IMU Validation for Coaching: Reliability, Validity, and Practical Use

A meta-analysis of 119 clinical studies on IMU-based assessment tools (Camomilla et al., 2018) reported that 800Hz IMU systems agree with force plates (gold standard) at ICC 0.94 with mean absolute error of 1.2cm for jump height. The same review warned that uncertified consumer IMUs can drift to errors of 4.7cm. In other words, the question is not "are you using an IMU" but "which IMU, validated how?" This guide walks coaches and sport scientists through the practical validation procedure needed to judge whether their IMU data is trustworthy. We cover the conceptual distinction between reliability, validity, and sensitivity; on-floor calibration; gold-standard comparison protocols; and how to write an error-range report that supports coaching decisions. The PoinT GO 800Hz IMU has documented ICC 0.92+ agreement with force plates in independent research, and this article shares the validation workflow transparently.

Three concepts must be cleanly separated when evaluating measurement tools.

Reliability answers "how consistent are repeated measurements under identical conditions?" It is typically expressed via ICC (intraclass correlation coefficient) and CV (coefficient of variation). ICC at or above 0.9 is considered excellent.

Validity answers "how accurately does the measurement reflect the true target?" It is evaluated against gold standards (force plates, optical motion capture) using correlation, mean absolute error (MAE), and Bland-Altman analysis.

Sensitivity answers "how well does the tool detect meaningful change?" Common statistics include SDD (smallest detectable difference) and MDC (minimal detectable change).

The table below summarizes typical thresholds for jump assessment.

All three must be satisfied. High reliability with low validity means "consistently wrong." For specific jump variables, see countermovement jump and drop jump technique.

Rather than rely solely on manufacturer-published validation, perform minimal re-validation in your own environment. The following five-step procedure can be completed in 1-2 days.

Step 1: Calibration check — Place the sensor on a level surface and confirm acceleration outputs 0g (excluding gravity axis). Confirm gyroscope drift remains within 0.5 degrees per 10 seconds.

Step 2: Test-retest reliability — Five athletes perform three CMJs each on the same day, yielding 15 measurements. Compute ICC and confirm a value at or above 0.9.

Step 3: Gold-standard comparison — Ideal: force plate. Alternative: standardized Vertec or tape measure performed simultaneously. Collect 30+ paired observations and run Bland-Altman analysis.

Step 4: Multi-day stability — Test the same athletes at the same time of day over five consecutive days. CV ≤ 5% indicates good stability.

Step 5: Detection of meaningful change — Induce controlled fatigue (e.g., five squat sets) and check whether the pre/post difference is statistically significant. Applying Hopkins's (2000) SWC (smallest worthwhile change) framework is recommended.

Completing these steps produces a quantitative report of data quality in your environment. We recommend it as a prerequisite before launching an athlete testing battery.

IMUs can extract many variables, but each has different error characteristics. Coaches must know which decisions a given metric can support.

Jump height is the highest-fidelity output, with mean absolute error around 1-2cm on an 800Hz IMU. That is sufficient for tracking individual baseline shifts. Barbell velocity (VBT) errors fall in the 0.02-0.04 m/s range, supporting squat velocity zone intensity prescription.

Rotational power carries somewhat larger error (around 5-8%) but is still adequate for trend tracking. The rotational power measurement guide details the full protocol. Joint range of motion sits at ±2-3 degrees, enabling clinical decisions in ankle dorsiflexion or shoulder ROM tests.

The table below maps each metric to recommended and caution-flagged use cases.

The takeaway is to never treat all metrics with equal trust. For absolute-value comparisons such as 1RM estimation, pair the IMU output with the correction formulas in the 1RM calculation methods guide.

To translate validation results into decisions rather than reports gathering dust, follow these principles.

First, attach an error band to every reported measurement. Example: "Today's CMJ: 42.5cm ±1.5cm." Coaches and athletes immediately grasp precision intuitively.

Second, set decision thresholds to at least double the error band. If jump-height error is ±1.5cm, meaningful change requires ±3cm or more. Smaller fluctuations are likely measurement noise and should be ignored.

Third, weight trends over single observations. A 5-7 day moving average suppresses noise and amplifies real signal. Buchheit (2014) reported that moving averages improve decision accuracy by 23% over single-day values.

Fourth, standardize the measurement environment. Same time of day, same warmup, same shoes, same sensor placement. Environmental variance often exceeds sensor variance.

Fifth, re-validate periodically. Repeat the five-step procedure every six months to detect sensor drift or algorithm-update effects. Cross-modality validation with medicine ball throw or broad jump test adds confidence.