Research theme
This is the pivotal paper in the cluster — the one where the individual-specific programme is finally run at scale, to ask whether optimal technique looks the same across bowlers.
The setup is a genuine methodological standoff. Group studies give “on average” statements that Glazier and Mehdizadeh (2019) argued cannot be applied to individuals — the paper’s own example is that run-up speed relates to ball speed linearly within a group but quadratically within individuals, and that numerous elite international bowlers do not display all the supposedly optimal characteristics. But individual-specific simulation studies (Felton’s own 2020 paper) have the mirror-image problem: you cannot know whether one bowler’s optimum is suitable, or even attainable, for anyone else. The way out: run the individual-specific method on many individuals and look for trends. That had never been done with forward-dynamics models because of the computational cost.
Method, plainly: the same 16-segment planar torque-driven simulation model as Felton et al. (2020), customised and evaluated separately for ten elite male fast bowlers, then both the landing position and the subsequent movement pattern optimised for each — 112 parameters per bowler (98 torque activation parameters across seven joints; 12 initial joint angles and angular velocities at front ankle, knee, hip, rear hip, front shoulder and bowling shoulder; 2 initial trunk orientation angle and angular velocity), run on Loughborough’s high-performance computing system. Then paired t-tests on discrete parameters (SPSS) and SPM1D on continuous time histories.
What they measured
- How fast the ball leaves the hand (ball release speed, m/s).
- How long the phase lasts (front foot contact → ball release, ms).
- How hard the front foot hits the ground, forwards and downwards (peak horizontal braking and vertical ground reaction force, bodyweights).
- How fast that force arrives (average horizontal braking and vertical loading rates, BW/s).
- The total forwards and downwards force applied over the phase (horizontal braking and vertical impulse, BW·s).
- Nine joint angle time histories: trunk orientation, front ankle, front knee, front hip, rear hip, front shoulder, bowling shoulder, bowling elbow, bowling wrist.
- Six joint torque time histories: front ankle, front knee, front hip, rear hip, front shoulder, bowling shoulder.
Findings
Causal within the model; the sample is ten bowlers, each with their own model.
Model evaluation across ten bowlers. Mean objective difference 5.0% (SD 0.8; range 4.0–6.5): kinematic difference 1.9% (SD 0.5; range 0.9–2.7) and force difference 11.3% (SD 1.9; range 9.2–15.0). All ten models were passed for optimisation.
Optimised techniques were 13.5% faster. Ball release speed rose by 4.8 ± 1.3 m/s (13.5 ± 4.1%), from 35.9 ± 1.5 m/s to 40.7 ± 1.6 m/s, p < 0.001, effect size 3.1 — optimised speeds ranged 37.8 to 42.9 m/s. Phase duration was unchanged (101 ± 8.4 → 102 ± 6.0 ms, p = 0.888).
Kinetic results (Table 1), current → optimised:
- Peak horizontal braking force: 4.13 ± 0.7 → 3.94 ± 0.4 BW, p = 0.394 (not significant), ES 0.33
- Peak vertical force: 6.03 ± 1.3 → 5.68 ± 0.8 BW, p = 0.093 (not significant), ES 0.33
- Horizontal braking loading rate: 154 ± 39 → 128 ± 45 BW/s, p = 0.008, ES 0.60
- Vertical loading rate: 249 ± 118 → 172 ± 70 BW/s, p = 0.022, ES 0.70
- Horizontal braking impulse: 0.15 ± 0.05 → 0.15 ± 0.05 BW·s, p = 0.882
- Vertical impulse: 0.31 ± 0.05 → 0.29 ± 0.06 BW·s, p = 0.383
So: loading rates fell significantly; peak forces and impulses did not. The abstract and discussion state peak forces were also lower; the statistics in Table 1, and the SPM1D continuous analysis in Fig. 1, do not support that — see Caveats below.
Commonalities in the landing position (all ten bowlers). At front foot contact the optimised techniques had:
- More extended front knees
- More flexed front shoulder (front arm higher)
- More flexed bowling shoulder (arm further back / more delayed)
Commonalities in the subsequent movement (all ten bowlers). Delays to the onset of:
- Trunk flexion
- Front arm shoulder extension
- Bowling arm shoulder extension
- Wrist flexion
…followed by increased magnitude of trunk, rear leg and both arms moving towards the target. The front leg stayed more extended throughout.
Joint torque commonalities: lower front hip extensor torque, lower front shoulder flexor torque, greater bowling shoulder extensor torque. Notably there were no significant differences in front knee joint torque — despite every optimised technique having a more extended knee. And not all optimised techniques used a fully braced knee throughout. The authors infer that factors outside the study (e.g. centre-of-mass velocity at front foot contact) influence front knee kinematics.
A genuinely new observation about the front arm. The optimised front shoulder starts more flexed (higher), which lowers front shoulder flexor torque in the first half of the phase, and then extends further and faster. The authors’ proposed mechanism: a high front arm helps align the shoulder girdle and stabilise the upper body, aiding the bowling arm’s delay — but it raises the upper body’s rotational inertia, so the front arm must then accelerate downwards to align with the torso and let the trunk flex. They state: “this is the first study to highlight this movement.”
The claim the paper makes for itself. From the conclusion: “This study has resolved the controversy on whether individual and group optimisation studies of fast bowling reflect underlying commonalities.” It does retain the individual caveat — whether a given bowler can adopt these characteristics depends on their organismic constraints (range of motion, strength, anthropometry) and self-organisation processes — and notes evidence that gender-related constraints may alter optimal technique for female bowlers (Felton et al., 2019a).
Applied use is confirmed: “The individual optimisations… have been used indicatively to support applied coaching.”
What a coach should look for on video
This is the strongest evidence base in the cluster for general coaching cues, because every cue below held for all ten elite bowlers, not one. But note that the targets are common while attainability is not.
Cue 1 — Front knee more extended at landing and held
- The cue: Front knee angle in the frame the foot lands, and whether it holds.
- Camera view + frame: Side-on, at front foot contact, then scrub to ball release.
- What “good” looks like: More extended at landing than the bowler’s habit, and staying more extended throughout. Important nuance: the paper explicitly notes not all optimised techniques used a fully braced (dead straight) knee — so “straighter than you currently are” is the coachable target, not “locked straight”.
- What the fault looks like: Landing with a soft knee that then collapses; head and hips sinking after contact.
- Why it matters: Present in all ten optimisations. Mechanism: brakes the lower half more efficiently and converts run-up linear momentum into angular momentum.
Cue 2 — Front arm high at landing, then pulled down hard
- The cue: Two things — where the non-bowling arm is when the foot lands, and how aggressively it comes down afterwards.
- Camera view + frame: Front-on for the height and alignment; side-on for the pull-down. Frames: front foot contact, then through to release.
- What “good” looks like: Front arm distinctly higher/more flexed at landing than the bowler’s habit; then a later but faster and further pull-down into the torso.
- What the fault looks like: A low or already-dropping front arm at landing; or a passive front arm that never actively accelerates down; or one that flies out sideways.
- Why it matters: Common to all ten bowlers, and this paper is the first to identify the mechanism (stabilise the shoulder girdle to permit the bowling arm’s delay, then unload the rotational inertia so the trunk can flex). It also gives coaches a mechanical reason for a cue they already used on instinct.
Cue 3 — Bowling arm further back at landing, and starting later
- The cue: Bowling arm position at front foot contact, and the frame at which circumduction begins.
- Camera view + frame: Side-on, at front foot contact and then frame by frame.
- What “good” looks like: More flexed bowling shoulder (arm further back) at landing than the bowler’s habit; extension starting later.
- What the fault looks like: Arm already coming over at landing.
- Why it matters: Common to all ten. A more delayed arm permits more trunk flexion while still delivering to the target. Note this comes with greater bowling shoulder extensor torque — the delay is paid for by working the shoulder harder later.
Cue 4 — Trunk flexion: delayed onset, larger total
- The cue: When the fold starts and how far it goes.
- Camera view + frame: Side-on, front foot contact → ball release.
- What “good” looks like: Trunk staying up briefly after landing, then folding further than the bowler’s habit.
- What the fault looks like: Folding immediately at landing.
- Why it matters: Common to all ten, and achieved with lower front hip extensor torque — i.e. it is a mechanism of the action, not a matter of muscling the torso down.
Cue 5 — Wrist flexion delayed
- The cue: When the bowling wrist starts to flex.
- Camera view + frame: Behind-the-arm or side-on, high frame rate, last few frames before release.
- What “good” looks like: Wrist held back longer before snapping through.
- What the fault looks like: Early wrist flexion.
- Why it matters: Common to all ten. Caveat: the model constrained the bowling wrist to match current performance in earlier work and the authors list the wrist among joints whose torque was not investigated — treat this as the weakest of the five cues.
Cue 6 — A load-management observation, with a caveat
- The cue: How abruptly the front foot loads, rather than how hard.
- Camera view + frame: Side-on at high frame rate — count the frames between first foot contact and the deepest point of the landing.
- What “good” looks like: Load arriving less abruptly.
- What the fault looks like: An instantaneous, jarring landing.
- Why it matters: The optimised (faster) techniques had significantly lower loading rates in both directions (horizontal −25.5 BW/s, p = 0.008; vertical −76.7 BW/s, p = 0.022). This is the cluster’s best support for the message that faster technique is not automatically higher-load technique. Caveat: peak forces and impulses did NOT change significantly — do not tell a bowler this technique reduces the force through their back.
What is NOT supported: the rear leg. The paper reports delayed rear hip flexion in the optimised techniques but attributes it to changed proximal-to-distal sequencing (there were no significant rear hip torque differences), and calls the function of rear hip flexion “uncertain”. The direction found here differs from Felton et al. (2020), a single-bowler study; both papers call the function of rear hip flexion uncertain. Do not cue it.
Caveats and limits
- n = 10. Ten male ECB elite fast bowling group members (age 20.7 ± 2.4 y; height 1.91 ± 0.08 m; mass 86.9 ± 8.5 kg). The authors call the statistical power “limited despite a ten-fold increase on any previous theoretical approach”.
- Simulation, not intervention. Nobody was coached into the optimised technique and re-measured. The 13.5% is what the models say is available.
- Planar (2D) model — non-planar torso and pelvis rotations cannot be investigated.
- Front foot contact phase only. Initial centre-of-mass position and velocity were taken from the bowler’s current performance and not optimised — so nothing here tells you about run-up speed, and the authors flag that front knee kinematics may be limited by exactly that unmodelled variable.
- Optimum is a single parameter set, robustness to perturbation untested (the “pinnacle” problem carried over from 2020).
- Some joints excluded: front MTP, bowling elbow and bowling wrist torques were not investigated (contribution minimal or constrained to current performance).
- Elite males only. The authors explicitly warn that gender-related organismic constraints may alter optimal technique for female bowlers, and the whole conclusion is caveated: “the ability of an individual to adopt these depends on their organismic constraints… and self-organisational processes”.
- A compound claim in the abstract and discussion. Both state the optimised techniques resulted in “lower peak ground reaction forces and loading rates”. In Table 1, the loading rates differ significantly (horizontal braking loading rate p = 0.008; vertical loading rate p = 0.022) but the peak forces do not (peak horizontal braking force p = 0.394; peak vertical force p = 0.093). The paper’s own Results section reports that the continuous SPM1D analysis in Fig. 1 found the same thing: “no significantly different periods between the ground reaction time histories were observed.” Neither the discrete peaks nor the continuous analysis support a peak-force difference — only the loading rates do.
Relationship to other Felton work
- Uses the model developed and validated in Felton 2020 — optimising the front foot contact phase, applied to ten bowlers instead of one.
- The conference companion is Felton 2023 — investigating commonalities of optimal technique in maximal effort movements (19th ISCSB, Kyoto), same data, framed as a methodological case study.
- The baseline optimisations here (40.7 ± 1.6 m/s) are the reference condition for Felton 2024 — effect of increasing isometric strength on technique and Felton 2025 — effect of increased strength on ball release speed.
Superseded: (the individual-specific message) Earlier work — especially Felton 2017 — optimising individual performance — argues that group research “is not suitable to understand the changes required to optimise an individual’s performance” and that the point of the method is a bespoke prescription for one named athlete. This paper, running the same individual-specific method on ten bowlers instead of one, concludes it “has resolved the controversy on whether individual and group optimisation studies of fast bowling reflect underlying commonalities” and reports that the same characteristics emerged for all ten bowlers. The individuality that survives is about attainability and constraint, not about what the target technique is: the optimal shape is common across all ten; whether a given bowler’s range of motion, strength and anthropometry let them reach it is individual.
Superseded: (rear hip flexion timing) the 2020 single-bowler optimisation produced earlier rear hip flexion; the 2023 ten-bowler study produced delayed rear hip flexion. The 2023 paper cites the 2020 mechanism and proposes that delaying rear hip flexion delays upper trunk flexion for a more efficient transfer of momentum. Both papers state the function of rear hip flexion is uncertain. Felton 2020 — optimising the front foot contact phase — same model, same lab — found the optimised technique had earlier and faster rear hip flexion, with the proposed benefit of reducing the moment of inertia about the front hip.
Open question: (peak forces) the abstract states optimal technique “lowers peak ground reaction forces and loading rates.” In Table 1, the loading rates differ significantly (horizontal braking loading rate p = 0.008; vertical loading rate p = 0.022) but the peak forces do not (peak horizontal braking force p = 0.394; peak vertical force p = 0.093). Fig. 1, the continuous SPM1D waveform analysis of the same force curves, found the same result — the paper’s own Results section states “no significantly different periods between the ground reaction time histories were observed.” Neither the discrete peaks in Table 1 nor the continuous analysis in Fig. 1 supports a peak-force difference; only the loading rates do.
Superseded: (the 22% vs 13.5%) the earlier conference papers claimed a 22% gain from optimising the landing position for a single bowler. This group study, running the same both-position-and-movement optimisation, found 13.5 ± 4.1% across ten bowlers, with a maximum optimised speed of 42.9 m/s.