2020 — Optimising the Front Foot Contact Phase of the Cricket Fast Bowling Action

Research theme

The peer-reviewed publication of the PhD’s whole-body model. Two aims: (a) prove the 16-segment torque-driven model is accurate enough — kinematically and kinetically — to be trusted, and (b) optimise ball release speed for one bowler and see what technique the optimiser chooses.

The methodological point the authors press hardest is evaluation. They quote King et al. (2006): if simulation research is to be “scientific”, models must be evaluated against both kinematics and kinetics. Many published simulation models never establish their accuracy at all.

Method, plainly: a 16-segment planar torque-driven simulation model of the front foot contact phase (front foot landing → ball release, ~0.1 s), built in AUTOLEV. 14 rigid segments plus wobbling masses (shanks, thighs, trunk) connected by non-linear spring-dampers; two massless segments (pelvis and shoulder girdle) with variable length and orientation allowing non-coincident hip and shoulder joint centres; trunk length varied to represent side-flexion; three spring-damper ground contact points on the front foot (heel, MTP, toe); the ball attached to the hand by a viscoelastic spring for smooth release. Torque generators (contractile + series elastic) at both shoulders, both hips, and the front knee, ankle and MTP plus bowling elbow and wrist; rear MTP/ankle/knee and non-bowling elbow angle-driven.

Data: one bowler, 12 maximal-effort good-length deliveries, 18 MX13 Vicon cameras at 300 Hz in a 7×3×3 m volume, Kistler force plate at 1800 Hz, 50 retroreflective markers, 95 anthropometric measurements, and Con-Trex MJ isovelocity dynamometer strength testing at the non-dominant ankle, knee and hip and the bowling shoulder.

What they measured

• How fast the ball leaves the hand (ball release velocity, m/s).
• How bent the front ankle and knee are through the phase (front ankle and front knee joint angle time histories).
• How the front hip behaves (front hip angle — extension then flexion).
• How far the bowler folds forward (trunk orientation angle time history).
• When the bowling arm starts to come over (bowling shoulder extension timing — onset of arm circumduction).
• When the back leg folds up (rear hip flexion timing and rate).
• How hard and in which direction the front foot hits the ground (horizontal and vertical ground reaction force time histories).
• Where the bowler’s mass centre is and how fast it is moving (centre of mass position and velocity).
• How long the phase lasts (duration front foot contact → ball release).
• How strong each joint is (subject-specific nine-parameter maximal voluntary torque functions).

Findings

Causal within the model, single bowler.

1. Model accuracy — the headline methodological result. Overall objective score 3.99% (reported as 4.0%). Components (Table 1): force 9.59%, centre of mass 0.06%, orientation 0.67°, ball velocity 0.03%, time 0.19%, joint angles 1.81°. The kinematic-only difference was 0.9%. No anatomical-limit penalties incurred. The authors argue the 9.6% kinetic difference is “reasonable agreement in a model which incorporates pin joints” (Allen et al., 2012).

2. Optimising the torque activation timings gave +9.8% — 38.8 m/s vs 35.3 m/s, a gain of 3.5 m/s. The initial pose at front foot contact was held identical to the matched simulation; only the 91 torque activation parameters were varied.

3. What the optimiser chose:

   • Front ankle and knee remained more extended throughout the phase.
   • Front hip stayed more extended before flexing, allowing more trunk flexion.
   • Bowling shoulder extension occurred later — the arm was delayed, starting only after trunk flexion had begun.
   • Rear hip flexion occurred earlier and faster.
   • Trunk flexion was initially delayed but ultimately went further.

4. The optimised technique raised horizontal ground reaction force while vertical stayed similar. The authors read this as evidence that the horizontal component of front-foot ground reaction force matters more to ball speed than the vertical component, consistent with Portus et al. (2004) (faster bowlers had higher peak horizontal GRF) and King et al. (2016) (faster bowlers had greater horizontal impulses).

5. A sanity check on strength calibration. The optimised 38.8 m/s sits comfortably inside the range observed in elite bowlers (32.8–39.7 m/s, Worthington et al., 2013a). The authors argue this shows the subject-specific strength estimates are appropriate: too high and the optimum would be implausible; too low and the evaluation match would have been poor.

6. Trunk flexion is a mechanism, not muscular work. The optimised solution produced more trunk flexion via straighter front-leg kinematics, which brake the lower body more efficiently and increase angular momentum. This supports Worthington et al. (2013a)’s proposal that trunk flexion during this phase is a mechanism of the action rather than a product of torso muscular effort.

7. A proposed relationship, not yet tested: the authors speculate that there is a relationship between the amount of trunk flexion and the delay in the bowling arm which maximises performance — if the arm is delayed, the trunk can flex further while still releasing towards the target — and call for investigation.

8. Rear hip flexion — a new claim. Rear leg motion has not previously been linked to ball speed, though coaching literature advocates it. The authors propose that earlier rear hip flexion reduces the moment of inertia of the body about the front hip, maximising trunk flexion. They call for future research.

9. An important honest limitation, quoted in full. The optimum is a single set of activation parameters, and the authors quote Yeadon’s landscape metaphor: “an optimisation routine may find the top of a pinnacle which stands on a narrow base high above the surrounding terrain. Even if this is the global optimum it is a summit that should not be attempted, since any small location error will land on the low terrain.” They note elite performers likely develop techniques that are insensitive to perturbation, and that robustness was not assessed.

What a coach should look for on video

Cue 1 — Front ankle and knee stay extended

• The cue: Not just the knee — the whole front leg. Watch the ankle too.
• Camera view + frame: Side-on, scrub frame by frame from front foot contact to ball release.
• What “good” looks like: Ankle and knee both holding extension through the phase; the leg working as one strut.
• What the fault looks like: The ankle collapsing (heel dropping / shin rotating forward over the foot) even when the knee looks reasonable. The knee flexing after landing, with head and hip height dropping.
• Why it matters: This is where most of the 3.5 m/s came from, and it is the change that produced the higher horizontal ground reaction force — i.e. more effective braking and momentum conversion.

Cue 2 — Trunk flexion: late start, big finish

• The cue: The timing as well as the amount of the forward fold.
• Camera view + frame: Side-on, front foot contact → ball release.
• What “good” looks like: The trunk staying up briefly after landing (the front hip staying extended), then folding further than the bowler’s habit.
• What the fault looks like: Folding immediately on landing and running out of range before the arm arrives.
• Why it matters: The optimised solution delayed the onset and increased the magnitude. The delay is attributed to changed proximal-to-distal sequencing when the front leg is straighter.

Cue 3 — Bowling arm starts after the trunk does

• The cue: The relative order of two events: trunk starting to fold, and the bowling arm starting to come over.
• Camera view + frame: Side-on, from front foot contact; find the first frame of trunk flexion and the first frame of arm circumduction.
• What “good” looks like: Trunk first, arm second. In the optimised simulation, “the optimised solution delays arm circumduction to occur after trunk flexion has started”.
• What the fault looks like: The arm leading the trunk.
• Why it matters: Tyson (1976) and Worthington et al. (2013a) identified arm delay as a key explainer of ball-speed variance across elite bowlers; this paper gives the mechanism — the delay creates room for the trunk to keep folding while the ball still goes to the right length.

Cue 4 — Back leg folds up early

• The cue: When the trailing (back) leg starts to fold at the hip.
• Camera view + frame: Side-on, front foot contact → ball release.
• What “good” looks like: The rear hip flexing earlier and faster than the bowler’s habit — the back leg gathering in rather than trailing out behind.
• What the fault looks like: The back leg left dangling behind through the phase.
• Why it matters: Proposed to reduce the body’s moment of inertia about the front hip, maximising trunk flexion. Flag this cue as provisional — the authors state rear leg motion “has not previously been linked to ball release speed” and explicitly call for future research. See the contradiction note below: the direction of this cue reverses in the 2023 group study.

Cue 5 (interpretive) — front-foot load: forwards, not just down

• The cue: How the bowler drives against the ground at front foot contact.
• Camera view + frame: Side-on. You cannot measure force on video, but you can see the consequence: a braced leg braking hard against forward travel vs a leg absorbing downwards.
• What “good” looks like: A firm, forward-resisting front leg.
• What the fault looks like: A cushioning, vertical-only landing.
• Why it matters: The optimised solution raised horizontal GRF with vertical unchanged; the authors conclude the horizontal component is the one that matters for speed. Caution: the force component was the model’s least accurate (9.59% difference), and see the loading tension below.

Caveats and limits

• n = 1. One male bowler, 18 years, 85.0 kg, 1.94 m, ECB elite fast bowling group. Same bowler as the thesis and the 2015/2017 conference papers.
• Simulation, not intervention. The +9.8% is what the model says is available.
• Planar (2D) model. Non-planar pelvis and torso rotations are driven as Fourier-series functions of trunk angle. The authors state the resulting limitation explicitly: these non-planar rotations cannot be perturbed, because there is no way to check whether new rotations would be feasible or within the bowler’s strength.
• Front foot contact phase only. The run-up, back foot contact and delivery stride are not modelled.
• Kinetic accuracy is the weak point — 9.59% force difference vs 0.9% kinematic. Weight the ground-reaction-force conclusions accordingly.
• Robustness untested — the “pinnacle” caveat above. The optimum may be a technique no human could reliably reproduce.
• The bowling shoulder torque generator is a constant, not a proper torque–angle/torque–angular-velocity function, because the planar bowling shoulder angle falls outside the range testable on a dynamometer (real bowling involves internal/external rotation and abduction/adduction). The bowling elbow has no active torque generator at all — it sits at its anatomical limit through the phase and is restrained passively.
• Male, elite, adolescent.

Relationship to other Felton work

• This is the peer-reviewed publication of PhD Chapter 9’s first optimisation Felton 2015 — PhD thesis: factors limiting fast bowling and of the conference papers Felton 2015 — optimising fast bowling performance and Felton 2017 — optimising individual performance. It reports the same 9.8% and the same 3.99% match score.
• The model built and validated here is used, essentially unchanged, in Felton 2023 — optimal initial position and technique for the front foot contact phase, Felton 2024 — effect of increasing isometric strength on technique and Felton 2025 — effect of increased strength on ball release speed.
• Cites Felton & King (2016) (the elbow paper) for the passive viscoelastic element used at the bowling elbow, and Felton, Yeadon & King (2019) for the non-coincident-joint-centre method.

CONTRADICTION: (rear leg timing reverses) This paper’s optimised technique had earlier and faster rear hip flexion, and proposes this reduces the moment of inertia about the front hip and thereby maximises trunk flexion. The 2023 J Biomechanics paper — same model, ten bowlers — found the optimised techniques had delayed rear hip flexion, and proposes that delaying it delays trunk flexion until after front foot contact, “providing a more efficient transfer of momentum through the kinetic chain”. These are opposite directions for the same joint from the same laboratory using the same model, and the 2023 paper does not acknowledge the reversal. A coach should not currently cue rear leg timing in either direction on the basis of this cluster.

TENSION: (the 22% went missing) the conference versions of this work (2015, 2017) and the thesis report a 21.5–22% gain when the landing position was also optimised. This journal paper reports only the 9.8% movement-only optimisation and states that varying the initial position of the bowling arm “was outside the scope of this paper”, relegating it to future work. The largest claim in the earlier conference literature was not carried into the peer-reviewed journal record for this bowler.

TENSION: (does optimal technique raise or lower front-foot load?) this paper’s optimisation increased peak horizontal ground reaction force. The thesis’s initial-configuration optimisation and the 2023 group study both found optimised technique lowered peak forces and loading rates. The resolution appears to be that changing only the movement raises horizontal braking force, while also changing the landing position (longer stride, straighter knee at landing) lowers peak load — but the cluster never states this reconciliation, and a coach reading only one paper would draw opposite injury conclusions.