Velocity-Based Fatigue Detection: Using Bar Speed to Manage Training Load

In a 2017 study by Pareja-Blanco et al., athletes who terminated squat sets at a 20% velocity loss retained 98% of their pre-training jump height, while those who continued to a 40% velocity loss showed a 7.3% CMJ decrement the following morning — despite performing the same total volume in subsequent weeks. The critical insight: velocity is not just a training variable but a real-time fatigue signal. An 800 Hz IMU sensor on the bar knows the athlete is fatigued before they do, before blood lactate peaks, and before subjective RPE reaches 9–10. This article reviews the mechanisms behind velocity-based fatigue detection and the protocols that translate the science into daily training decisions.

Why Bar Velocity Detects Fatigue Earlier Than Other Markers

Neuromuscular fatigue manifests as a reduction in motor unit firing rate, synchronisation, and recruitment — all of which reduce the rate of force development (RFD) and thus bar velocity at any given load. Because velocity is proportional to RFD across the concentric phase of a lift, an IMU sensor captures fatigue-induced motor unit impairment in real time, within the same set it occurs.

Comparison with other fatigue markers:

• Blood lactate: Peaks 3–5 minutes after a set, lags behind velocity changes, and requires invasive sampling.
• Rating of perceived exertion (RPE): Correlates with velocity loss (r ≈ 0.78) but shows wider inter-individual variance and cannot detect set-by-set changes within minutes.
• Creatine kinase (CK): Reflects muscle membrane damage accumulated over hours to days — a lagging indicator, not a real-time signal.
• Heart rate variability (HRV): Useful for between-session readiness but not sensitive enough for intra-session fatigue during resistance training where HR fluctuates rapidly.

Mean propulsive velocity (MPV) — the average velocity measured only during the phase when bar acceleration exceeds gravity — is the most stable and reproducible velocity metric for fatigue detection, showing a coefficient of variation of 2–4% across calibrated linear position transducers and high-frequency IMUs.

Intra-Set Velocity Loss: The Primary Monitoring Signal

Intra-set velocity loss (VL%) is calculated as the percentage drop from the fastest repetition in a set to the final repetition: VL% = ((fastest rep MPV − final rep MPV) / fastest rep MPV) × 100.

The relationship between VL% and the percentage of repetitions-in-reserve (RIR) is well-established in the squat and bench press literature. Gonzalez-Badillo et al. (2011) demonstrated that a 20% VL in the squat corresponds to approximately 3–4 RIR at loads between 60–80% 1RM, while a 40% VL corresponds to 0–1 RIR (near-maximal set). This means VL% functions as an objective rep-in-reserve counter without requiring athletes to self-report.

Critically, the neuromuscular cost of a set increases exponentially, not linearly, beyond 30% VL. In the Pareja-Blanco et al. (2020) study comparing 20% vs. 40% VL protocols over 8 weeks, both groups showed similar strength gains but the 40% VL group required 48 h longer to recover CMJ to baseline after each session. For athletes training twice or more per week per movement pattern, this recovery extension meaningfully restricts subsequent training quality.

Velocity Loss Thresholds by Training Goal

Different training objectives tolerate different levels of intra-set fatigue. Evidence-based VL thresholds:

These thresholds assume loads between 60–85% 1RM. At loads below 60% (primarily used for plyometric or speed-strength work), fatigue-induced VL is less reliable because initial velocities are near-maximal and small absolute changes in velocity represent large percentage losses without equivalent neuromuscular impairment.

Daily Readiness Testing with Reference Loads

Between-session fatigue accumulation can be detected before training begins using a standardised reference-load velocity test. The protocol:

1. Select a load corresponding to approximately 40–50% 1RM in the primary lift of the day's session (e.g., 70 kg back squat for an athlete whose 1RM is approximately 150 kg).
2. Perform 3 maximally-intentioned repetitions with full recovery between each.
3. Compare mean of the 3-rep MPV to the athlete's 7-day rolling baseline MPV at the same load.
4. If today's MPV is within ±5% of baseline: proceed as planned.
5. If MPV is 5–10% below baseline: reduce session volume by 20% and cap intensity at 80% of planned peak.
6. If MPV is more than 10% below baseline: consider a full deload session or complete rest.

This protocol takes under 5 minutes to execute, requires no blood sampling, and provides an individualised load prescription that accounts for day-to-day variability in readiness — variability that RPE-based or percentage-of-1RM-based programming cannot capture.

Balsalobre-Fernandez et al. (2017) validated this approach over a 12-week training block, finding that athletes whose sessions were auto-regulated via reference-load velocity testing achieved 11% greater 1RM improvements than a matched group following a traditional periodised programme — primarily because their highest-quality sessions were not under-prescribed and their lowest-readiness sessions were not over-prescribed.

Jump Testing as a Neuromuscular Readiness Proxy

For sports and phases where bar-velocity testing is impractical, the countermovement jump (CMJ) provides an equivalent readiness signal. CMJ height correlates strongly with neuromuscular fatigue state (r = 0.82–0.91 across power-sport populations) and takes under 2 minutes to administer.

Key CMJ readiness norms (based on Gathercole et al., 2015 and Claudino et al., 2017 meta-analyses):

• Morning CMJ within 3% of 7-day baseline: full training readiness.
• Morning CMJ 3–7% below baseline: moderate fatigue; reduce training volume or intensity by 15–25%.
• Morning CMJ more than 7% below baseline: high fatigue accumulation; consider active recovery only or rest.

CMJ flight time: contraction time ratio (FT:CT) adds a second dimension — it captures not just the height achieved but the relative speed of the take-off movement. A reduced FT:CT indicates slower force production, which is more sensitive to central fatigue (CNS-driven) than peripheral fatigue (metabolic). Central fatigue typically requires longer recovery windows (48–72 h) and responds less to nutritional interventions than peripheral fatigue.

Practical Fatigue Detection Protocols

Three operationally viable VBT-based fatigue detection protocols, ordered by complexity and information density:

Protocol 1 — Set Termination Alarm (Basic)

Set a target VL% alarm on the sensor (e.g., 20% for power, 30% for hypertrophy). Terminate the set when the alarm sounds. Record total reps completed per set — declining reps at the same load over days indicates accumulated fatigue.

Protocol 2 — Daily Reference Load Test (Intermediate)

Administer the 3-rep MPV test at the reference load described above before each training session. Log MPV, compare to rolling baseline, adjust session accordingly. Review weekly trends to identify accumulated fatigue across a training block.

Protocol 3 — Dual-Metric Battery (Advanced)

Combine a morning CMJ test (3 jumps, best height recorded) with the reference-load MPV test pre-session. Use a decision matrix: if both metrics are within 5% of baseline, proceed at full load; if either exceeds a 10% decrement, apply a full load reduction or postpone to the following day. This dual-metric approach reduces false positives from single-metric fluctuations driven by non-fatigue factors (hydration, sleep, mood) and provides complementary information about central vs. peripheral fatigue status.

Limitations and Confounders

Velocity-based fatigue detection requires methodological consistency to produce reliable signals. Key confounders:

• Motivational variability: MPV is only a valid fatigue indicator when athletes apply maximum concentric intent on every repetition. Submaximal effort produces velocity reductions that mimic fatigue. Verbal encouragement and real-time velocity feedback (showing the athlete their last rep) increase intent consistency.
• Sensor placement and calibration: Bar-mounted IMUs and linear position transducers give different absolute MPV values. Readiness decisions must use data from the same sensor across sessions; switching equipment mid-block invalidates the baseline comparison.
• Load-specific velocity profiles: Each exercise has a unique load-velocity relationship. A 20% VL at 60% 1RM in the squat represents a different physiological stimulus than 20% VL at 80% 1RM. Threshold application should be exercise- and load-specific, not applied universally across a session.
• Fatigue type mismatch: VBT detects peripheral neuromuscular fatigue with high sensitivity but is less sensitive to central fatigue, accumulated training monotony, or systemic stress (illness, life stress, poor sleep). A multi-metric approach (HRV + CMJ + MPV) provides a more complete picture.