Source note: full accepted manuscript read, including Table 2 and all reported SPM cluster results. All numbers below are transcribed directly.
Research theme
Almost all cricket biomechanics research, and almost all winter training, happens indoors on artificial turf. Almost all cricket is played on natural grass. Nobody had checked whether the action is the same on both. This matters twice over: for coaches (does net footage transfer to the middle?) and for the entire research literature (are the artificial-surface studies that underpin coaching even valid?). Eight sub-elite male fast bowlers (University Centre of Cricketing Excellence, comparable to English County 2nd XI; age 19.4 ± 1.3 yr, height 1.83 ± 0.09 m, mass 77.4 ± 9.1 kg) bowled a minimum of six maximal good-length deliveries outdoors on a natural grass pitch on day one, then indoors on artificial turf the next day — an 11 mm pile monofilament polyethylene carpet over a 15 mm bound rubber shockpad over concrete, with room for a full run-up. Both sessions captured with an 18-camera Vicon system at 300 Hz, 47 markers, the same researcher placing markers both days. Analysis: paired t-tests on discrete variables plus statistical parametric mapping (SPM) on the whole time-normalised joint-angle waveforms, so a difference anywhere in the phase would be detected, not just at key instants. Exploratory; no hypothesis posed. Measured comparison, within-subject.
What they measured
- Ball release velocity, run-up velocity
- Shoe slip distance at back foot contact and front foot contact — how far the shoe slides along the ground after landing
- Phase durations: back foot contact → ball release, back foot contact → front foot contact, front foot contact → ball release
- Whole-body and pelvic centre-of-mass acceleration through both phases
- Performance-related joint angles across the whole waveform: bowling shoulder flexion, front knee flexion, thoracolumbar flexion
- Injury-related joint angles across the whole waveform: rear hip flexion, lumbopelvic flexion, rear knee flexion, thoracolumbar side flexion, thoracolumbar rotation, front hip flexion, pelvic tilt, lumbopelvic side flexion, pelvic twist
- Footwear differed with surface by design, matching real practice: 6 mm spiked cricket shoes on grass, non-spiked athletic trainers indoors
Findings
The one significant difference: the feet slide more indoors
- Back foot contact slip distance: 7.1 ± 3.5 cm on grass vs 8.8 ± 3.6 cm on artificial; mean difference −1.7 ± 0.8 cm, p = 0.001, d = 2.12 (95% CI 0.81–3.39) — a very large effect.
- Front foot contact slip distance: 6.6 ± 2.5 cm on grass vs 8.7 ± 2.0 cm on artificial; mean difference −2.1 ± 1.8 cm, p = 0.013, d = 1.16 (95% CI 0.23–2.06) — large.
- Attributed to traction: non-spiked trainers on carpet grip less than 6 mm spikes in turf.
Everything else: no difference 4. Ball release velocity: 30.2 ± 2.2 m/s (grass) vs 30.4 ± 2.2 m/s (artificial), p = 0.438, d = 0.29. Bowlers were, if anything, a fraction quicker indoors. 5. Run-up velocity: 5.4 ± 0.4 vs 5.4 ± 0.3 m/s, p = 0.282, d = 0.41. 6. Back foot contact → ball release: 0.328 ± 0.032 vs 0.327 ± 0.038 s, p = 0.813, d = 0.09. 7. Back foot contact → front foot contact: 0.218 ± 0.033 vs 0.215 ± 0.034 s, p = 0.336. 8. Front foot contact → ball release: 0.110 ± 0.008 vs 0.111 ± 0.009 s, p = 0.331. 9. Whole-body COM acceleration BFC→FFC: −3.7 ± 0.6 vs −3.8 ± 0.7 m/s²; FFC→BR: −11.0 ± 2.5 vs −11.4 ± 1.8 m/s². Pelvic COM acceleration BFC→FFC: 1.2 ± 1.0 vs 1.0 ± 1.3; FFC→BR: −25.5 ± 4.3 vs −26.4 ± 4.2. All non-significant (p = 0.28–0.81 across the whole discrete set, d = 0.09–0.41). 10. SPM found no statistically significant kinematic difference anywhere, in any joint angle, at any point in either phase (p > 0.05). This is the headline. Not “no difference at the key instants” — no difference across the entire time-normalised waveform.
Exploratory sub-threshold trends (NOT statistically significant — reported by the authors where the 95% CI of the mean difference excluded zero) 11. Back foot contact phase, artificial vs grass: less front knee flexion (99–100% of phase, peak mean difference 4°); greater ipsilateral pelvic twist (73–100%, peak 7°); greater contralateral lumbopelvic side flexion (43–48%, peak 2°). 12. Front foot contact phase, artificial vs grass: greater rear knee flexion (0–21%, peak 3°); less rear hip flexion (50–87%, peak 2°); greater ipsilateral pelvic twist (0–49%, peak 7°); greater contralateral lumbopelvic side flexion (92–100%, peak 4°); less ipsilateral thoracolumbar rotation (85–100%, peak 2°). 13. The largest of these is the 7° pelvic twist, present in late back-foot and early front-foot phases. The authors speculate this is the adaptation that absorbs the extra slip and keeps the business end of the action (thoracolumbar flexion, shoulder extension at front foot contact) unchanged — i.e. the bowler reorganises upstream so that ball speed is preserved.
The injury-risk verdict 14. The two best predictors of prospective lumbar bone stress injury (Alway et al. 2021 — rear hip flexion at back foot contact, and lumbopelvic flexion/extension at front foot contact) showed no difference between surfaces. The one rear-hip trend occurred at 50–87% of the front foot contact phase, i.e. long after back foot contact, and the authors judge it “unlikely to influence lumbar bone stress injury risk.” 15. Lumbopelvic flexion, front hip flexion, pelvic tilt and thoracolumbar side flexion: no differences at all. Other injury-relevant variables (thoracolumbar rotation, lumbopelvic side flexion, rear knee flexion, pelvic twist) showed only sub-threshold trends, and at phase timings different from those previously linked to injury. 16. Practical consequence stated by the authors: artificial-surface deliveries “should be included as part of any workload monitoring” — they load the back the same way grass deliveries do.
What a coach should look for on video
The practical verdict a coach actually needs: on this evidence, net-session footage on artificial turf transfers to the middle. Everything you would coach off video — front knee, trunk flexion, arm position, hip/shoulder alignment, front foot contact timing — was statistically indistinguishable between surfaces, across the entire movement, in the same eight bowlers two days apart. Ball speed was also unchanged. You do not need to discount your indoor footage.
Cue 1 — Foot slide at landing (the one thing that genuinely changes indoors)
- Camera view + frame: Side-on, low, tight on the feet. Scrub from the frame the back foot lands to the frame it stops moving; repeat for the front foot.
- What to expect: On grass in spikes, about 7 cm of back-foot slide and 6–7 cm of front-foot slide. Indoors in trainers on carpet, about 9 cm at each — roughly 1.5–2 cm more, a very large effect (d = 2.12 and d = 1.16).
- What the fault looks like: Slide markedly beyond that indoors, or a visible skid-and-recover, usually means the wrong footwear rather than a technique problem.
- Why it matters: This is the only measured surface effect. Note the confound the study deliberately accepted: footwear changed with surface, so this is a shoe-and-surface effect, not a surface effect. The fix is footwear, not action.
Cue 2 — Hip/pelvis alignment at back foot contact, if you are comparing indoor to outdoor footage of the same bowler
- Camera view + frame: Behind-the-arm or high overhead, from back foot landing through to front foot landing.
- What to expect: Up to about 7° more ipsilateral pelvic twist indoors — the pelvis a touch more turned towards the bowling arm side.
- Why it matters: This is the largest trend in the paper and it is not statistically significant (n = 8, underpowered). Do not chase it. Its value is knowing that a small pelvis-alignment difference between your indoor and outdoor clips of the same bowler is probably the surface, not a technique regression.
Cue 3 — Workload counting
- Not a video cue. Because the loading kinematics are the same, indoor artificial-net deliveries count exactly the same as match deliveries for back-injury workload monitoring. Do not treat winter net volume as “free.”
Cues this paper explicitly does NOT support: any claim that artificial nets push a bowler towards or away from injury-risk positions, or that they cost ball speed. Both were tested and both came back null.
Caveats and limits
- n = 8, sub-elite (County 2nd XI standard), all male. The authors state plainly that the study is “likely underpowered” and that “results should be interpreted with caution.” A null result in an underpowered study is weak evidence of no effect — this paper cannot rule out real differences smaller than roughly a large effect size. The honest reading is: no large surface effect exists, and the small trends observed are of no coaching consequence.
- Sample size was limited by outdoor-facility availability, not by design.
- Footwear was confounded with surface by design (6 mm spikes on grass, non-spiked trainers indoors). This mirrors real practice, but it means the slip finding cannot be attributed to the surface alone.
- Two examples of surfaces, not two categories. The authors are explicit: natural turf varies with soil type, grass species, moisture and management; artificial varies with carpet, pile length, shockpad and sub-base. Results “should be limited to the surfaces tested.” A worn club astro over tarmac is not the surface tested here.
- The two sessions were on different days, so markers were re-applied — inter-session marker placement variability could mask small real differences (the authors used the same skilled researcher and repeated static/dynamic calibrations to mitigate).
- No ground reaction forces were measured. The authors flag that shoe-surface differences may well change vertical and horizontal GRF, which are known to be high at front foot contact. GRFs are not linked to lumbar bone stress injury, but may matter for lower-limb injury. So “no kinematic difference” does not license “no kinetic difference.”
- No correction for multiple comparisons (exploratory design).
- The exploratory sub-threshold trends in findings 11–12 are not significant and are reported by the authors only as candidates for future work. They should not be quoted as findings.
Relationship to other Felton work
- Directly validates the surface used in the rest of this cluster: Felton et al. (2015, 2019) and Bull et al. (2026) were all collected indoors on artificial turf at the ECB National Cricket Performance Centre. This paper is the reason those results can be applied to grass cricket. That makes it load-bearing for the whole Felton fast-bowling corpus.
- Uses the Worthington et al. (2013) marker set, joint-angle conventions and event definitions, so it is methodologically continuous with everything else here.
- Tests the injury variables from Alway, Felton, Brooke-Wavell, Peirce & King (2021), “Cricket fast bowling technique and lumbar bone stress injury” (the 88%-classification model), which sits in the injury cluster.
- Cites Felton, Shine, Yeadon & King (2023) simulation work on optimal front-foot-contact technique.
- No contradictions with other Felton work. This paper’s function is corroborative: it removes a threat to the validity of the rest of the programme.
- TENSION (mild, methodological): the paper’s conclusion that “previous research utilising artificial surfaces in fast bowling research is likely to be valid” is a strong claim resting on a null result from n = 8. It is the right conclusion on the available evidence, but it is a licence granted by an underpowered study, and the authors themselves call for corroboration.