Why two channels instead of one score
Recent evidence separates two things a single "load" score blends.
Total match volume — not high-speed running specifically — tracks the MD+2 neuromuscular drop and the adductor-strength drop (Springham 2026).
Injury risk is speed-band specific: hamstring injuries are predominantly a sprint / high-speed mechanism, and the acute:chronic ratio of the 15–20 and >25 km/h bands predicts risk better than total distance (Xia 2026; Fousekis 2025). One number can't tell you whether a flagged player is
fatigued from volume or
exposed to a high-speed spike — and those need different actions.
Readiness channel (volume fatigue)
| Marker | Captures | Flag threshold |
| CMJ mRSI (force plate) | Neuromuscular readiness | drop > ~9% |
| Adductor squeeze (isometric) | Groin/adductor recovery & risk | drop > ~11% or pain |
| Soreness (Likert) | Perceived NM response | ≥ 3/5 (or Z > 1.2) |
Thresholds are the "meaningful change" values from Table 3 of Hader & Buchheit (2026): a change only counts once it exceeds the smallest worthwhile change plus the typical error (SWC + TE). The verdict uses a
confidence rule: one marker moving = monitor, two together = action. That is why the CMJ and adductor are read separately — the CMJ can say "recovered" while the adductor says "not yet".
Adductor squeeze — when to test
| Day | Window | Reads |
| MD+1 | ~24 h | the acute drop (passive, mainly starters) |
| MD-4 (= MD+3) | ~72 h | whether it recovered (full NM battery day) |
MD+2 is often a day off — and that is fine. A pain-free post-match drop normally recovers within 48–72 h, so the 24 h and 72 h checks are a better recovery window than a single MD+2 test. The most recent test drives the read. A drop that
persists at 72 h, or one with
pain, is the flag.
Injury-risk channel (high-speed exposure)
The usable signal is the
week-on-week change in high-speed running and sprint distance (Breed 2026: OR ~1.5–1.7). It is shown next to the
absolute weekly metres and is
low-certainty on purpose — the best multivariate hamstring model still leaves ~80% unexplained, so we steer on exposure, not a predictive score. A large rise from a low base is read with the absolute value, not as panic. Working bands: ≥ 100% = flag, 30–100% = monitor — calibrate on our own cohort.
Reading the verdicts
Each player gets
two badges, each with a reason. "Action" = multiple markers crossed threshold; "monitor" = one did; "clean" = none. Treat a readiness flag with same-day session-content changes (reduce neuromuscular load); treat an injury-risk flag with exposure management — not a generic "reduce load".
Data sources
GPS (StatSports, weekly HSR & sprint), force-plate CMJ and daily wellness/soreness are the club's real data
real. The
adductor squeeze is mock data in this build
demo — swap in the real isometric-adductor test once it is logged on MD+1 / MD-4.
References. Springham 2026 (JSCR) · Xia 2026 (Front Public Health) · Fousekis 2025 (Medicina) · Breed 2025 (Sports Med Open) · Hader & Buchheit 2026 (SPSR #307, Table 3) · Buchheit & Hader 2025 (SPSR #258). Working thresholds are defaults to be calibrated in-house.