Cluster 03 — Fast Bowling Technique and Context

Measured (not simulated) fast bowling. Seven items covering three questions a coach actually asks: what separates a fast bowler from a slow one, whether women’s mechanics work the same way as men’s, and whether the environment and the approach change the action — the pitch surface underfoot, and the bound before the delivery stride.

Everything in this cluster is correlational or descriptive. Nothing here is an intervention study. These papers tell you what fast bowlers do; the simulation cluster (01) is where causal claims live.

Ranked coachable determinants of ball speed supported by this cluster

Ranked by strength of evidence × size of effect × how easily a coach can see it on a phone.

1. How far back the bowling arm still is at ball release (arm delay / shoulder angle at BR). Female: r = 0.95, p < 0.001; alone explains 89% of ball-speed variance in 11 bowlers; 0.224 m/s per degree (≈0.8 km/h per degree). Independently reproduced on the same squad in the 2022 poster (r = 0.95, adj. R² = 0.89). Male: the equivalent finding sits in Worthington et al. (2013), outside this cluster. Film: side-on, square to the plane of motion, pause on the release frame. ⚠️ But the 2015 female abstract found the opposite direction at front foot contact. See contradictions.

2. Run-up speed into the crease. The only variable in this cluster that shows up as speed-relevant in men and women, in multiple papers. Female elite: one of three variables explaining 74.1% of variance (2015). Female sub-elite: r = 0.75, p = 0.01 (2023). Male: the strongest known predictor (Worthington 2013). ⚠️ Two heavy caveats. (a) Control for height: in the female data the relationship collapses to r = 0.41, p = 0.28 once height is held constant — taller bowlers run in faster, and that’s most of the effect. (b) In elite men, pre-bound run-up velocity predicts arrival speed at back foot contact superbly (r = 0.844) but predicts ball release speed not at all, and predicts a worse front knee at release. Optimal run-up speed is individual, not a squad target.

3. Front knee straight and braced at front foot contact, and not collapsing after it. Female: r = 0.68, p = 0.04 once height is controlled; elite female squad sat at 167.0 ± 2.9° (all within ~13–19° of straight). Male: central to the male model. ⚠️ Only significant as a partial correlation; not significant bivariately (r = 0.49, p = 0.13) and not significant in the 2022 poster. And the 2015 female abstract found the fast women more collapsed at release. Film: side-on, the frame the front foot lands, then the release frame — look for the knee angle held, not just achieved.

4. The shape of the time budget: back foot → front foot vs front foot → release. Large effect size, and measurable with a frame counter on any phone. Elite men: ≈192 ms then ≈103 ms. Elite women: ≈172 ms then ≈128 ms — same total, redistributed. A short second window is direct evidence the front leg is braking and momentum is converting. Film: side-on, count frames. At 240 fps: men ≈46 then ≈25 frames; women ≈41 then ≈31.

5. Height and arm length — for selection, not coaching. Female: height r = 0.76, p = 0.007, alone explains 53% of variance; arm length r = 0.61, p = 0.05. Leg length (r = 0.24) and body mass (r = 0.07) predict nothing. Measure the top half. And note the authors’ warning: “bigger is not continuously better.”

6. Take-off angle and pelvic position in the bound. New from 2026, elite male, and the most coachable material in the cluster because it is upstream of the delivery. Elite range 11 ± 3° (5–17°) — flatter than most club bowlers bound. Flatter → faster at the crease (r = −0.400) and bigger front-leg plant angle (r = −0.428). Bound length 122 ± 18% of standing height; longer → faster arrival (r = 0.505). ⚠️ No bound characteristic correlates with ball release speed at all. The bound shapes the technique you arrive in; it does not directly make you fast.

7. Trunk flexion between front foot contact and release — MALE ONLY. One of the four pillars of the male speed model. In 11 female bowlers: r = −0.19, p = 0.57. Nothing, despite a 33° spread in the squad. Do not coach “get your chest over the front leg” as a speed intervention for a female bowler.

Variables this cluster tested and found NOT related to ball speed: delivery stride length (ns between sexes, unrelated to speed), body mass, leg length, pelvis–shoulder separation / counter-rotation (ns at every instant in the 2019 comparison), jump height in the bound, and every pre-delivery stride variable vs ball speed.

Coaching female fast bowlers

The short version: the same run-up and the same arm delay; a different engine, and a different reason the front knee collapses.

The elite gap is real and large: 34.9 ± 1.7 m/s (men) vs 27.9 ± 1.4 m/s (women) — about 126 vs 100 km/h — at comparable age, with women 21 cm shorter and 17 kg lighter.

What actually differs mechanically (2019, 20 v 20, large effect sizes):

• Women arrive at the crease slower (5.31 vs 5.76 m/s) and lose more momentum crossing from back foot to front foot.
• Women take 24% longer from front foot contact to ball release (127.7 vs 103.3 ms) — with the same total delivery-stride time. The time is redistributed, not added.
• Women finish more front-on at release (pelvis 298° vs 288°, shoulders 323° vs 310°) and rotate the pelvis further through the action (78.6° vs 68.7°).
• Interpretation: with less linear momentum to convert, women recruit the large trunk rotators to drive pelvis and torso round the long axis — a sequence “more akin to throwing.” The speed comes from rotation, not just from the linear-to-angular conversion the male model describes.

What this means in practice:

• Do not import the trunk-flexion cue. It is a male finding and it does not transfer (r = −0.19, p = 0.57). Female bowlers already flex the trunk more at release than men (153.1° vs 159.5°) and it is not buying them speed.
• Do import the arm-delay cue. It is the strongest relationship in the female literature by a distance.
• Individualise run-up speed by height. The run-up/speed relationship in women is largely a height effect.
• Treat front-leg collapse as a strength problem, not a technique target. The 2015 abstract found the faster women more collapsed at release, and explicitly attributes it to a strength deficit that “caused the front knee to collapse in the majority of the female fast bowlers.” Note that in the 2019 comparison the female mean front knee angle was fine — what was different was the variability (SD 24.9° vs 18.8°). The “female knees collapse” story is about the spread, not the average.
• Same for the position at back foot contact. Elite men held a higher front knee (hip 105.6° vs 123.0°) and a trunk still leaning back (185.2° vs 177.2°). The authors’ reading is that the trunk strength to hold that simply wasn’t there. Coaching a taller position onto a bowler who cannot support it moves the collapse elsewhere.

Two honest caveats on the whole female evidence base. First, everything rests on n = 18, n = 20 and n = 11, with the smallest study using 2D iPhone footage that cannot see the axial rotation the 2019 paper identifies as the defining feature of the female action. Second — the circularity problem, raised by Lyons et al. themselves — these women have been coached on male models their whole careers. Finding that male-derived cues correlate with their speed may be measuring the coaching, not the mechanics. No study in this cluster has established the optimal female technique. It has only established that the male one doesn’t fully describe them.

Artificial vs grass nets

Verdict: your indoor net footage transfers. Coach from it with confidence.

Eight sub-elite male bowlers, filmed on natural grass one day and indoor artificial turf the next, 18-camera Vicon, whole-waveform statistical parametric mapping across every performance and injury-related joint angle:

• Ball release velocity: 30.2 vs 30.4 m/s (p = 0.438). No cost to bowling indoors.
• Run-up velocity, and all three phase durations: no difference.
• No statistically significant kinematic difference anywhere, in any joint angle, at any point in either phase. Not just at the key instants — across the entire time-normalised movement.
• The only significant difference was foot slide: ~1.7 cm more at back foot contact (7.1 → 8.8 cm, d = 2.12) and ~2.1 cm more at front foot contact (6.6 → 8.7 cm, d = 1.16) on artificial. That is a footwear effect as much as a surface one — the study used 6 mm spikes on grass and non-spiked trainers indoors, exactly as bowlers really do.
• Injury risk: the two variables that best predict prospective lumbar stress fracture (rear hip flexion at back foot contact, lumbopelvic flexion at front foot contact) showed no surface difference. Practical consequence: indoor artificial deliveries load the back the same way. Count them in workload monitoring — winter net volume is not free.

Three limits worth stating to anyone who leans on this:

1. n = 8, and the authors call the study “likely underpowered.” A null from an underpowered study is weak evidence of no effect. The honest reading: no large surface effect exists, and the small trends observed (the biggest being ~7° of pelvic twist) are of no coaching consequence.
2. Two surfaces, not two categories. An 11 mm monofilament carpet over a 15 mm shockpad over concrete, vs a maintained clay-soil county pitch. A worn club astro over tarmac is not what was tested.
3. No ground reaction forces were measured. “Same kinematics” does not license “same forces.” Lower-limb loading may still differ.

This paper is also quietly load-bearing for the whole Felton corpus: the 2015, 2019 and 2026 studies were all collected indoors on artificial turf at the ECB National Cricket Performance Centre. This is the paper that lets those results be applied to grass cricket.

Every contradiction and tension flagged in this cluster

Female arm timing — the big one

• CONTRADICTION: Felton et al. (2015) found the faster elite women had an earlier onset of arm circumduction (shoulder angle at front foot contact). Lyons et al. (2023) found a more delayed arm at ball release was the strongest predictor of speed in women (r = 0.95). Worthington et al. (2013) in men says delayed. Different instants are being measured, which softens it, but the coaching implications point opposite ways.
• TENSION: Felton et al. (2019) found elite women have a more delayed arm at release than men (233.9° vs 219.4°). So women as a group are late, but the 2015 abstract says the fast ones among them are relatively early — while in men and in Lyons (2023) later is faster. The arm-timing evidence in the female game does not hang together.

Female front knee

• CONTRADICTION: Felton et al. (2015) — faster women had a more flexed front knee at ball release. Lyons et al. (2023) — straighter front knees at front foot contact went with faster speed (r = 0.68, height controlled). Worthington (2013) and Felton’s simulation work — straighter is better in men.
• TENSION: Felton et al. (2019) found front knee angle at release did not differ between sexes at all (172.5° women vs 167.3° men, ns) — women were marginally straighter on average. What differed was the SD (24.9° vs 18.8°). The “female knee collapse” narrative is about spread, not mean.

The 2022 poster vs the 2023 journal paper — same 11 bowlers, opposite conclusions

• CONTRADICTION: Run-up speed. Poster: moderate, non-significant (P > 0.05). Journal: r = 0.75, p = 0.01, significant. (Probable reconciliation: the journal’s partial correlation controlling for height gives r = 0.41, p = 0.28 — but the two published accounts of one dataset say opposite things.)
• CONTRADICTION: Front knee at front foot contact. Poster: non-significant. Journal: r = 0.68, p = 0.04 with height as covariate.
• TENSION: The poster’s headline conclusion is “don’t extrapolate from men” — but it is drawn from finding that run-up and front knee don’t matter, while the one variable that did matter (delayed arm) is itself a male finding. The rhetoric outruns the evidence.

Trunk flexion — male cue that fails in women

• CONTRADICTION: Trunk flexion FFC→BR is one of four pillars of the male speed model (Worthington 2013). In Lyons et al. (2023): r = −0.19, p = 0.57 — nothing, across a 24.4°–57.2° spread. And Felton et al. (2019) shows women actually flex the trunk more at release. It is happening, and it is not buying speed.
• TENSION: Felton et al. (2019) attributes female speed to pelvis/trunk rotation. Lyons et al. (2023) agrees in the discussion but could not test it — the 2D method cannot measure axial rotation. The rotational hypothesis has never been tested in a female-only performance regression.

The bound (2026), internal

• TENSION: Faster pre-jump run-up and longer bound produce a larger front leg plant angle at FFC (good — more braking) but a more flexed front knee at ball release (bad — worse momentum transfer). The authors call this “conflicting” in their own discussion. Two variables the male literature treats as pointing the same way are driven in opposite directions by the same upstream cause. Their resolution: the relationships are non-linear and these bowlers were running in above their individual optimum.
• CONTRADICTION: There is no take-off angle that optimises both goals. Lower take-off angle → faster at BFC and bigger plant angle. Higher take-off angle → straighter front knee at release. Any squad-wide “flatten the bound” or “get more air” instruction is guaranteed to be wrong for some bowlers.
• TENSION with the male speed model: Worthington (2013) makes run-up speed the top predictor of ball release speed. In Bull et al. (2026), pre-bound run-up velocity predicts arrival speed at BFC brilliantly (r = 0.844) but predicts ball release speed not at all (r = 0.262, ns) and predicts a worse front knee at release. “Run in faster” does not survive contact with the bound.

Grass vs artificial

• TENSION (methodological): The claim that “previous research utilising artificial surfaces is likely to be valid” is strong, and it rests on a null result from n = 8 from the 2024 Alway study that the authors themselves call underpowered. It is the right conclusion on the available evidence, but it is a licence granted by a small study, and the authors call for corroboration.

Source gaps

• 2017 Felton is marked SOURCE NOT RECOVERED. The ISB 2017 Brisbane proceedings are not openly indexed at abstract level and this item is not on Felton’s own site. It is almost certainly the conference version of the 2019 journal paper; no figure should be cited to it.
• The 2015 item is a one-page conference abstract with no r values, coefficients, or means — only the 74.1% model R² and the direction of each relationship.
• The 2022 item was recovered as a JPG poster image and transcribed. Note its cited title says “kinetic variables” but its printed title says “kinematic,” and no kinetics were measured.
• Two internal reporting discrepancies in Bull et al. (2026): a landing-velocity → bowling-shoulder correlation appears in the text (r = 0.379, p = 0.043) but not in Table 2 (0.064, unstarred); and the pelvic-tilt → rear-hip sign differs between text (−0.716) and Table 3 (+0.716). Prefer the tables. One in Lyons et al. (2023): the Results text says “knee angle at BR” where every table and the abstract say “knee angle at FFC” — read it as FFC.