Navigating the Data Processing Maze: A Systematic Review of Jump Height Calculations Using Force Platforms

Authors: Ingrid Eythorsdottir, Øyvind Gløersen, Hannah Rice, Amelie Werkhausen, Gertjan Ettema, Paul Solberg, Gøran Paulsen

Abstract

Vertical jump height estimates the ability to oppose gravity and lower body neuromuscular performance in athletes and various clinical populations. The use of force platforms for measuring jump height is increasingly popular due to technological advancements and the equipment's relative ease of use in various settings. However, when utilizing the force platform, ground reaction force (GRF) data must be processed to calculate jump height. While processing the GRF‐time data, several factors could alter the data, leading to inaccurate jump height estimates. These factors include sampling frequency, filtering, cut‐off frequencies of the filter, averaging periods of body weight, integration procedures, selection of take‐off/landing thresholds, and selection of the gravity constant. These data processing steps can alter jump height estimates, with effects ranging from minor (< 0.5%) to major (> 25%). Despite some guidelines on data processing, there is no consistency in the literature or in practice regarding how the GRF‐time data should be processed. Consequently, jump height without specifying the data processing steps may be of limited use to others. The aim of this review was to assist researchers and practitioners in navigating the complexities of data processing to better understand how it influences jump height.

Background

Force platforms are widely used to estimate maximal vertical jump height, a common metric for assessing lower‐body neuromuscular performance in athletes and clinical populations (Eythorsdottir et al. 2024; Xu et al. 2023). Among available technologies—such as contact mats (Bosco et al. 1983), photoelectric cells (Glatthorn et al. 2011), linear position transducers (Cronin et al. 2004) or smartphone applications (utilizing video and/or an accelerometer) (Balsalobre‐Fernández et al. 2015)—force platforms offer a distinct advantage as they allow direct kinetic analysis by capturing the ground reaction force (GRF) signal throughout the jump (Figure 1) (Beckham et al. 2014).

However, the advantage of force platforms also introduces complexity. The GRF signal must be processed through multiple analytical steps, and each decision along the way can meaningfully influence the final jump height estimate. Eythorsdottir et al. (2024) demonstrated that applying different jump height equations to the same jump could result in differences of up to ∼15 cm, emphasizing that these equations cannot be used interchangeably and that the most suitable equation depends on the jump modality tested, the reason for testing, and how you define jump height. Yet, even when a single equation is used, decisions related to force signal processing can independently affect jump height by at least 26% (Street et al. 2001; Harry et al. 2022; Donahue et al. 2021; Pinto and Callaghan 2021, 2022; Pérez‐Castilla, Rojas et al. 2021; Pérez‐Castilla et al. 2019; Hori et al. 2009; Wank and Coenning 2019); a critical issue that, until now, has not been systematically reviewed.

Data Processing Factors

The factors influencing jump height estimates in the data processing procedures include a) the sampling frequency of the force data (Street et al. 2001), b) the choice of applying a filter to the force data (Street et al. 2001; Harry et al. 2022; Pinto and Callaghan 2021), and c) the cut‐off frequency of that filter (Street et al. 2001; Harry et al. 2022; Pinto and Callaghan 2021), d) averaging periods of body weight (Street et al. 2001), e) method of integration (Street et al. 2001), f) start of integration (Street et al. 2001; Donahue et al. 2021; Pérez‐Castilla et al. 2019), g) direction of integration (Wank and Coenning 2019; Wade et al. 2020), h) selection of take‐off threshold (i.e., stop of integration) (Street et al. 2001; Pérez‐Castilla, Fernandes et al. 2021), and i) selection of gravity (Street et al. 2001) (Figure 2). Indeed, calculating jump height using a force platform involves a multitude of factors that must be carefully considered.

Given the widespread use of jump height as a performance variable in research and practice, data processing procedures must be clearly defined, goal‐aligned, and transparently reported to ensure meaningful and comparable results (Street et al. 2001; Harry et al. 2022; Donahue et al. 2021; Pinto and Callaghan 2021, 2022; Pérez‐Castilla, Rojas et al. 2021; Pérez‐Castilla et al. 2019; Hori et al. 2009; Wank and Coenning 2019). To achieve this, it is essential to understand how specific data processing steps influence the final outcome. The following subsections outline the key factors that can affect jump height estimates using force platform data.

Sampling Frequency

Sampling frequency refers to the number of GRF data points sampled per second. Force platforms used in jump assessments typically sample at frequencies ranging from 100 to 2000 Hz, with lower sampling frequencies often observed in portable force platform systems (Bobbert et al. 1996; Hatze 1998; Baca 1999; Aragón‐Vargas 2000; Moir et al. 2005; Domire and Challis 2007; Cormack et al. 2008; Coh and Mackala 2013; Centeno‐Prada et al. 2015; Loturco et al. 2018; García‐Ramos and et al. 2020; Wilder et al. 2021; Lindberg et al. 2022). These variations in sampling frequencies among studies suggest that the amount of data available to detect critical phases in jump height equations (see Eythorsdottir et al. (2024)) may differ significantly. Consequently, the effect of sampling frequency on jump height must be carefully considered (Figure 3).

Sampling at low frequencies (e.g., 200 Hz) means that certain data points are missed, potentially leading to inaccuracies. According to the Nyquist–Shannon sampling theorem, accurate reconstruction of a signal requires sampling at least twice the highest frequency present in the signal to avoid aliasing. However, a sampling rate of at least 5 times the highest frequency component is recommended to ensure decent replication of signal amplitude and accurate event detection. If the sampling frequency is too low, critical features such as take‐off may be misidentified or entirely missed due to insufficient temporal resolution, leading to variable errors in jump height calculations (Street et al. 2001).

When using simplistic data analysis, for example, by defining take‐off and landing as the last/first force samples above a given threshold, take‐off and landing are determined with a time resolution of 1/sample rate. As an example, a flight‐time based jump height from a 200 Hz recording will have a height‐resolution of about 0.5–1.0 cm (depending on jump height). At 1000 Hz, this is reduced to 0.1–0.2 cm. It should be noted that this resolution limitation is theoretically avoidable by interpolating between samples, but only to the extent incurred by the Nyquist–Shannon theorem. At take‐off and even more at landing, that is, when abrupt changes occur, the force signal may contain relatively high frequencies beyond 1/5 x Nyquist frequency for a 200 Hz recording. Thus, identifying the time of take‐off and landing may be affected by aliasing‐like artifacts. This affects jump height calculations using both flight time, impulse‐momentum, and double‐integration methods.

Filtering

GRF‐time data from force platforms is a time‐varying signal that combines the true force signal with noise. Noise can arise from electrical interference in the force sensors, mechanical vibrations in the force platforms, or external factors, particularly in portable force platforms placed on varied surfaces. Low‐pass filtering is used to remove high‐frequency noise from the force measurements. Both the filter type and the cutoff frequency affect the GRF signal used to determine jump height (Figure 3).

Studies on jump height often analyze either unfiltered GRF‐time data or use low‐pass filters (where frequencies above the cut‐off are removed as noise) with cut‐off frequencies ranging from 4 to 50 Hz (Moir et al. 2005; Hasson et al. 2004; Barker et al. 2018; Ferreira et al. 2010; Philpott et al. 2020; Harry et al. 2020; Montalvo et al. 2021; Eagles et al. 2017).

Given this wide variation in reported cut‐off frequencies, it is important to understand how different filtering choices affect jump height estimates. Without this knowledge, comparisons across studies or between systems may be compromised.

Integration

Integration is required when using the impulse‐momentum methods for jump height estimates (Figure 3). For the equations where integration is needed, the start of integration, method of integration, direction of integration, and stop of integration (take‐off or landing, in the equations where integration is used) need to be accounted for. The rationale for integrating GRF‐time data to estimate jump height is detailed in Eythorsdottir et al. (2024). Yet, most studies do not report the specific integration procedure used, raising the question of whether this step has a meaningful influence on jump height estimates.

Take‐Off Threshold

Take‐off is a critical phase when estimating jump height through equations that use take‐off as an input (Eythorsdottir et al. 2024). Take‐off marks the beginning of the flight phase (stop of integration in the impulse‐momentum equations), and its precise identification is essential for accurate outcome measures (Figure 3). In the literature, various force thresholds have been used to define take‐off, ranging from ∼2 to 40 N, from both filtered and unfiltered data (Barker et al. 2018; Harry et al. 2020; Chandler et al. 2018; Guess et al. 2020; Heishman et al. 2020), raising important questions about how different thresholds influence jump height estimates. This also applies to the backward integration approach starting from quiet standing after landing in a reversed time direction (see example in Wank and Coenning (2019)).

The actual force value at which take‐off is detected may differ from the intended threshold, depending on the noise and offsets in the force recordings. This issue can be particularly pronounced when using portable force plates or when testing on non‐rigid surfaces, where residual noise may exceed low thresholds (L. Smith and Jones 2025). Inspection of the force trace during the airborne phase for offset and other disturbances (particularly for portable systems) before deciding on calculation algorithms and thresholds is always recommended.

Body Weight

Newton's second law of motion establishes that acceleration is directly proportional to the force applied to the ground and inversely proportional to body mass (Eythorsdottir et al. 2024). Indeed, assuming that the athlete produces the same force over a period of time (e.g., some months), a change in body mass over the same period will directly affect jump height. Despite its importance for jump height outcomes, most studies fail to report the method used to determine body weight for jump analyses. In the studies that do, body weight is typically obtained by averaging the GRF‐time data over periods ranging from 0.5 to 4 s prior to movement onset. Body weight is then used to normalize the GRF‐time data to calculate acceleration (Figure 3), which is a key variable in several jump height equations (Eythorsdottir et al. 2024).

Given the lack of information on the procedure used to obtain body weight in the literature, it seems to be assumed that body weight is accurately measured regardless of the method used. This is an interesting and potentially problematic assumption and highlights the need to discuss how the procedure used to obtain body weight affects jump height.

Gravity

Jump height is directly influenced by the relationship between the applied net impulse and the opposing gravitational force, g. In practice, there are two ways to increase jump (i) increase the net impulse, or (ii) reduce the mass of the jumper. Of course, there is a theoretical third option, which is to reduce the gravitational acceleration. For example, on Mount Everest, g drops to approximately 9.77 ms^−2^ compared to the standard 9.81 ms^−2^ at sea level.

However, variations in gravitational accelerations across the Earth's surface are minimal and have negligible effects on jump height outcomes. For instance, applying the flight time equation with a typical flight time of 0.5 s results in jump heights of 30.66 cm at g = 9.81 ms^−2^, 30.63 cm at g = 9.80 ms^−2^, and 30.69 cm at g = 9.82 ms^−2^. Similarly, Street et al. (2001) found that using the gravitational constant for the equator instead of the actual test site (latitude ∼45°) resulted in a 0.25% underestimation in CMJ height using the take‐off velocity equation, while using the polar value caused a 0.25% overestimation.

Although these discrepancies were statistically significant, they are practically irrelevant in most testing scenarios. Users should nonetheless be aware that small variations in g, depending on location, can affect jump height estimates. Therefore, although gravitational acceleration will not be discussed further in this review, its potential influence should not be entirely overlooked in contexts requiring high precision.

Current Recommendations

Guidelines have been proposed regarding how to process GRF‐time data when calculating jump height using force platforms (Street et al. 2001; Harry et al. 2022; Donahue et al. 2021; Pinto and Callaghan 2021; Pérez‐Castilla, Fernandes et al. 2021; Aragón‐Vargas 2000; Moir 2008; Perez‐Castilla and Garcia‐Ramos 2018; Chiu and Dæhlin 2020; Vanrenterghem et al. 2001; Pérez‐Castilla et al. 2019; Wank and Coenning 2019; Wade et al. 2020; Kibele 1998); please see Table 1. Yet, these guidelines have not been well implemented, either in the literature or in practice. One reason may be that current guidelines do have important they are highly technical, differ substantially in their recommendations, and are rarely validated across different populations, jump types, or testing environments. Moreover, most guidelines address only a single aspect of the data processing steps influencing jump height estimates (see Figure 2), rather than providing an integrated framework. These shortcomings may make it difficult for practitioners and researchers to navigate the data processing maze.

Indeed, despite some published guidelines, the technical nature of data processing may pose a barrier to widespread implementation. As a result, data processing steps may be overlooked or simplified, potentially affecting the validity of the reported jump height outcome. This is reflected in the literature, where considerable variation is shown. Moreover, commercial software, which most practitioners rely on, rarely discloses how the data has been processed, further contributing to the lack of transparency. Merrigan and colleagues have demonstrated this challenge in whereas some commercial systems showed acceptable agreement for CMJ height estimates, others systematically overestimated jump height compared with reference methods (Merrigan et al. 2024, 2022). These discrepancies were attributed to differences in data processing steps, such as defining system weight and applying integration techniques (Merrigan et al. 2022), illustrating how processing decisions can introduce error into jump height calculations and limit comparability across studies.

The absence of standardized procedures and poor reporting severely limit reproducibility and undermine the comparability of jump height values across studies, systems, and applied settings. The aim of this review is to synthesize the available literature and provide researchers and practitioners with clearer insights into how data processing choices influence jump height estimates from force platform data.

Methods

A literature search was conducted between April 2021 and January 2025 in PubMed, SPORTDiscus, and Web of Science.

In PubMed the following search was (“vertical jump*” OR “squat jump*” OR “countermovement jump*” OR “drop jump*” OR “jump height*”) AND (“force plate*” OR “force platform*” OR « biomechanical phenomena » [mesh] OR “ground reaction force*”) AND (velocity OR filter* OR “body weight”[mesh] OR “flight time*” OR movement/physiology[mesh] OR motion[mesh] OR exercise test/methods[mesh] OR movement/physiology[mesh]) NOT (phone OR app OR device* OR photocell* OR ACL OR inertial*).

In Web of Science and SPORTDiscus the following search was “vertical jump*” OR “squat jump*” OR “countermovement jump*” OR “drop jump*” OR “jump height*”AND “force plate*” OR “force platform*” OR “biomechanical phenomena” OR “ground reaction force*” AND velocity OR filter* OR “body weight” OR “flight time*” OR movement OR motion OR “exercise test*” NOT phone OR app OR device* OR photocell* OR ACL OR inertial*.

The inclusion criteria (i) used a force platform for vertical jump height calculations; (ii) calculating jump height by at least one of the four equations presented in Eythorsdottir et al. (2024): flight‐time (FT), take‐off velocity (ToV), take‐off velocity + the displacement of the CoM until take‐off (ToV + D), or maximum displacement of the CoM (DIS); (iii) different variations of at least one or more of the following data processing steps (alone or in combination) were sampling frequency, filtering, take‐off threshold, integration, body weight; (iv) jump height values were reported separately for each of the different data processing steps investigated in criterion (iii); (v) the participants were healthy and had no reported disabilities; (vi) the papers were peer‐reviewed, (vii) the papers were original empirical studies, and (viii) the papers were written in English. Studies where equipment such as contact mats was placed on top of the force platform were excluded.

Results

Literature Search

The search from the three databases retrieved a total of 2528 results. After removing duplicates between databases, 1568 peer‐reviewed articles were obtained. First, papers were screened by title, excluding papers that did not clearly address jump height. From these exclusions, 82 papers were included for further analysis.

The abstract was carefully read, and 41 articles were excluded at this stage as they failed to provide any direct comparisons in jump height when calculated with the different equations presented in the inclusion criteria, and/or data processing steps using a force platform. The remaining 41 papers were thoroughly read. Two additional papers were retrieved from the reference list at this stage. Of the remaining 43 papers, 9 papers addressed different data processing methods for jump height calculations using the force platform and were included in the present review (Figure 4).

Quality Grading

The quality grading system used in this study was adapted and modified from Baca (1999) based on the relevance of the topic of this review. The quality grading was conducted by researchers I.E., P.S., and G.P. Each paper was graded using seven criteria on a scale from 0 to 2 (0 = no, 1 = maybe/unclear, 2 = yes). Thus, each paper's maximum score was 14 (Table 2). Zero to 6 points were considered low quality, 7 to 10 points were considered moderate quality, and 11 to 14 points were considered high quality. A, b, and c graded four out of the nine articles equally; the remaining five articles were discussed and agreed upon. Of the 81 individual scoring decisions across all articles, only five required discussions before consensus was reached. All included studies were considered high quality, with scores ≥ 13 (Table 3).

Results

For an overview of what each of the papers included, see Table 4.

Sampling Frequency

In Street et al. (2001), CMJ height was progressively underestimated at sampling frequencies below 1080 Hz. The largest underestimation was observed at 180 Hz (−4.4%), whereas smaller errors were seen at 300–900 Hz (ranging from −1.1% to −0.3%). Sampling frequencies at or above 1000 Hz resulted in jump heights within ±0.1% of the reference value (sampled at 5400 Hz).

Filtering

Harry et al. (2022) found that CMJ heights differed by 1.9 cm between a 99% signal power filter and a standard 50 Hz filter, whereas the difference between unfiltered and 50 Hz filtered data was only 0.3 cm. Pinto and Callaghan (2021) observed the smallest difference in CMJ height from the optical motion capture criterion using a 5 Hz dual‐pass filter (0.57 cm), followed by the 5 Hz single‐pass filter (1.05 cm). The largest difference from their reference was found in the unfiltered condition (1.86 cm). Street et al. (2001) reported that jump height was increasingly underestimated as cutoff frequency decreased, with the largest difference observed at 6 Hz using a dual‐pass filter (−26%) and minimal differences at 580 Hz (< 1%), compared with unfiltered data.

Take‐Off Thresholds

J. C. Smith et al. (2024) observed CMJ heights ranging from 28.7 cm (< 1 N) to 30.1 cm (5SD), with a maximum difference of 1.4 cm between thresholds. Pérez‐Castilla, Fernandes et al. (2021) reported CMJ heights of 30.7 cm (five standard deviations [5SD]), 30.5 cm (peak residual force [PRF]), and 30.9 cm (10 N) at 0.5 kg; 18.9 cm (5SD), 18.8 cm (PRF), and 19.1 cm (10 N) at 30 kg; and 9.1 cm (5SD), 9.0 cm (PRF), and 9.4 cm (10 N) at 60 kg. The largest difference between thresholds in their study was 0.4 cm at the 60 kg load. Street et al. (2001) reported CMJ heights increasing from the reference threshold (the thresholds ranged from 0.7 to 2.0 N) to 1% higher at 6 N and 1.5% higher at 10 N.

Integration

Method of Integration

Street et al. (2001) compared the trapezoid and histogram methods for integrating the GRF‐time signal to calculate CMJ height. Using a 1080 Hz sampling rate, the histogram method resulted in a 0.3% underestimation in jump height relative to the trapezoid method. At 5400 Hz, the difference was reduced to 0.06%. The trapezoid method, used as the reference, showed CMJ height differences within ± 0.02% across different sampling rates.

Start of Integration

Meylan et al. (2011) reported CMJ heights of 28 cm (integration 2.5% BW), 29 cm (integration 5% BW), and 29 cm (integration 10% BW), with CVs ranging from 2.4% to 2.8%. No statistically significant differences were reported between thresholds.

Pérez‐Castilla et al. (2019) reported that SJ height varied slightly across thresholds, with the 50 N and 10% of system weight thresholds consistently underestimating jump height relative to the 10 N, 1% of system weight, and 5SD of system weight. At the lowest load, jump height differed by up to 0.3 cm; at 30 kg, the difference reached 0.5 cm; and at 60 kg, the largest difference was again 0.5 cm, with the 10% of system weight threshold yielding the lowest value.

Donahue et al. (2021) reported SJ heights ranging from 33.3 ± 5.6 cm (20 N) to 33.8 ± 5.5 cm (5% of system weight), with all thresholds producing nearly identical values. No statistically significant differences were observed in jump height across conditions.

Street et al. (2001) reported that starting integration exactly at movement onset underestimated CMJ height by 0.5% compared with starting integration 2 s prior to movement onset. Starting integration 0.5–1.5 s prior to movement onset led to deviations below 0.1%, compared with the 2 s reference value.

Direction of Integration

Wank and Coenning (2019) reported that the backward integration method produced CMJ heights nearly identical to forward integration during both maximal and submaximal CMJs, with mean differences of −0.1 cm from the kinematic reference in both conditions. The standard deviation of the bias was slightly higher for backward integration (1.2 vs. 0.9 cm for forward integration in maximal jumps; 0.9 vs. 0.7 cm in submaximal jumps).

Body Weight

Street et al. (2001) reported that averaging body weight over 1.0 or 1.5 s resulted in a CMJ height error of less than ± 1%, whereas a 0.5 s period led to an error of ± 1.4%, and 0.1 s increased the error to ± 3.3%, compared to measuring body weight as the average voltage over the first 2 seconds of the sampling period.

Discussion

The aim of this review was to provide an overview of the current guidelines for the processing of GRF‐time data and to examine how data processing affects jump height results when using force platforms.

Across the included studies, differences in sampling frequency, filtering, take‐off thresholds, integration procedures, and body weight determination each led to variations in calculated jump height, ranging from negligible (< 0.5%) to substantial (> 25%). Although some parameters, such as integration method or direction, showed minimal impact, others, particularly low filter cut‐off and sampling frequencies, resulted in meaningful under‐ or overestimation of jump height. Taken together, these findings emphasize that even small variations in these choices can influence reported jump height values, underscoring the need for greater standardization.

Sampling Frequency

Street et al. (2001) recommended sampling frequencies ≥ 1080 Hz (Table 1 (Street et al. 2001)), as lower frequencies were shown to underestimate CMJ height calculated through the ToV equation, compared to a reference calculation (where the GRF‐time data was sampled at 5400 Hz). This underestimation could be attributed to the late identification of take‐off (based on their take‐off definition; see Section 4.3). Late identification of take‐off increases the negative impulse prior to take‐off during CMJs, which, when integrated with the positive impulse, decreases the net impulse and subsequently underestimates jump height (Eythorsdottir et al. 2024).

Interestingly, despite recommendations to use high sampling frequencies, many researchers and commercial force platform systems continue to sample at frequencies < 1000 Hz when assessing jump height. In applied contexts, the continued use of lower sampling frequencies may be driven less by methodological preferences than by practical constraints built into commercial systems. Portable and commercial platforms often prioritize affordability, ease of use, data storage efficiency, and wireless transmission, all of which can be facilitated by reduced sampling frequencies. However, lowering the sampling frequency to minimize computational demands is arguably less of a constraint with modern technology. Moreover, proprietary software may restrict users to default settings or limit flexibility in adjusting sampling frequency. Together, these factors help explain why many systems and research applications continue to sample below 1000 Hz, despite recommendations to use higher frequencies for optimal accuracy in jump height calculations.

Hori et al. (2009) investigated the lowest acceptable sampling frequency for CMJ assessments, focusing on power, force, and velocity variables, but not jump height directly. Using 500 Hz data as the reference, they compared data between 25 and 400 Hz. Their findings indicated that sampling at frequencies ≥ 200 Hz resulted in deviations of < ± 2% from reference values sampled at 500 Hz. This difference was deemed negligible for practical applications and has since been used to justify the use of lower sampling frequencies (< 1000 Hz) for jump assessments, including jump height calculations (Satkunskiene et al. 2021; Moir et al. 2012; Gómez‐Molina et al. 2018; Beattie et al. 2020). Still, it remains unclear whether such findings generalize to jump height estimates.

It is important to note that the recommendations of sampling > 1000 Hz apply specifically to jump height calculations using the ToV equation (Street et al. 2001). If the primary limitation of low sampling frequencies lies in the incorrect identification of take‐off, these issues may not apply to the DIS equations, where take‐off is not a direct input. However, the DIS equation relies on finding the maximum point in the displacement signal (Eythorsdottir et al. 2024), which could also be affected by the sampling frequency. Nonetheless, no evidence currently supports sampling GRF data at frequencies < 1000 Hz when estimating jump heights using a force platform.

Filtering and Cut‐Off Frequencies

The effect of filtering on CMJ height outcomes has been explored in three key studies, all using the Butterworth filter.

The Effect of Filtering on ToV

Both Street et al. (2001) and Harry et al. (2022) found that CMJ height was underestimated by up to 26% with a dual‐pass filter when applying low cut‐off frequencies, compared with unfiltered data. The underestimation in CMJ height was a result of a decrease in net impulse, caused by the filtering process attenuating high‐frequency components in the net force‐time signal around take‐off. Since take‐off involves a rapid transition from high force to zero (GRF = 0), applying a low cut‐off frequency can distort this high‐frequency changeover, effectively removing part of the true signal. This distortion can affect different parts of the impulse curve unequally and reduce the positive impulse (typically when GRF > body weight) and exaggerate or introduce a negative impulse (when GRF < body weight). This ultimately leads to a lower net impulse and, subsequently, an underestimated jump height with the ToV equation. This point is illustrated in Figure 3. Especially, a combination of low‐pass filtering, choice of sample rate, and integration algorithm may cause large discrepancies in jump height.

Both Street et al. (2001) and Harry et al. (2022) recommended treating GRF‐time data as unfiltered for CMJ height calculations, though Harry et al. (2022) suggested a 50 Hz cut‐off frequency as a viable secondary option. A cut‐off frequency of 50 Hz resulted in ∼2% underestimation in CMJ heights, compared with unfiltered data. Harry et al.'s (2022) conclusion on filtering with a cut‐off frequency of 50 Hz, as a secondary option to unfiltered data, was based on findings of a ∼5% longer unloading phase seen in the unfiltered data, compared to all other filtering methods, which was possibly due to an artifact affecting the force measurements. Indeed, it is unlikely that jump height is the single variable extracted from jump assessments. Thus, a more holistic view is needed when examining errors associated with force plate measurements.

In contrast to the findings presented above, Pinto and Callaghan (2021) recommended filtering the GRF‐time data with a single‐pass, low‐pass filter using a cut‐off frequency of 5 Hz for CMJ height calculations for both ToV and ToV + D. Pinto and Callaghan (2021) stated that their findings conflicted with those of Street et al. (2001) and Harry et al. (2022), but this is incorrect. The reason Pinto and Callaghan (2021) reported their recommendation to differ from those of Street et al. (2001) and Harry et al. (2022) is simply because they included a different criterion, namely a criterion of 3D motion capture versus a criterion of unfiltered data, which was the criterion in Street et al. (2001) and Harry et al. (2022).

Both Street et al. (2001) and Harry et al. (2022) applied a dual‐pass filter, using unfiltered data as the criterion. When comparing the dual‐pass filter results from Pinto and Callaghan (2021) to their unfiltered results, the outcomes from the three studies are remarkably similar (Section 3.3.2). Although these findings suggest that unfiltered data may produce more accurate jump height estimates when using the ToV method, it remains unclear whether these values are actually more accurate or merely more sensitive to noise. As such, it is not possible to determine a definitive “true” value. Therefore, when analyzing jump height using the force platform, the priority should be transparent reporting and consistency in processing decisions, rather than assuming one condition is inherently superior.

The Effect of Filtering on ToV + D

If using the ToV + D equation for CMJ height estimates, more nuances seem to be present. Pinto and Callaghan (2021) observed how CMJ height was overestimated in the filtered version using cut‐off frequencies of 5 Hz when using ToV + D, compared to any of the other cut‐off frequencies. This applied both to the single‐ and dual‐pass filters. As take‐off is selected late using cut‐off frequencies of 5 Hz, resulting in underestimated jump heights with the ToV equation (see discussions above), overestimated jump heights in ToV + D at cut‐off frequencies of 5 Hz must be the result of double integrating force up until the point of take‐off.

Indeed, Pinto and Callaghan (2021) observed how filtering the data using a 5 Hz cut‐off frequency resulted in the lowest take‐off velocity (due to late selection of take‐off) and the highest take‐off displacement, compared with any of the other filtering methods. Interestingly, the increased displacement at take‐off was due to the take‐off definition being time‐shifted relative to the take‐off definition of the unfiltered data. Thus, at the same take‐off definition, the position of the CoM (obtained by double integrating force over time) was lower in all instances for the filtered data using a 5 Hz cut‐off frequency than that of the unfiltered data.

It was merely speculated how the phase shift induced by the low cut‐off frequency might better replicate the assumed true biomechanics of human movement as it “models the second‐order system behavior in excitation‐contraction dynamics” (Pinto and Callaghan (2021), pp. 349).

However, caution is warranted when interpreting these findings. The reference measure of jump height used by Pinto and Callaghan (2021) appears to be based on a single point marker placed on the lower back, rather than a segmental CoM model, which limits its validity as a gold standard reference. Furthermore, filtering the GRF‐time signal with a 5 Hz cut‐off clearly affects both the shape of the force signal and the timing of key events used to estimate jump height (see Figure 3). A phase shift caused by filtering is always an artifact and introduces uncertainty when it (by chance) counteracts another possible artifact. This makes the interpretation of their conclusions, especially the suggested benefits of combining 5 Hz low‐pass filtering with the ToV + D method, highly questionable and in need of further validation.

The Effect of Filtering on DIS

The effect of different filtering cut‐off frequencies on jump height using the DIS equation remains unexplored. Unlike other methods, DIS is not reliant on take‐off to calculate jump height but involves integration over a relatively substantial time period (force is double integrated from standing until the maximum CoM position while in the air (Eythorsdottir et al. 2024), which can lead to accumulation of offsets in the GRF‐time data. Indeed, Chiu and Dæhlin (2020) discussed how integrating the GRF‐time data over the flight phase of the jump could result in calculation errors, as the force platform signal may entail small offsets that are easily unnoticed, leading to increased (accumulated) error during the integration process. Thus, Chiu and Dæhlin (2020) recommended ToV + D over DIS for CMJ height estimations. These observations were made even though Chiu and Dæhlin (2020) filtered their data with a fourth order bi‐directional Butterworth filter using a 40 Hz cut‐off frequency. Other researchers have also cautioned against using the DIS method for SJ estimates on unfiltered data, due to this issue (Kibele 1998). However, such warnings may reflect challenges with uncorrected offsets rather than limitations inherent to the DIS method itself.

Practical Considerations

Before drawing any conclusions on the topic of filtering, it is important to consider some practical factors. The studies discussed above were conducted in laboratory‐based environments. The presence of environmental noise in more practical settings might lead to different conclusions (Harry et al. 2022). In particular, when using portable force platforms that are bound to contain artifacts (the force platform might move slightly during testing when not anchored to the ground) and are placed on surfaces of various materials, the effect of filtering the GRF‐time data might render different implications. Therefore, it is strongly recommended that researchers investigate the impact of filtering on GRF‐time data collected with portable force platforms in practical settings. This would provide a more comprehensive understanding of the role of filtering in biomechanical analyses and enhance the generalizability of the findings.

At present, when measuring CMJ heights using ToV in a laboratory setting, it is generally recommended to use unfiltered force data. Further research is needed before more generalized conclusions can be drawn. Given the sensitivity of results to processing choices, a more robust practice would be for researchers and proprietary software solutions to also publish their unfiltered data alongside processed results. This would allow others to apply consistent signal processing methods and improve transparency and comparability across studies.

Take‐Off Threshold

Three studies have investigated how different take‐off thresholds affect jump height outcomes, providing consistent yet complementary insights into this issue (Street et al. 2001; J. C. Smith et al. 2024; Pérez‐Castilla, Fernandes et al. 2021).

Street et al. (2001) examined the effects of using take‐off thresholds from 2 to 10 N above a calculated offset (see below) on unloaded CMJ height. More specifically, each of the take‐off thresholds was compared against a reference threshold determined by finding the first force value that exceeded a baseline (offset) during the flight phase. The offset was established using a 0.4 s moving average during the flight phase (all flight times were > 0.5 s), and the force value with the smallest standard deviation within this window was selected. The final reference threshold was obtained by lowering this value by 2.4 mV (from the analog signal). The results showed that jump height was significantly overestimated at take‐off thresholds above the reference thresholds (> 2 N), where the overestimation increased with increases in threshold values (6 N threshold resulted in a 1% overestimation, 10 N threshold resulted in a 1.5% overestimation in jump height) (Street et al. 2001).

J. C. Smith et al. 2024 assessed six different take‐off four absolute values (20 N, 10 N, 5 N, and < 1 N), one statistical threshold (five standard deviations above the vertical GRF during the flight phase; 5SD) and one dynamic threshold (the point when GRF dropped below the peak residual force that occurred during the flight phase; PRF). These thresholds were applied to unloaded CMJs to investigate their influence on several performance metrics, including jump heights.

In line with Street et al. (2001), J. C. Smith et al. (2024) found that higher force thresholds tended to result in significantly greater jump heights compared to lower thresholds (< 1 N). However, unlike Street et al. (2001) a consistent pattern of overestimation with increasing thresholds was not clearly observed in J. C. Smith et al.'s (2024) data, as several thresholds produced overlapping results. This overlap was likely due to signal noise around the point of take‐off, which led to multiple thresholds being reached simultaneously in a notable portion of the trials. Notably, J. C. Smith et al. (2024) collected their force data at a sampling frequency of 1000 Hz, whereas Street et al. (2001) used a much higher sampling rate of 5400 Hz. The lower sampling frequency in J. C. Smith et al. (2024) may have limited the temporal resolution of force detection and increased the risk of threshold overlap, possibly obscuring more distinct differences between thresholds.

The overestimations reported in Street et al. (2001) and J. C. Smith et al. (2024) were attributed to early take‐off detection caused by the use of higher force thresholds. Just before actual take‐off, GRF typically drops slightly below body weight, resulting in a brief deceleration phase—a negative impulse. When take‐off is detected too early (i.e., at a higher force threshold), this deceleration phase is partially or completely missed, leading to an overestimation of take‐off velocity and consequently jump height. Both studies advised against relying on arbitrarily selected thresholds, recommending methods aimed at preventing errors caused by detecting take‐off either too early or too late. Street et al. (2001) recommended using their reference threshold (see above) whereas J. C. Smith et al. (2024) recommended that practitioners account for signal noise specific to their force plate system and consider using thresholds such as the PRF (see above) which reflects actual noise during flight, helping to ensure that take‐off detection occurs outside the range of signal fluctuations.

Similarly, Pérez‐Castilla, Fernandes et al. (2021) reported that using arbitrary thresholds, such as 10 N, significantly overestimated loaded CMJ heights. The degree of overestimation increased with heavier loads, reaching up to 4% for jumps performed with 60 kg. In contrast, more refined approaches, such as the 5SD and PRF methods (see explanations above), resulted in more accurate estimates. Interestingly, the degree of overestimation for unloaded CMJs was nearly identical to that reported in Street et al. (2001), with both studies reporting an overestimation of ∼ 1.5% when using a 10 N threshold. In summary, these studies consistently show that arbitrary thresholds can lead to systematic overestimation of jump height, whereas relative thresholds (e.g., PRF, 5SD, or reference‐based methods) better reflect the actual behavior of the signal and are therefore more appropriate for accurate detection of the take‐off.

Reliability Considerations

Both Pérez‐Castilla, Fernandes et al. (2021) and J. C. Smith et al. (2024) concluded that the choice of take‐off threshold does not compromise the reliability of CMJ height measurements, whether in loaded or unloaded conditions, as long as the same threshold is applied consistently within the same force plate system. This supports the idea that within‐system test‐retest reliability remains robust across thresholds. However, J. C. Smith et al. (2024) also noted slight differences in reliability across thresholds, with the lowest threshold (< 1 N) showing higher variability due to its proximity to signal noise. These findings underscore the importance of using a consistent approach within a system and highlight potential limitations when comparing results across different methodologies or force platforms, particularly when signal noise differs between setups.

Practical Considerations

Although the methods for detecting take‐off using relative thresholds outlined above offer improved accuracy and reliability compared to arbitrary thresholds, they are not widely implemented in practice due to their relative complexity. Street et al. (2001) proposed a method based on analog voltage signals, which, although highly precise, is inherently complex and deeply tied to academic contexts. In contrast, Pérez‐Castilla, Fernandes et al. (2021) and J. C. Smith et al. (2024) introduced the 5SD and PRF methods, which dynamically adjust the take‐off threshold based on force signal stability during the flight phase. These approaches are more adaptable but still require additional data processing and assumptions about signal behavior, which may reduce their appeal for applied settings.

Importantly, both the 5SD and PRF methods depend heavily on the quality of the GRF‐time signal. This reliance poses challenges in settings with increased signal noise, such as when using portable force platforms or testing on uneven or compliant surfaces (L. Smith and Jones 2025). To date, no studies have systematically evaluated how these dynamic methods perform under varying conditions of signal noise, sampling frequency, or filter settings. In scenarios of signal noise, using a fixed threshold such as 10 N may serve as a practical compromise, albeit with a known tendency to overestimate jump height.

In summary, the findings of Street et al. (2001), J. C. Smith et al. (2024), and Pérez‐Castilla, Fernandes et al. (2021) collectively argue against the use of arbitrary take‐off thresholds. Their results highlight the importance of tailoring take‐off detection methods to the specific context and ensuring adequate signal quality. When measuring CMJ heights (loaded or unloaded) in laboratory settings with unfiltered data sampled at high frequencies (> 1000 Hz), these studies have recommended using a take‐off threshold near 2 N for unloaded jumps or the 5SD/PRF methods for unloaded and loaded jumps. Further research is necessary to explore the suitability of these recommendations in other settings, such as field environments, portable platforms, and different jump modalities.

Integration

Three critical aspects of integration procedures must be (i) the integration method, (ii) the start of integration, and (iii) the direction of integration.

Method of Integration

Integration of force over‐time approximates the area under the GRF‐time curve, and this approximation can be achieved through various methods (Kong et al. 2020). The trapezoid rule is the most commonly cited method in the jump height literature (Moir et al. 2009; Dias et al. 2011; Toft Nielsen et al. 2019; McMahon et al. 2021; Wank and Coenning 2019), though the Simpson's rule has also been reported (Chiu and Dæhlin 2020; Pérez‐Castilla et al. 2019). Notably, most studies fail to disclose the integration method used, making it difficult to assess its impact on results.

The influence of different integration methods on jump height outcomes has been studied in limited detail. Street et al. (2001) compared the trapezoid rule with a histogram‐based approach for integrating GRF‐time data. At a sampling frequency of 1080 Hz, histogram integration produced a small but systematic underestimation of jump height (0.3%). This error was further reduced to 0.06% when sampling at 5400 Hz. Although the magnitude of the error was negligible, the consistent underestimation led the authors to recommend the trapezoid rule as a more reliable method (Street et al. 2001).

The underestimation associated with histogram integration when jump height was calculated using the ToV method until take‐off was attributed to errors in the force signal just prior to take‐off. Specifically, histograms caused an increase in negative impulse near take‐off, whereas errors in other portions of the GRF‐time signal tended to cancel each other out (Street et al. 2001). The findings of Street et al. (2001) suggests that different integration methods might disproportionately affect specific parts of the force‐time curve. Despite this, no studies have yet investigated whether commonly used integration methods, such as the trapezoid rule versus Simpson's rule, produce different jump height estimates across various jump modalities with distinct force‐time curves.

Street et al. (2001) also noted that the accuracy of integration methods may depend on the sampling frequency of the force signal. However, it remains unclear how integration methods perform at lower sampling frequencies (e.g., 500 or 200 Hz), which are commonly reported for jump height estimates in the literature. This knowledge gap highlights the need for further research to assess the impact of integration methods on jump height outcomes, particularly when data is collected at lower sampling rates.

Start of Integration

Calculating jump height using the impulse‐momentum theorem requires measuring the net impulse during the propulsive phase between clearly defined start and stop points in the GRF‐time series (Eythorsdottir et al. 2024). The choice of integration start point is critical, as the jump height equations using integration require the integration to start when velocity is zero (Eythorsdottir et al. 2024). Various approaches to define the start point have been reported in the literature, (i) time thresholds (e.g., starting at a fixed time before movement), (ii) absolute thresholds (e.g., force exceeding or falling below a set value such as 50 N), (iii) relative thresholds (e.g., a percentage of body or system weights) or, (iv) hybrid versions (combing an absolute or relative thresholds with a time component).

Time‐Based Thresholds

Time‐based thresholds define integration starting at a fixed time before movement onset. Street et al. (2001) investigated different time thresholds ranging from 0 to 1.5 s prior to movement initiation and found no significant differences in CMJ height outcomes. However, they recommended starting integration at least 0.2 s before movement onset, as random error was reduced by half compared to shorter intervals. Although this approach ensures a stable starting point, it relies on the specific definition of movement onset by Street et al. (2001). In their method, the start of movement was identified based on deviations in the GRF relative to body weight. First, the peak residual force was determined during a two‐second quiet standing period by averaging the GRF signal and identifying its maximum deviation from body weight. The movement threshold was then set as 1.75 times this peak residual value. To detect movement onset, the GRF signal was scanned forward to find the first point where the GRF exceeded or fell below body weight by more than the threshold. A backward search was then conducted to identify the last instance where the GRF crossed body weight, which was defined as the true start of movement. Although this method provides a highly reliable way to detect the start of movement, it has not been widely adopted in the scientific literature, likely due to its relative complexity.

Relative Thresholds

Relative thresholds determine the start of integration based on a percentage of body or system weight. Meylan et al. (2011) examined different relative thresholds (when the GRF fell below a threshold of 2.5%, 5%, and 10% of body weight) and found that starting integration at 5% or 10% of body weight led to an overestimation of CMJ height by 3.6% when estimated based on the ToV equation. This occurred because a portion of the eccentric phase was excluded, meaning that integration began whereas the athlete was already in motion. In contrast, starting integration when the GRF dropped below 2.5% of body weight, retained more of the force signal, and produced more accurate results. However, this threshold exhibited higher variability in reliability, though the practical impact was considered minor.

Similarly, Pérez‐Castilla et al. (2019) investigated relative thresholds in loaded SJs, comparing starting integration at 1% and 10% of system weight with absolute and hybrid thresholds (see discussion below). They found that higher thresholds, such as 10% of system weight, led to systematic underestimation of SJ height, particularly with heavier loads. The underestimation was attributed to a loss of the early movement phase, leading to a reduced positive impulse. In contrast, the 1% of system weight threshold retained more of the force signal and yielded more accurate estimates, aligning with the findings of Meylan et al. (2011).

Absolute Thresholds

Absolute thresholds define the integration start at a fixed force level, such as 10 N or 50 N. Pérez‐Castilla et al. (2019) compared absolute (10 and 50 N) thresholds to the relative thresholds discussed above in loaded SJs, and found that higher thresholds (50 N) underestimated SJ height by on average 1.9%, with the error increasing as load increased (0.9% at 0.5 kg, 1.8% at 30 kg, and 2.9% at 60 kg). This underestimation was attributed to the loss of the early phase of the SJ movement, which, in the case of SJs, reduced the measured positive impulse. In contrast, lower thresholds (10 N) retained more of the GRF‐time signal and produced more accurate SJ height estimates.

The use of both the relative and absolute thresholds leads to a start of integration slightly after the true onset of movement. Starting integration after the onset of movement voids the assumption of zero initial velocity, which is required to apply the impulse‐momentum theorem (Eythorsdottir et al. 2024). Hence, starting integration after the onset of movement—even slightly—introduces systematic errors in jump height measurements and should be avoided.

Hybrid Threshold

To address the limitations of relative and absolute thresholds where integration is started a time point where the velocity is not zero, a hybrid version has been introduced, combining a relative or absolute threshold with a time‐based component.

The 5SD method, proposed by Pérez‐Castilla et al. (2019) for loaded SJs, defines movement onset as the first point where the GRF exceeds system weight by 5SD of system weight, then going back 30 ms in time. In their study, SJ height was significantly underestimated at all loads during higher thresholds, such as 50 N and 10% of system weight. The underestimation increased with greater loading (1.1% at 0.5 kg, 2.3% at 30 kg, and 4.7% at 60 kg). Despite similar levels of reliability across thresholds, the 5SD method showed slightly better consistency in loaded SJ height estimates, particularly at low and high loads (Pérez‐Castilla et al. 2019). This performance stability, combined with its ability to avoid the underestimation seen with higher thresholds, likely explains why the authors recommended it over more arbitrary or fixed alternatives.

Donahue et al. (2021) implemented a hybrid approach in which integration was consistently initiated 0.3 s prior to the defined movement onsets, regardless of the threshold type. They examined both relative thresholds (2.5%, 5%, and 10% of system weight) and absolute thresholds (20 and 50 N), along with the 5SD method proposed by Pérez‐Castilla et al. (2019) for SJ heights. Using this hybrid approach, Donahue et al. (2021) observed small differences in SJ height outcomes across thresholds, with a maximum difference of 1.5% between the 20 N and 5% of system weight thresholds. Notably, they reported a smaller difference of 0.6% between the 10% of system weight and 5SD thresholds, compared to the 1.1% differences observed by Pérez‐Castilla et al. (2019) under the same conditions. The smaller difference in Donahue et al. (2021) can likely be attributed to their consistent earlier integration start point (0.3 s before 10% of system weight was reached), which captured more of the force signal and reduced discrepancies. Taken together, both studies strongly suggest that starting integration earlier than the onset of movement yields more valid results.

Summary and Considerations

It is important to recognize that commonly used threshold‐based methods, whereas convenient and supported in parts of the literature, do not necessarily represent theoretical best practice. By definition, thresholds risk missing the initial portion of the force development, particularly in loaded SJs where force is variable. This may lead to more substantial errors than the potential drift caused by slightly longer integration windows. Thus, literature “recommendations” should not be mistaken for good practices. The key requirement integration should begin early enough to ensure that velocity is zero and the full force signal is captured.

Direction of Integration

Integration is typically initiated at the beginning of the jump (forward integration). However, backward integration—starting integration from the landing—has been proposed as a viable option for certain jump modalities, such as SJs and DJs (Eythorsdottir et al. 2024). For the DJ, backward integration has primarily been discussed as a method to calculate DJ heights and has thus been included in Eythorsdottir et al. (2024). For other jump modalities, research on backward integration is limited, with only one study to date investigating its effect on CMJ height (Wank and Coenning 2019).

Wank and Coenning (2019) compared forward and backward integration for submaximal and maximal CMJ heights using the ToV + D equation, where heel‐lift distance (+D) was measured via 3D motion capture. Forward integration began at the onset of movement, defined as the first force value exceeding 0 N (identified by searching backward from a reference force threshold). Backward integration followed an identical approach but started from the landing, searching forward to identify the first force value exceeding 0 N.

When compared to a 3D motion capture criterion, both forward and backward integration produced equivalent CMJ height estimates, differing by only 0.3 cm on average. However, the backward integration method displayed slightly larger variability (95% confidence 2.25 cm for backward integration vs. 1.85 cm for forward integration). The increased variability was attributed to an average of 0.3 s longer landing time during backward integration, compared to the push‐off time in forward integration. A longer landing time suggests that relying on landing data for integration may introduce variability, potentially due to differences in returning to a stable standing position post‐jump. The authors also noted that while no direct evidence suggests landing dynamics differ between jump modalities, these findings are currently only validated for CMJ height estimates (Wank and Coenning 2019).

For SJs, several studies caution against calculating SJ heights using ToV + D or DIS with forward integration (Wade et al. 2022; Wank and Coenning 2019; Wade et al. 2020). This is due to noisy force data in the squat position and the longer integration time required when forward integration begins as the participant approaches the squat position. For example, (Kibele 1998) reported an integration time approximately 1.7 s longer than the current recommended start point for ToV (see discussion above), as participants held the squat position for 2 s. Integration for SJs should ideally begin when the participant is standing to ensure an initial velocity of zero. However, if the squat data is noisy, backward integration may provide more accurate results.

It is important to note that all studies on backward integration have been conducted using in‐ground laboratory force platforms. For portable force platforms, landing forces may be attenuated due to lower force thresholds (e.g., platforms “maxing out” at lower values), potentially affecting data accuracy during backward integration (Wade et al. 2022). Furthermore, achieving consistent landings on portable platforms, especially with both feet planted firmly, can pose challenges (Baca 1999).

As suggested by Wank and Coenning (2019) future research should evaluate the accuracy of backward integration on portable force platforms, particularly for modalities such as SJs and DJs where forward integration is less reliable.

Averaging Periods of Body Weight

Street et al. (2001) examined the influence of body weight averaging periods ranging from 0.1 to 1.5 s on CMJ height estimates and found no differences in CMJ height calculated through ToV across averaging periods. However, the random error increased as the averaging period shortened (Section 3.3.5). Based on these findings, they recommended averaging body weight over at least 1 s to minimize error.

Notably, the small error in body weight observed for shorter periods (e.g., 0.13% for a 0.1 s period) resulted in a 26 times greater error in jump height due to the accumulation of errors. To emphasize, Kibele (1998) observed how body weight tended to vary by < 1% between trials within the same participant. Even though an error of 1% is a relatively small error, it impacts jump height substantially. Street et al. (2001) illustrated this by recalculating jump height for the 22 participants included after increasing each body weight by as little as 0.25%. An increase in body weight of 0.25% caused a 6.5% underestimation in jump height, highlighting the importance of determining accurate body weight when calculating jump height. Similarly, in a simulation study, Vanrenterghem et al. (2001) found that variability in body weight, which reached 1.7% between trials, caused jump height errors of 4.5 cm. More recently, Burnett et al. (2023) reported that a 1 kg change in body mass caused jump height differences of 7 ± 7 cm, with variability ranging from 11.5 to 0.4 cm.

Pinto and Callaghan (2022) investigated different body weight averaging periods for CMJ and SJ assessments, though not examining its influence on jump height. They found that shorter body weight averaging periods (e.g., 0.5 s) resulted in larger errors in body weight (0.22%), which resulted in 23% differences in take‐off displacement compared to longer averaging periods. Averaging body weight over 1.5 s reduced body weight errors to 0.04%, decreasing displacement differences to 4%. A 2‐second averaging period further minimized these errors. Thus, for ToV + D and DIS, where displacement is used for jump height estimates, averaging body weight over 2 s should be recommended over a 1‐second averaging period.

Additionally, Pinto and Callaghan (2022) observed that small errors in body weight significantly influenced the definition of the onset of movement when it was determined relative to body weight (see examples in Section 4.4.2.2). For example, an error of 0.02 ± 0.80 N (95% CI: ± 1.6 N) in a 0.5‐second body weight averaging period led to a 95% CI of ± 72% for the start of movement threshold. A 1.5‐second averaging body weight period resulted in a 95% confidence interval of ± 23% for the threshold of start of movement, where body weight differed by only 0.01 ± 0.15 N (95% CI: 0.29 ± N). When averaging the force signal over 1 s, the body weight errors were smaller (0.00 ± 0.29 N, 95% confidence 0.57 N), though resulting in a 95% confidence interval of ± 43% for the threshold of start of movement. These findings underscore that short averaging periods can introduce significant variability in both body weight and start of movement thresholds, potentially compromising jump height calculations in the equations where these variables are used (Eythorsdottir et al. 2024).

Practical Considerations

Even though the findings above offer valuable information on the effect of determining body weight, the results can only be interpreted according to their protocol and equipment set‐up. To exemplify, as of today, the recommendations from Street et al. (2001) and Pinto and Callaghan (2022) are only valid when calculating maximal CMJ or SJ heights performed with arms on the hips, using unfiltered data sampled at 1000 Hz in a laboratory‐based environment. Moreover, Pinto and Callaghan (2022) instructed their participants to hold their arms crossed over their chest when body weight was averaged. Such a procedure is unlikely to be used in practice, as body weight is usually averaged during the “3‐2‐1” commands prior to the jump when arms are placed on the hips. Considering the effect a small error in body weight has on jump height, it is likely that the findings of Street et al. (2001) and Pinto and Callaghan (2022) are inflated further if averaging body weight over noisy GRF‐time data. Also, as the results of Street et al. (2001) were taken in relation to their definition of the start of movement, they could differ if other definitions were applied. Finally, given that landing data tends to be more variable than push‐off data (Wank and Coenning 2019), care must be taken if averaging body weight from the landing (e.g., when applying backward integration procedures, which have been recommended for certain jump modalities (Eythorsdottir et al. 2024)).

Given the significant influence that body weight estimations have on jump height calculations, future research should investigate the most appropriate method for determining body weight across different jump modalities. In the meantime, it is crucial that both researchers and proprietary software solutions clearly report how body weight was obtained. Analyses and interpretations should be made with an understanding of how variations in body weight determination may impact jump height results.

Practical Applications

The purpose of this review was to help researchers and practitioners navigate through the data processing maze to better understand its effect on jump height estimated using force platform data. More specifically, we aimed to demonstrate the expected differences in jump height and the reasons for these, considering the different data processing steps that can be chosen.

First and foremost, the lack of information in the literature on how the GRF‐time data has been processed when calculating jump height using force platforms is surprising. Due to this, much of the jump results reported in the literature cannot be used by others, unless the same testing and data analysis methods have been used by the practitioner/researcher, which is unfortunate. We strongly encourage researchers and practitioners to report the necessary information on how jump height has been calculated. The necessary information is as follows.The equations used to calculate jump height (Eythorsdottir et al. 2024).The sampling frequency of the force data.The details of any signal filtering that has been applied, including information on the cut‐off frequency.Averaging period of body weight/procedure used to obtain body weight.Method of integration, start of integration, and direction of integration.End of integration (i.e., take‐off/landing threshold (where applicable)).

Including this information in the methods section will help make the results reproducible and allow others to interpret the results appropriately. Moreover, as most practitioners rely on the calculations obtained from commercial software, the commercially available software should list the information addressed above in their product documentation or ideally allow these settings to be selected within the software to allow comparisons with other data. Doing so will enhance the transparency, replicability, and comparability of results across studies and force platform systems.

If using the same calculation procedure for repeated testing, the exact data processing steps are not of practical importance as the reliability appears to be good—although only examined for the take‐off thresholds and start of movement. If planning to compare test results with those obtained by others, it is advised to use the same data processing procedures as those reported previously (assuming the necessary information has been provided). In cases where the necessary information is lacking, the test results are simply not suitable for comparison.

It is important to note that the different data processing steps are interconnected and can influence each other. Therefore, the current recommendations for the different data processing steps should not be interpreted in isolation. For example, even if the same take‐off threshold definition is used as in previous studies, the results may not be comparable if sampling frequency and filtering procedures are not consistent. Therefore, when conducting jump analysis, researchers and practitioners should carefully consider all the relevant data processing steps and strive to maintain consistency with previously established procedures to ensure accurate and comparable results.

In Figure 5, an illustration of the potential interdependence of the various data processing steps is given.

Currently, we only have information on the magnitude of one of the variables depicted in Figure 5, in isolation. In isolation, we could expect differences of up to 15 cm in jump height, depending on the equation chosen (Eythorsdottir et al. 2024), and differences of up to 26% can arise from variations in one of the data processing steps. However, we do not yet know the total expected error in jump height when taking into account the interplay between all the factors influencing jump height estimates. Although some of the individual errors addressed in this review may seem small, the cumulative effect of these errors could lead to significant discrepancies in jump height results. Therefore, it is crucial for future research to explore the combined effects of all the factors that influence jump height estimations, for example, through Monte Carlo simulations, such that standardized methodological frameworks could be established. In addition, the establishment of open datasets and shared processing pipelines would allow direct comparison of methods and improve reproducibility. Until we gain further knowledge on the interplay between the factors addressed in this review, in conjunction with those discussed in Eythorsdottir et al. (2024), caution should be exercised when interpreting jump height calculated using force platforms.

Conclusions

This review underscores the significant impact of data processing choices on jump height outcomes derived from force platform data. From sampling frequency and filtering to integration methods and take‐off threshold definitions, each step can meaningfully alter results, both independently and interactively.

Current guidelines on data processing are largely based on maximal CMJs performed in laboratory settings using the ToV equation, with arms on hips. These narrow conditions limit the generalizability of recommendations, leaving their applicability to other jump types, settings, and systems uncertain. This lack of alignment between research protocols and real‐world practice creates a methodological gap that undermines the comparability and practical utility of reported jump heights.

In this review, we provide guidelines instead of a generally applicable recipe due to the great number of possible differences between data collection and analysis methods and the current state of research. Until future research addresses how these variables interact across diverse settings, caution is warranted when interpreting and comparing jump height results between systems or studies. Crucially, jump height cannot be considered a valid or interpretable metric without full transparency regarding how it was calculated, including the equation and all data processing steps involved. Without this context, comparisons between systems or studies risk being misleading. These findings highlight the importance of greater standardization in data collection and analysis procedures, particularly when practitioners or researchers aim to compare their data with published results.

Author Contributions

Conceptualization and IE, ØG, HR, AW, GE, PS, and GP. Literature IE, GP, and PS. Data IE, ØG, HR, AW, GE, PS, and GP. Writing – original draft IE, GP, and ØG. Figures: IE, AW. Writing – reviewing and IE, ØG, HR, AW, GE, PS, and GP.

Funding

The authors have nothing to report.

Ethics Statement

The authors have nothing to report.

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.