2023 — Female Pace Bowling: Kinematic and Anthropometric Relationships with Ball Speed

Source note: full paper read, including Tables 1–3. All numbers below are transcribed directly.

Research theme

The peer-reviewed version of the 2022 BASES poster. The question: do the kinematic and anthropometric relationships with ball release speed that were established on male pace bowlers actually hold in female pace bowlers, or is female coaching pedagogy built on borrowed and possibly wrong assumptions? Eleven high-performance right-handed female pace bowlers — senior Irish International or Interprovincial squad members — bowled six maximal-effort good-length deliveries indoors on a standard-size pitch with a full run-up, using a 141.75 g women’s ball. Method was deliberately field-grade: two-dimensional, two iPhone 11s at 240 Hz on tripods at 0.91 m, 6 m either side of the crease and perpendicular to the plane of motion; radar gun 2 m behind the bowler’s stumps for ball speed; two pairs of Brower photocell timing gates 5 m apart at hip height for run-up speed; stadiometer, scales, and tape for anthropometry. Best three trials averaged (ICC 0.81–1.00, mean 0.92). Pearson correlations plus forward stepwise regression, with partial correlations controlling for anthropometry. Correlational — no intervention, no simulation.

What they measured

• Ball release speed (radar gun)
• Height, body mass, front leg length (ASIS to lateral malleolus), bowling arm length (acromion to fingertip)
• Run-up velocity (timing gates over the 5 m ending 0.5 m behind the crease)
• How straight the front knee is when the front foot lands (front knee angle at FFC)
• How far back the bowling arm still is at release (bowling shoulder angle at BR)
• How much the trunk folds forward between front foot landing and release (trunk flexion FFC→BR)

Findings

Sample

1. Age 22.3 ± 4.7 yr, height 1.68 ± 0.08 m, mass 73.0 ± 8.3 kg. Ball release speed 23.0 ± 1.8 m/s, range 20.3–26.4 m/s (roughly 73–95 km/h) — a club-to-international-development band, well below the 27.9 m/s elite female mean in Felton et al. (2019).
2. Descriptives: leg length 0.90 ± 0.08 m; arm length 0.71 ± 0.05 m; run-up velocity 5.05 ± 0.61 m/s (3.72–5.77); front knee angle at FFC 167.0 ± 2.9° (160.7–171.4); bowling shoulder angle at BR 180.3 ± 7.7° (172.7–194.4); trunk flexion FFC→BR 43.5 ± 8.5° (24.4–57.2).

Bivariate correlations with ball release speed (Table 2) 3. Bowling shoulder angle at BR: r = 0.95, 95% CI 0.82–0.99, p < 0.001. By far the strongest. 4. Height: r = 0.76, CI 0.29–0.93, p = 0.007. 5. Run-up velocity: r = 0.75, CI 0.22–0.94, p = 0.01. 6. Arm length: r = 0.61, CI 0.02–0.89, p = 0.05. 7. Front knee angle at FFC: r = 0.49, CI −0.16–0.84, p = 0.13 — not significant bivariately. 8. Trunk flexion FFC→BR: r = −0.19, CI −0.71–0.46, p = 0.57 — no relationship at all, and the sign is negative. 9. Body mass: r = 0.07, p = 0.85 — nothing. Leg length: r = 0.24, p = 0.48 — nothing.

Partial correlations (controlling for size) 10. Controlling for height: shoulder angle at BR r = 0.94, p < 0.001 (survives); front knee angle at FFC becomes significant, r = 0.68, CI 0.18–0.98, p = 0.04; run-up velocity collapses to r = 0.41, p = 0.28 (no longer significant); trunk flexion r = −0.05, p = 0.90. 11. Controlling for arm length: shoulder angle at BR r = 0.94, p < 0.001; front knee at FFC r = 0.67, p = 0.05; run-up r = 0.60, p = 0.09; trunk flexion r = −0.07, p = 0.87. 12. The run-up result is the interesting one: run-up speed correlates with ball speed largely because taller bowlers run in faster. Once height is held constant, the relationship disappears. The authors conclude “individual-specific optimal run-up speeds exist and are most likely based on height.”

Regression models (Table 3) 13. Best anthropometric model: height alone, 53% of variance (coefficient 17.379 m/s per m, CI 5.429–29.329, p = 0.010). 14. Best kinematic model: bowling shoulder angle at BR alone, 89% of variance (coefficient 0.224 m/s per degree, CI 0.168–0.279, p < 0.001). That is roughly 0.8 km/h of ball speed per extra degree of arm delay across the observed 22° spread. 15. A two-variable model (height + shoulder angle) reached 93% but was rejected because the height coefficient’s 95% CI included zero (p = 0.052). The accepted overall model is shoulder angle alone.

Interpretation offered by the authors 16. Arm delay is the mechanism: the arm further back at release “allows greater amounts of trunk flexion, while still allowing the arm to deliver the ball towards the intended target.” 17. But trunk flexion itself did not relate to speed here, unlike in men. The authors’ explanation: female bowlers have proportionally longer trunks (larger transverse moment of inertia, harder to flex), and may achieve arm delay via trunk rotation rather than trunk flexion — consistent with Felton et al. (2019)’s “more akin to throwing” conclusion. 18. Leg length did not predict speed in women, contrary to male findings. Body mass did not either — which matches male findings. 19. Explicit warning against over-reading the height result: “bigger is not continuously better, and there is a tipping point based on the ability to maintain the power to weight ratio.” 20. Explicit warning against circularity: the alignment with male findings “may occur due to female pace bowlers being coached based on male pace bowling philosophy, and not because this is the optimal method to generate BRS for female pace bowlers.”

What a coach should look for on video

Cue 1 — Arm delay at ball release (the highest-yield cue in this cluster)

• Camera view + frame: Side-on, phone on a tripod at about hip height (0.91 m in the study), roughly 6 m from the crease, square to the bowler’s plane of motion — alignment matters, this is a 2D angle. 240 fps. Pause on the exact frame the ball leaves the hand.
• What “good” looks like: The bowling arm still swept back behind the line of the upper trunk at release. In this sample the range was 172.7° to 194.4°, and the regression says each extra degree is worth about 0.22 m/s (≈0.8 km/h).
• What the fault looks like: Arm already vertical or past vertical relative to the trunk at release — the ball is being pushed out rather than slung. Often paired with the trunk already having folded forward and the bowler “reaching” for the target.
• Why it matters: r = 0.95; 89% of the ball-speed variance in this squad from this one frame. Nothing else in the female literature comes close.

Cue 2 — Front knee at front foot contact, judged against the bowler’s own height

• Camera view + frame: Side-on, pause on the frame the front foot first contacts the ground.
• What “good” looks like: Front knee close to straight. Mean here 167.0 ± 2.9°, i.e. within about 13–19° of straight, and the whole squad sat between 160.7° and 171.4°. Straighter went with faster once height was accounted for (r = 0.68, p = 0.04).
• What the fault looks like: Knee visibly bent at the moment of landing, before any braking has even happened.
• Why it matters: Real but conditional — it only shows up as significant after controlling for height, and it was not significant on the raw correlation (r = 0.49, p = 0.13). Compare like-for-like bowlers, not a 1.55 m bowler against a 1.80 m one.

Cue 3 — Run-up speed: compare a bowler to herself, not to the squad

• Camera view + frame: Side-on wide shot; or better, two timing gates 5 m apart ending ~0.5 m behind the crease, at hip height — the exact setup used here and cheap to replicate.
• What “good” looks like: Squad mean 5.05 ± 0.61 m/s, range 3.72–5.77 m/s. Faster is associated with faster ball (r = 0.75) — but that association is mostly height doing the work (r drops to 0.41, p = 0.28 once height is controlled).
• What the fault looks like: A short bowler being pushed to match a tall team-mate’s approach speed, arriving out of control.
• Why it matters: The paper’s clearest practical statement is that optimal run-up speed is individual and scales with height. Track a bowler’s own run-up speed against her own ball speed over time. Do not set a squad-wide target.

Cue 4 — Trunk flexion: a male cue that does NOT transfer

• Camera view + frame: Side-on, compare the front-foot-contact frame with the release frame; measure how far the chest folds over between the two.
• What this paper found: r = −0.19, p = 0.57. Nothing. Despite a huge spread in the squad (24.4° to 57.2°).
• Why it matters: Trunk flexion between front foot contact and release is one of the four pillars of the male model (Worthington et al. 2013). In these eleven women it carried zero information about ball speed. This is the single clearest instance in the cluster of a male cue failing to transfer. Coaching “get your chest over the front leg” as a speed intervention for a female bowler is not supported.

Cue 5 — Not a video cue: leg length and body mass

• Leg length (r = 0.24, p = 0.48) and body mass (r = 0.07, p = 0.85) had no relationship with ball speed. Height (r = 0.76) and arm length (r = 0.61) did. For talent ID in the female game, measure the top half.

Caveats and limits

• n = 11. The authors concede this “limits the power of the statistical tests conducted,” while noting it is “still relatively large for this population.” Correlations this strong on eleven data points will shrink on replication; the r = 0.95 in particular should be expected to regress.
• Two-dimensional analysis from iPhone footage, which the authors explicitly rank as “secondary in terms of accuracy” to 3D lab capture. Crucially, 2D cannot see the axial rotation that Felton et al. (2019) identified as the defining feature of the female action — so the paper is partly blind to its own preferred explanation.
• Only four kinematic variables were tested (run-up, front knee at FFC, shoulder angle at BR, trunk flexion). The authors state that “a greater number of kinematic parameters are required to fully appreciate the characteristics of technique which influence BRS.”
• Correlational. No causal claim is available. Increasing a given bowler’s arm delay is not shown to increase her speed.
• Sub-elite speeds (23.0 ± 1.8 m/s) relative to the elite female cohort in Felton et al. (2019) (27.9 ± 1.4 m/s). Applicability to international-standard bowlers is unproven.
• Stepwise regression on a small n with candidate variables screened by correlation is a known overfitting risk; the authors mitigated it by checking all same-size models, verifying residual normality (Shapiro-Wilk), removing arm length for multicollinearity with height (r > 0.80), and rejecting the 93% model on CI grounds — which is more discipline than most papers show.
• The circularity problem, raised by the authors themselves: these women have been coached using male models their whole careers. Finding that male-derived cues correlate with their speed may just be measuring the coaching, not the mechanics.
• No adjustment for multiple comparisons.
• Internal inconsistency in the source: the Results text says “a further kinematic parameter (knee angle at BR) was added as a candidate variable,” but Table 1, Table 2 and the abstract all report knee angle at front foot contact, and knee angle at BR was never measured. Read it as FFC.

Relationship to other Felton work

• The full, peer-reviewed version of the 2022 BASES poster (same 11 bowlers, same equipment). Prefer this paper’s numbers.
• Explicitly framed as a test of Worthington, King & Ranson (2013) male findings and Felton et al. (2019) male/female differences, both of which it cites.
• Cites Felton, Yeadon & King (2020) simulation work as the basis for expecting a straight front leg to be optimal.
• CONTRADICTION: run-up speed vs the 2022 poster. The poster (same data) reported run-up speed as a moderate non-significant correlate of female ball speed; here it is r = 0.75, p = 0.01, significant. The reconciliation is probably the partial correlation (r = 0.41, p = 0.28 controlling for height), but the two published accounts of one dataset say opposite things.
• CONTRADICTION: front knee at FFC vs the 2022 poster. Non-significant on the poster; r = 0.68, p = 0.04 here (height controlled). Same eleven bowlers.
• CONTRADICTION: arm timing vs Felton 2015. This paper: a more delayed bowling arm at release goes with faster ball speed in women (r = 0.95). The 2015 abstract: the faster elite women had an earlier onset of arm circumduction (measured at front foot contact). The instants differ (BR vs FFC), and the samples differ (11 Irish sub-elite vs 18 elite), but the coaching implications point in opposite directions.
• CONTRADICTION: trunk flexion. This paper finds no relationship (r = −0.19, p = 0.57) between trunk flexion FFC→BR and ball speed in women. Worthington et al. (2013) found it to be one of four key predictors in men, and Felton et al. (2019) reports that female bowlers actually flex the trunk more at release (153.1° vs 159.5°). Trunk flexion is real in women and it is not buying them speed.
• TENSION: Felton et al. (2019) attributes the female action to pelvis/trunk rotation driving speed. This paper agrees in the discussion but could not test it — the 2D method cannot measure axial rotation. The rotational hypothesis remains untested in a female-only performance regression.