajectory in the snatch and clean, the 1st pull, transition, and 2nd pull display smaller SEM_IPI at submaximal loads. In contrast, the SEM_IPI of vertical trajectory in the jerk is less at maximal loads. The vertical barbell velocity displays smaller SEM_IPI at maximal loads for the snatch (except 1st pull), the clean (except transition), and the jerk (except turnover and catch). Finally, in the snatch, the vertical barbell acceleration displays smaller SEM_IPI at maximal loads (except 1st pull). In contrast, the SEM_IPI of vertical barbell acceleration in the jerk is less during the dip at maximal loads and less during the catch at submaximal loads, respectively.

When looking at the time series data of SEM and ICC across the snatch, the clean, and the jerk (Figures 3–5), it can be summarized that the variability of barbell trajectory (x and y), vertical barbell velocity, and vertical barbell acceleration tend to increase as the lift progresses. In addition, it is worth pointing out that during the lift in the snatch and the clean, the variability of vertical barbell velocity is smaller (i.e., smaller SEM) at the end of the single lifting phases than within the lifting phases (except transition).

This study aimed to analyze the intra-session total variability, in terms of absolute (SEM) and relative (ICC) reliability, of barbell kinematics in elite weightlifters during repeated lifts in the snatch, the clean, and the jerk at submaximal (85% 1RMp) and maximal loads (97% 1RMp). Concurring with our hypothesis, the total variability during the snatch, the clean, and the jerk depends on the barbell load and lifting phase. In general, variability in the snatch (primarily in barbell velocity) decreased at maximal compared to submaximal barbell loads. In addition, total variability of time series kinematics increased at both loads as the lifts in the snatch and the clean and jerk progressed.

In the training of elite weightlifters, objectifying specific barbell handling skills using biomechanical analyses (e.g., video analysis) is an important factor to control and monitor lifting performance (12). Accordingly, attention should be given to the inherent total variability of biomechanical data for a specific test (9). In this context, the total variability of maximal barbell velocity during the power clean (40–100% of 1RM) in competitive weightlifters was previously analyzed, showing an SEM ranging from 0.05–0.09 m·s⁻¹ measured with a linear position transducer (16). In comparison, the assessed total variability (i.e., SEM_IPI) in our study for vertical barbell velocity time series data of the snatch, clean, and jerk is slightly smaller (majority of SEM_IPI = 0.01–0.06 m·s⁻¹). Additionally, for total variability of barbell acceleration in the power snatch, power clean and jerk in strength-trained athletes, the SEM has previously been reported to be 1.77 m·s⁻², 1.0 m·s⁻², and 0.55 m·s⁻², respectively, using an accelerometer attached to the barbell (31). Again, in our study, the total variability of vertical barbell acceleration time series data in the three exercises is slightly smaller (SEM_IPI = 0.26–1.0 m·s⁻²). The aforementioned differences in SEM can be explained by the technical mastery of the elite weightlifters in our study and the high precision of the used barbell tracking software. Although barbell trajectory is of major interest in weightlifting research (6, 32), to our knowledge, no studies have assessed total variability (i.e., SEM) of barbell trajectory in the snatch, the clean, and the jerk. Therefore, our study outcomes may give additional insights into how to interpret individual changes in barbell trajectory during training and competition. Furthermore, based on the smaller SEM_IPI at maximal compared to submaximal barbell loads, future studies on variability assessment in weightlifting and strength training should account for barbell load as an additional component that can moderate the total variability of movement outcomes.

Typically, reliability indices of biomechanical variables often rely on single points/events (i.e., discrete parameters) from an entire time series of the measured data. As previously presented for cycling (33) and one-leg hops (26), reporting and analyzing reliability indices as time series data (waveforms or aggregated as integrated pointwise indices), rather than as discrete points, provide more detailed insight into changes in variability over time. In the case of barbell kinematics in the snatch, the clean and the jerk, calculating the reliability indices as time series showed that total variability (i.e., SEM, SRD) increased in later stages of the lift, in particular for the horizontal barbell trajectory. For example, the SRD_IPI for horizontal barbell trajectory in the snatch is about 4 cm in the turnover phase and 6 cm in the catch phase (Table 1). Furthermore, time series variability waveforms can be used to select time points with low variability from which discrete parameters can be derived. In this context, during the 1st pull and 2nd pull of the snatch, the variability of vertical barbell velocity is lowest at the end of the lifting phases. These time points would serve well to define discrete vertical barbell velocity parameters that the coach can use to determine the effects of training on barbell kinematics.

As mentioned previously, the total variability during repeated lifts mainly reflects the variability of the motor performance (i.e., movement). With ongoing skill development, the movement pattern shows less variable and more stable aspects of the movement (34). In this context, the more stable movement patterns (i.e., less variability) were assumed to be those which are goal-directed (34) and hence related to the movement outcome (i.e., performance). Further, it has been discussed that motor variability is regulated that more reproducible output is aimed at situations with a high “reward” (35). Viewed the other way around, specific parts of a movement with lower trial-to-trial variability might be important concerning the outcome. Transferring this idea to the time series SEM data it can be concluded that during maximal lifts, achieving a desired v_max in the snatch, the clean, and the jerk, is one of the “goals” within the lifting movement since variability at these time points is lower compared to others. This fact may underline the importance of the vertical barbell (threshold) velocity to lift a maximal load (18). In contrast, the higher variability in vertical barbell velocity at submaximal barbell loads may be interpreted as less “rewarded” movements with less “precision” needed to achieve the desired goal. Specifically, at submaximal loads, a good lift can be achieved with a wider range of possible v_max values. However, at a maximal barbell load only “one solution” of v_max exists that is related with a good lift, i.e., the threshold velocity of the barbell at the end of the second pull. However, the above-mentioned theoretical assumptions may be of interest for further research.

In conclusion, guiding the weightlifter’s training process based on kinematic measurements of the barbell can improve weightlifting performance. For practical use, knowledge of the typical total variability (SEM) of kinematic barbell parameters can be useful to detect the smallest “real” difference (SRD) in kinematic barbell parameters during training and competition. Our study presents reference values of SEM/SRD for barbell kinematics (i.e., x, y, vy, ay) of the snatch, the clean, and the jerk at submaximal and maximal barbell loads. These reference values can be used to estimate changes in barbell kinematics in response to training. In this context, during training, coaches should take into account that the variability of barbell kinematics can vary depending on the exercise and the barbell load.

For this study, some limitations need to be acknowledged. First, from research on vertical jumps, it is known that variability in performance changes over time (36). In our study, the intra-session test-retest reliability was assessed during one singular training session during the preparation phase of a macrocycle. Therefore, variability at other time points during a macrocycle may have been higher or lower than reported in our study. Second, our kinematic analyses of the barbell were based on a custom build video tracking software with limited access. Using another device for the kinematic barbell analysis may affect the total variability (e.g., SEM). Third, the presented SEM/SRD is specific to the population under investigation (i.e., elite male weightlifters). Therefore, care should be taken when using these reference values for other populations. Finally, the cohort size was small. However, given that the study was conducted with elite athletes performing on an international level (German national team), the overall population to draw the sample is small. This is a well-known limitation when performing research in elite sports.