A Dual-Tech Approach to Measuring Defensive Physical Demands in Basketball Pick-and-Rolls During Official Games: Inertial Sensors and Video Analysis

Featured Application

This study helps basketball coaches design drills that reflect the physical demands of defensive pick-and-rolls. Understanding how different options influence physical load can improve workload management, optimize defensive profiles, and guide team rotations for better efficiency.

Abstract

This study aimed to quantify the physical load of defensive pick-and-roll (PnR) actions according to court location (middle or side), defensive option employed (switch, drop/ice, or trap), and effectiveness (successful or unsuccessful) during official basketball games. Twenty-four male basketball players (age: 20.5 ± 1.1 years; stature: 191.5 ± 8.7 cm; body mass: 86.5 ± 11.3 kg; playing experience: 8.5 ± 2.4 years) from two teams competing in the Lithuanian third division were recruited, with data collected across six official games. Participants were monitored using a combination of video-based time–motion analysis (TMA) and inertial measurement units (IMUs), allowing the calculation of duration, PlayerLoad (PL), and PL·min⁻¹ for each of the 364 defensive PnR actions identified. No significant differences were found based on court location or defensive option employed (p > 0.05). By contrast, unsuccessful plays resulted in significantly higher physical loads than successful ones (duration: p < 0.001, ES = 0.46; PL: p < 0.001, ES = 0.41; PL·min⁻¹: p = 0.047, ES = 0.24). Overall, these findings highlight a consistent physical load based on court location and defensive option adopted and an increased physical load when the defensive effort failed. Therefore, basketball coaches are suggested to consider the physical load of different defensive PnR scenarios when planning training drills, defining performance profiles of defensive strategies, and managing team rotations during games.

1. Introduction

Basketball is an intermittent-based team sport characterized by repeated neuromuscular efforts and high cardiorespiratory requirements [1,2]. The physical load during basketball games includes high-intensity actions such as sprints, jumps, changes of activity, and screening actions [1]. In this regard, quantifying and evaluating players’ workloads is of vital importance to optimizing their performance, and it is useful to generate a large flow of information for basketball coaching staff to characterize the physical load of official competitions [3].

Being a team sport, basketball encompasses collective actions, both at the defensive and offensive ends. From a defensive standpoint, various strategies are used by teams during competitions to stop the opponent offense from scoring [4,5]. In particular, the minimum expression of collective offensive play is the pick-and-roll (PnR) action, also known as “on-ball screen” [6,7,8,9], which has been defined as the legal interposition of an offensive player towards a defender with the aim of freeing a teammate to allow them to shoot or receive a pass [10]. Specifically, during the PnR action, the screener performs a pick to allow the screened player (i.e., ball handler) to gain an advantage over his/her direct defender and then advances (i.e., rolls) towards the basket to receive a pass [11].

While a large body of knowledge exists on the use and effectiveness of PnR actions from a tactical standpoint—both offensively [10,12] and defensively [13]—there is a paucity of studies examining the physical load imposed on players during different PnR actions [6,7,9]. In this area, a recent study [7] analyzed the physical load characterizing small-sided games (SSGs) featuring middle PnR actions which were defended using commonly implemented defensive options, such as switch, drop, and trap. Findings from this study demonstrated higher physical loads [measured through inertial movement units (IMUs)] for the trap defense compared to the other two defensive options. Additionally, when comparing the physical load imposed by defensive actions on middle and side PnR performed during SSGs, higher PlayerLoad (PL) values were found when defending middle PnR actions compared to side PnR [6]. Nevertheless, these studies mostly focused on quantifying the physical load of PnR actions during basketball training sessions, while—to the best of our knowledge—no previous studies have focused on official basketball games, which would provide more relevant information to help coaches understand the physical load imposed on players during competition.

An interesting and innovative way to quantify the specific physical load relative to PnR actions might be the integration of multiple performance analysis systems, such as combining video-based time–motion analysis (TMA) with IMUs. This performance analysis approach could be implemented to evaluate the physical load related to basketball tactical actions such as PnR, with a particular focus on the load imposed on defensive players. To date, only one study [14] has combined TMA and IMU approaches to describe the physical load of activities recorded during official basketball games, focusing on high-intensity specific movements, sprints, and jumps. This study found higher external load intensities during sprinting and high-intensity specific movements compared to jumping and when players were in possession of the ball [14]. However, to the best of our knowledge, no information is available about players’ physical loads when defending PnR actions during official basketball competition. Such detailed analysis could be useful to better describe the performance profiles of basketball games, integrating the physical and tactical domains and thus providing particularly valuable information which basketball practitioners can use to optimize training and competition loads in consideration of key tactical aspects. Therefore, the aim of this study was to quantify the physical load measured during defensive PnR actions according to court location (middle vs. side), defensive options (switch, drop/ice, or trap), and the effectiveness of the defensive action (successful vs. unsuccessful) during official basketball games.

2. Materials and Methods

2.1. Participants

Data were collected over 6 official home basketball games played by two teams competing in the Lithuanian 3rd division [Regionu Krepšinio Lyga (RKL)] and the Lithuanian Student League [Lietuvos Studentu Krepšinio Lyga (LSKL)]. A total of 24 players (age: 20.5 ± 1.1 years; stature: 191.5 ± 8.7 cm; body mass: 86.5 ± 11.3 kg; playing experience: 8.5 ± 2.4 years) were monitored during the experimental period. Throughout the study period, the players attended ~4 practice sessions per week and played one or two official games per week. Only players without injuries in the 4 weeks preceding the data collection were included in the study. Players were informed of the purpose, benefits, and risks of the study by the research staff and provided written informed consent before the start of the study. Ethical approval was acquired from the Institutional Scientific Board for Research of the Kaunas Regional Research Ethical Committee (number: BE-2-97).

2.2. Design

This was an observational study encompassing 364 PnR actions during basketball games played according to FIBA regulations. Each game consisted of four 10 min quarters, with 2 min rest between quarters and a 15 min halftime break. Each game was simultaneously recorded and analyzed via TMA and IMUs, to assess the physical load associated with each defensive PnR action identified. Specifically, TMA and accelerometer datasets were aligned to establish the duration and external load of each defensive PnR activity. Before the beginning of the study, the players were familiarized with the devices used to measure their physical load during official games. The participants were already familiar with each PnR defense typology included in the analysis, as these strategies were regularly implemented during team training sessions and games during the competitive in-season period.

2.3. Procedures

2.3.1. Video-Based Time–Motion Analysis

Games were recorded using a smartphone-embedded camera, sampling at 30 Hz (Redmi 5 plus; Xiaomi, Beijing, China), properly located on the court to allow a complete view of the game, as previously described [14]. A freely downloadable Android software system (Timestamp Camera Free, version 1.235; Bian Di, Google Play Store) was used to record videos. The software includes an on-screen timeline indicating the exact actual time of day (hours, minutes, seconds, and milliseconds) at which activities occurred, as reported in a previous investigation [14]. Each file was exported to a personal computer, where an expert video analyst (with 10 years of experience as a basketball coach) manually performed TMA using the freely available software LongoMatch (version 0.20.8; Barcelona, Spain) to record all PnR actions.

2.3.2. Inertial Movement Analysis

During each game, the players’ physical loads were measured using IMUs (ClearSky T6; Catapult Innovations, Melbourne, Australia). Before the beginning of each game, players were required to wear neoprene vests featuring a pocket placed between the scapulae where the IMU was positioned. Each IMU included a triaxial accelerometer sampling at 100 Hz to record movement in all three planes (transverse, coronal, and sagittal) [15]. The PL, which represents the square root of the sum of the squared instantaneous rate of change in acceleration across the transverse, coronal, and sagittal planes (x, y, and z, respectively) was used to quantify the volume of the physical load, due to its high reliability (within-device CV = 0.91–1.05%; between-device CV = 1.02–1.90%) [16] and its widespread use in basketball research [6,7,17,18]. Additionally, PL·min⁻¹ was calculated to evaluate the physical intensity [14,19]. All data were processed and exported on a personal computer using the OpenField software (version 1.18; Catapult Innovations, Melbourne, Australia).

2.3.3. Integration of Data from Time–Motion Analysis and Inertial Measurement Units

Immediately after each game, PL data were downloaded and stored via Catapult proprietary software (OpenField, version 1.18; Catapult Innovations, Melbourne, Australia). Once the TMA was completed, the TMA and IMU datasets were aligned, based on the start and end times of each defensive PnR action, identified using the “Timestamp Camera Free” software (version 1.235; Bian Di, Google Play Store). Specifically, IMU data were analyzed to identify each defensive PnR action, allowing alignment with TMA-detected events. Combining both datasets enabled the quantification of PL and activity duration for each coded PnR action. This process resulted in a fully integrated dataset containing all relevant activities and their corresponding PL values.

2.3.4. Pick-And-Roll Action Identification

During the investigated games, defensive PnR actions were categorized based on court location (middle or side) and the defensive option employed, including switch, drop (only for middle PnR), ice (only for side PnR), and trap. The definition of each PnR typology has been previously described [6,7] and is displayed in Figure 1.

Moreover, each defensive PnR action was classified as successful or unsuccessful, based on a previous classification [20] (

Table 1. Criteria for the classification of successful and unsuccessful defensive pick-and-roll actions.

Successful Defense	Unsuccessful Defense
Missed two-point shot (+2 pts)	Two-point shot made (−2 pts)
Missed three-point shot (+3 pts)	Three-point shot made (−3 pts)
Turnover (TO)	Foul (F)
Block shot (B)	Foul → 2 free throws (2 FT)
Steal (S)	Two-points shot + F (2 pts + 1 FT)
Offensive foul (OF)	Three-points shot +F (3 pts + 1 FT)
Shot clock violation (24”)	-
Pass out of the two PnR players	Pass + score within the two PnR players

Abbreviations: PnR, pick-and-roll.

).

2.4. Statistical Analysis

Means ± standard deviations and percentages were used as descriptive statistics. Due to the small number of traps occurring during both side (n = 8) and middle (n = 22) PnR actions, only descriptive statistics were used for this defensive option. Linear mixed models (LMMs) were adopted with duration, with PL and PL·min⁻¹ considered as dependent variables, while court location (middle vs. side), defensive option [switch vs. ice (middle PnR) and drop (side PnR)], and outcome (successful vs. unsuccessful) were considered as fixed effects and player and game were adopted as random effects. Cohen’s d effect sizes (ESs) and their 95% confidence intervals (95%CIs) were computed by converting the t-statistics calculated based on the LMMs and interpreted as follows: trivial < 0.20; small = 0.20–0.59; moderate = 0.60–1.19; large = 1.20–1.99; and very large > 2.0 [21]. Descriptive analyses and LMMs were run on the Jamovi software (the Jamovi project, version 2.3.21.0, 2024; Sydney, Australia), while ESs and 95%CIs were calculated using the “compute.es” package in the RStudio software (version 4.2.1). The level of significance was set at p < 0.05.

3. Results

Descriptive statistics for middle and side PnR actions are reported in

Table 2. Descriptive statistics for defensive pick-and-roll actions.

Court Location	Defensive Option	n	%	% Successful	% Unsuccessful
Middle	Switch	122	46.4	77.5	22.5
	Drop	119	45.2	67.5	32.5
	Trap	22	8.4	82.0	18.0
Side	Switch	49	48.5	75.5	24.5
	Ice	44	43.6	66.0	34.0
	Trap	8	7.9	50.0	50.0

, while the results calculated through inferential statistics are displayed in

Table 3. Inferential statistics of defensive pick-and-roll actions by court location, defensive option, and outcome.

Dependent Variable	Court Location		Defensive Option		Effectiveness		P Court Location	P Defensive Option	P Outcome
	Middle	Side	Switch	Drop/Ice	Successful	Unsuccessful
Duration (s)	2.35 ± 1.06	2.45 ± 0.97	2.50 ± 1.12	2.29 ± 0.94	2.30 ± 1.07	2.65 ± 0.91	0.199	0.544	<0.001
PL (AU)	0.54 ± 0.49	0.61 ± 0.52	0.58 ± 0.53	0.54 ± 0.48	0.51 ± 0.50	0.68 ± 0.49	0.075	0.238	<0.001
PL·min−1 (AU·min−1)	13.6 ± 9.87	14.1 ± 9.86	13.3 ± 9.27	14.2 ± 10.4	13.00 ± 9.76	15.8 ± 9.86	0.310	0.095	0.047

Abbreviations: PL, PlayerLoad; AU, arbitrary units.

. For all the investigated dependent variables, unsuccessful PnR actions showed higher values compared to successful PnR actions [duration: p < 0.001, ES (95%CI) = 0.46 (0.22; 0.70), small; PL: p < 0.001, ES (95%CI) = 0.41 (0.17; 0.66), small; PL·min⁻¹: p = 0.047, ES (95%CI) = 0.24 (0.00; 0.48), small]. Moreover, for each dependent variable, an interaction between court location, defensive option, and outcome was found (duration: p = 0.023; PL: p = 0.004; PL·min⁻¹: p = 0.027). Furthermore, an interaction between outcome and defensive strategy was found for PL (p = 0.006).

4. Discussion

This study aimed to investigate the physical loads, monitored by dual technologies, associated with defensive PnR actions during official basketball games, according to court location (i.e., middle vs. side), defensive option (i.e., switch, drop/ice, or trap), and effectiveness (i.e., successful vs. unsuccessful). The main results showed no effects of court location or defensive option on physical loads, while unsuccessful defensive PnR actions produced higher physical loads (i.e., duration, PL, and PL·min⁻¹) compared to successful actions. These results can provide valuable information for basketball coaches in defining the performance profiles of commonly used defensive options during PnR actions, which represent some of the most widely utilized tactics in basketball.

To the best of our knowledge, this is the first study to quantify the physical loads of defensive basketball actions using the combination of video-based TMA and IMUs. Comparing our results with those reported in a previous investigation using a similar approach [14], we found that the duration and PL of defensive PnR actions were generally greater (duration: ~2.5 s; PL: ~0.6 AU) compared to sprints (duration: 1.18 s; PL: 0.4 AU), high-intensity specific movements (duration: ~1.32 s; PL: ~0.4 AU), and jumps (duration: ~0.72 s; PL: ~0.2 AU) compared to official Lithuanian third division male basketball games. Accordingly, these results likely led to a lower PL·min⁻¹ for PnR actions (~14 AU·min⁻¹) compared to sprints (21.7 AU·min⁻¹), high-intensity specific movements (19.6 AU·min⁻¹), and jumps (18.7 AU·min⁻¹) found in the previous investigation [14]. Therefore, we can conclude that defending in various PnR scenarios can provide a higher load in terms of volume (i.e., duration and PL) but corresponds to a lower intensity (i.e., PL·min⁻¹) compared to some of the key physical actions characterizing basketball game-play. This can be attributed to the longer duration of defensive PnR situations compared to the above-mentioned high-intensity neuromuscular actions and to the increased cognitive demands of such defensive scenarios, in which players need to undergo pattern-recognition and decision-making processes which are crucial for success. The present findings provide an interesting snapshot of basketball performance, providing coaches with valuable information on how to manipulate the physical loads of defensive PnR scenarios by applying different tactical strategies.

No effect of court location was found for any of the investigated variables. This result is apparently in contrast with a previous investigation [6], where moderately higher PL values were reported during basketball SSGs (i.e., three vs. three, half-court) when executing middle PnR compared to side PnR. A possible reason for this discrepancy might be due to the different analyses performed across studies. Indeed, in the current study, physical load was monitored during official games and considering the two players involved in defending the PnR. Conversely, in the previous study examining three vs. three SSGs [6], the physical load was measured in training settings and for all the involved players, including the third defensive player and the offensive ones. Previous research agrees that the physical loads of official basketball games are higher than those of training [22]. Furthermore, the role of the third player in the PnR action is likely different to the ones played by the two other more involved players, likely leading to different findings compared to previous research [6]. Therefore, the detailed analysis of the current study suggests that the physical load, measured through video-based TMA and IMUs, is similar for the two defensive players directly involved in the PnR when comparing middle- and side-court locations.

Another interesting result of the current study is the absence of differences in duration, PL, and PL·min⁻¹ between switch (duration = 2.50 s; PL = 0.58 AU; PL·min⁻¹ = 13.3 AU·min⁻¹) and drop/ice (duration = 2.29 s; PL = 0.54 AU; PL·min⁻¹ = 14.2 AU·min⁻¹)], which are two of the most-implemented defensive tactical strategies on PnRs. The lack of statistical difference was somewhat expected, since neither switch nor drop/ice options represent highly physically demanding tasks for basketball players compared to the trap options [7]. Specifically, during the switch option, defensive players are just required to switch the assigned offensive players, which limits the distance which players have to cover [23]. During the ice/drop option, while a high effort is needed for the screened player to chase the ball handler as fast as possible, this effort is somewhat compensated by the stationary position of the defender with respect to the screener, who is just required to slip into the paint and/or near the free throw line. By contrast, when performing trap defense, both defensive players pressure the ball handler simultaneously, likely increasing their physical load, as shown by a previous study [7] indicating trap defense as the most physically demanding option (compared to switch and drop) when defending the middle PnR during basketball SSGs. However, in the current investigation, we found a limited number of traps performed during official games (n = 22, 8.4%, for middle PnR and n = 8, 7.9%, for side PnR), which did not allow us to include this option in the inferential statistical analysis (i.e., LMM). Therefore, future studies should assess a larger number of games to verify whether the higher physical loads elicited by trap options during basketball SSGs [7] can also be confirmed during official games.

Interestingly, the analysis of physical load based on PnR effectiveness demonstrated greater duration, PL, and PL·min⁻¹ in unsuccessful PnR actions compared to successful ones. No previous investigation assessed the physical load based on PnR outcome during official games, making our results hard to compare. A potential explanation for our results could be the occurrence of unorganized efforts and movements by the defensive players to overcome the offensive PnR advantage. In fact, a lack of communication between the two defensive players (i.e., announcing the upcoming offensive screen, its direction, and the agreed-upon defensive coverage according to the game plan) may lead to higher PnR duration and physical load, as players would need to recover from the initial incorrect movements performed during PnR and the advantages obtained by offensive players. Consequently, great footwork, help and recovery, and team rotations are instantly required by both the defensive players directly involved in the PnR and the off-ball screen players. These defensive multidirectional displacements induce a significant physical load coupled to the timing of its development. Another possible explanation could be that offensive players confused the defense coverage by faking to go right of the screen, and then reversed direction at the last second and ran the ball from the left side (the speed and the angle of the screen set by the screener, in order to interrupt the trajectory of the ball handler’s defender) [12]. This deceptive PnR offensive strategy leads to a disorganized defensive reaction, as players start running randomly to stop the ball and cover the open gaps due to the unexpected nature of the move. These findings are crucial for basketball coaches, since unsuccessful defense during a PnR action not only produces a negative impact on the ball possession outcome but also generates greater physical load volumes and intensities for the two defensive players involved. Therefore, coaching staff should consider this information when planning team rotations during official games.

Although this study provides interesting insights for basketball coaches, some limitations should be acknowledged. Firstly, the recruited basketball players competed in the male Lithuanian third division, and therefore the findings might not be generalizable to basketball players competing in higher-division leagues, where games are characterized by a higher physical load, nor to female players. Secondly, the current investigation only focused on quantifying the external load associated with defensive PnR actions, while no information is available on the physical load associated with offensive PnRs. Therefore, future investigations should focus on assessing the physical loads of PnR actions in high-level basketball players from both an offensive and a defensive standpoint.

5. Conclusions

This study provides novel insights into the physical loads (i.e., duration, PL, and PL·min⁻¹) of defensive PnR actions during official basketball games. The main findings showed that court location (middle vs. side) and the defensive option employed (switch, drop, or ice) do not significantly affect physical loads during defensive PnRs. Conversely, unsuccessful PnR defenses resulted in higher duration, PL, and PL·min⁻¹ compared to successful ones. Based on current findings, basketball practitioners should be aware of the higher physical load imposed by unsuccessful defensive PnR situations and should therefore aim at improving the players’ defensive performances to increase the team’s performance in ball possession alongside reducing the physical loads imposed.