05 — Spin bowling and batting

The non-fast-bowling half of Felton’s cricket work. Two separate research lines, connected only by the lab (18-camera Vicon at the ECB National Cricket Performance Centre, Loughborough), the funder (ECB, plus ICC for the batting work), and Mark King as senior author on everything.

Finger spin — Felton co-supervised Liam Sanders’ PhD. Three published outputs plus a 211-page thesis that contains two substantial studies that were never published anywhere else.

Power hitting — Felton was a co-author on Stuart McErlain-Naylor’s male-vs-female batting comparison, in conference and journal form.

Sourcing note: the drpaulfelton.com URL for the 2019 power hitting conference paper returns HTTP 404; it was retrieved from the first author’s own site instead. All other PDFs were retrieved and text-extracted in full. No numbers in these files were inferred or estimated — where a source did not report something, the files say so.

Coaching finger spin

The one-sentence version: the traditional side-on model is wrong for elite finger spinners; the bowlers who spin it hardest are already half-open at the hips when the front foot lands, and rotate past square by release.

Body cues (source: 2018 Sanders / Sanders PhD thesis Ch. 4–5, n = 23, correlational)

1. Hips half-open at front foot contact — about 45° between fully side-on and square (≈225°). Strongest single correlate of spin rate, r = 0.674, p < 0.001. High overhead camera, frame of front foot landing.
2. Shoulders slightly open of side-on, and lagging the hips by ~20° at front foot contact. Separation r = 0.521, p = 0.011. Same view, same frame.
3. Hips past square (>270°) at ball release. r = 0.676, p < 0.001. Only 23% of this elite group managed it — they were the big spinners.
4. Both feet pointing further down the wicket than “back foot parallel to the crease”. r = 0.433 (back foot at BFC), r = 0.416 (front foot at FFC). Treat as a check on hip position, not a target.
5. Sequence: hips, then shoulders, then arm, then forearm pronation last. Supported indirectly (via the separation finding), not measured as a timing variable.

Things elite finger spinners do NOT do, despite the coaching:

• Only 14.3% used a genuinely side-on action. 64.3% used a “mixed” action.
• Neither hips nor shoulders complete the 180° rotation the textbooks describe (158° and 171° on average, BFC to follow-through).
• The front knee flexes into release on average (−9.5 ± 17.0°), and knee extension had no relationship with spin (r = 0.156, p = 0.477). No support for bracing the front leg.
• Run-up speed doesn’t matter (r = −0.155). Elite finger spinners jog in at 2.5 m/s.
• Wrist flexion doesn’t matter (r = −0.133). Upper-arm internal rotation doesn’t matter (r = −0.031) — the opposite of what throwing biomechanics would predict.

Off-field screening (source: 2019 Sanders passive ROM, n = 16)

Before you coach the hip positions above, check the athlete can physically get there.

• Front-hip total rotation arc (supine, knee bent 90°, inclinometer on the shin). Only significant predictor: r = 0.552, 25.5% of spin variance. Group mean 82.8 ± 10.8°, range 67–107°. Flag anything under ~70°.
• Rear-hip internal rotation. r = 0.485, p = 0.057. Mean 34.3 ± 8.0°.
• Bowling-shoulder internal rotation. r = 0.476, p = 0.063. Mean 77.4 ± 11.7°.
• Injury-monitoring measure (not performance): compare total shoulder arc bowling vs non-bowling side. In this group they matched (214.8° vs 211.8°), with the normal protective shift — more external rotation, less internal. A loss of total arc on the bowling side is the pattern linked to injury in baseball.

Match-play cues — visible from the boundary, no lab needed (source: Sanders PhD thesis Ch. 7, 69,552 test-match deliveries, unpublished)

This is the most directly usable finding in the whole cluster and it exists only in the thesis.

• Length 4–5 m from the batter’s stumps. Explains 22.9% of economy variance on its own. Optimum 4.40 m. Costs rise quadratically both ways.
• Line within 0.15 m of middle stump. Takes the model to 52.8%. That corridor produced average 22.73 and economy 2.30 rpo against a baseline of 33.37 and 2.94.
• Release speed toward 56–57 mph, and no further. Takes the model to 54.4%. Group mean was 52.5 mph, so most are below the optimum. Above 57 mph, line and length scatter sharply (speed vs combined line/length SD: r = 0.934).
• Turn it away from the batter. 24.3% better average and 17.3% better economy than turning it in. Off-spinner to a lefty, slow left-armer to a righty.
• 2.2° of deviation is enough to beat the bat — 9 cm over 2.5 m. Batters cannot adjust inside ~200 ms.
• Flight height did nothing. No significant effect on average or economy across 69,552 deliveries.
• The optimised delivery: 57.4 mph, 4.40 m, 0.11 m → 1.67 runs per over. (A regression extrapolation, not an observed delivery.)

Wrist spin is a different sport — do not transfer these cues (source: Sanders PhD thesis Appendix A, n = 10, unpublished). See the contradiction list below.

Measure spin rate and release speed TOGETHER, never in isolation (thesis §9.6.2). Drift, dip and deviation come from Magnus force, which depends on both. A bowler who gains 3 mph while losing revs may be going backwards. If you can afford one piece of technology for a spinner, buy a radar, not a better camera.

Coaching power hitting

The one-sentence version: bat speed comes from a wind-up between hips and chest at the top, then a straightening front arm on the way down — and the front arm is where male and female batters differ most.

Cues (source: 2021 McErlain-Naylor power hitting, n = 15 + 15; prefer it over the 2019 conference abstract)

1. Lead elbow extension through the downswing — the highest-yield cue in this cluster. Front-on camera; two frames, top of backswing and contact. Good = elbow opens ~30° (male mean +29.7 ± 12.0°). Fault = elbow folds (female mean −3.0 ± 23.5°; eight of fifteen female batters flexed, by up to 34°). β = 1.28, p = 0.008, with near-complete separation between groups — the nine largest extensions were all male, the nine smallest all female.
2. X-factor at the start of the downswing — hips vs chest, viewed from above (transverse plane; a front-on camera will not give you this honestly). Good ≈ 18° (male mean 17.6 ± 8.3°); fault ≈ 12° or less. β = 1.14, p = 0.030. Note: both sexes gained the same extra stretch during the downswing (7.2° vs 7.0°) — coach the position at the top, not the stretch after it.
3. Rear elbow angle at impact — ~126° vs ~112°. β = 1.03, p = 0.044. Coach the position at contact; rear elbow extension through the downswing did not survive the covariate-controlled analysis (p = 0.086).
4. Check the bat before you coach the technique. The authors’ leading hypothesis for the lead-elbow difference is that bat moment of inertia is poorly scaled to the athlete’s strength — in baseball, too much bat inertia produces exactly this lead-arm-stabilising pattern. Cheap to check, expensive to get wrong.

Stop coaching these — the paper tested them and they failed:

• “Snap the wrists.” Wrist uncocking was a pre-registered hypothesis that was rejected: p = 0.819, β = −0.14 (trivial). Female batters uncocked marginally more. Wrist cock at the top was essentially identical (119.3° vs 118.7°, p = 0.968).
• Anything in the lower body. Lead knee angle at impact, lead knee extension, base width, forward CoM displacement — all non-significant. And the front knee flexes into impact in both sexes.
• Setup position. Bat angle at the top, bat height, wrist cock — all non-significant. The difference emerges during the downswing.

Does the same model predict distance in both sexes? — the honest answer is: not established. The 78%-of-bat-speed model (X-factor + lead elbow extension + wrist uncocking) was built on male batters only (Peploe et al., 2019). The 2021 paper shows two of those three variables differ between the sexes in the predicted direction and the third does not. It never tested whether they predict carry distance within a female cohort. That is not reported in any source in this cluster.

The 2021 paper makes exactly this point about its own field, citing Felton’s fast-bowling work: the parameters that differ between male and female fast bowlers are not the same parameters that predict ball speed in male fast bowlers. A sex difference in a variable is not evidence that the variable drives performance. Apply the same scepticism to the batting results.

The paper’s own framing is that “gender” is a container for anthropometry, strength, bat inertia, boundary size, incoming ball speed, and coaching history — not a biological claim. Some female batters in the sample did extend the lead elbow by up to 30°, and the parallel golf finding (female golfers flex the lead elbow, but professional female golfers do not) suggests the difference tracks skill and exposure rather than being fixed.

What video CAN and CANNOT resolve for spin

The decisive fact is buried in Sanders’ thesis methods and repeated in his limitations chapter:

Four reflective markers were placed on the index and middle finger knuckles of the bowling hand specifically to quantify finger motion during the delivery — and were discarded from all analysis because the markers sat too close together to track, leaving large gaps in the joint-centre trajectories.

An 18-camera Vicon system running at 300 Hz in a controlled 7 × 3 × 3 m volume could not resolve the fingers of a spin bowler. A phone at 240 fps from the boundary has no chance whatsoever.

CANNOT resolve

• Finger action of any kind. Not the spread angle between index and middle finger, not the release from the finger pads, not the “rip”. This is the thing coaches most want to see and it is unavailable at any consumer price point. The 2018 paper names the absence of finger motion as a formal limitation.
• Wrist flexion. ~30° of movement in the last few hundredths of a second. (No loss: r = −0.133, unrelated to spin.)
• Upper-arm internal rotation. (No loss: r = −0.031, unrelated to spin.)
• Elbow extension to legality standard. Requires the ICC’s marker-based protocol. Phone footage cannot adjudicate the 15° limit, and the thesis’ own two chapters report irreconcilable values for this variable on the same 23 bowlers (3.65 ± 3.24° in Ch. 5, 10.6 ± 6.1° in Ch. 4).
• Forearm pronation, quantitatively. ~11° in the final instants before release. You may see whether the forearm pronates on a good 240 fps side-on clip; you will not measure how much. Treat as a yes/no.

CAN resolve — and, fortunately, these are the variables that predicted spin

• Pelvis line at back foot contact, front foot contact and ball release (high overhead camera).
• Shoulder line at the same instants.
• The gap between them — hip-shoulder separation.
• Foot orientations at both contacts.
• Sequencing order (hips → shoulders → arm) at 240 fps side-on.
• All the Chapter 7 ball-flight variables — length, line, deviation direction — from behind the arm with a marked pitch.

The honest accounting

The strongest predictors of spin rate happen to be exactly the large, slow, high-contrast whole-body orientations a camera sees best. That is the good news, and it is genuinely useful. But pelvis orientation explains ~43% of the variance in spin rate. The other ~57% lives substantially in a hand you cannot film. Do not let a coach believe that a perfect pelvis position guarantees a big-spinning bowler, or that a video review has diagnosed a spin problem. It has diagnosed the part of the spin problem that is visible.

The proxies that DO work on video, in priority order: hip openness at front foot contact → hip openness at ball release → hip-shoulder separation at front foot contact → shoulder openness at release → foot orientations. Then leave the hand alone and buy a radar.

Contradictions and tensions flagged in this cluster

Substantive contradictions

• Finger spin vs wrist spin — the cues invert (Sanders PhD thesis Ch. 5 vs Appendix A). Finger spin: more pelvis rotation past front-on at release = more spin (r = +0.676). Wrist spin: pelvis orientation at release correlates negatively (r = −0.644), shoulder orientation at release negatively (r = −0.875), and less shoulder rotation BFC→BR (below 45°) is the single best predictor (r = −0.837, 80.8% of variance). Sanders states explicitly that the two must be coached differently, against the generic spin-bowling model. Never transfer finger-spin cues to a leg-spinner.
• Both spin studies vs coaching orthodoxy (Woolmer & Noakes, 2008). The side-on-to-front-on model is contradicted for finger spin and wrist spin. The authors say so directly.
• Sanders/Felton vs Beach et al. (2017/2018). Beach, on club bowlers, described finger spin as a push-like action with maximum internal shoulder rotation velocity correlating with spin. Sanders/Felton, on elite bowlers, find a sequential action and no upper-arm internal rotation relationship. Beach also proposed that wrist spinners’ faster run-ups drive their higher spin; the thesis’ wrist spin study found no run-up/spin relationship (r = −0.512, p = 0.130).
• Power hitting: plane of pelvis-thorax separation, conference vs journal. The 2019 conference abstract reports the significant difference in the FRONTAL plane (moderate evidence). The 2021 journal paper reports it in the TRANSVERSE plane (X-factor, β = 1.14, p = 0.030) and finds the frontal measure non-significant (X’-factor, p = 0.110). Opposite plane assignments for the same data. Prefer the journal version — peer-reviewed, and it controls for height and body mass (males were 12 cm taller and 12 kg heavier).
• Power hitting: rear elbow extension, conference vs journal. Conference: males extend the rear elbow more during the downswing, “very strong” evidence. Journal: non-significant (β = 0.831, p = 0.086). Rear elbow angle at impact survives in both.

Tensions and internal inconsistencies

• Spin vs legality (raised by Sanders himself, thesis §9.6.1). Spratford et al. (2018) advise bowlers wanting to reduce elbow extension to be more side-on at BFC — precisely the position Ch. 5 associates with lower spin. Sanders: the literature “illustrates a mixed message to the coaching community”, and coaches need “a safe space by which coaches do not fear coaching instruction with a sole purpose on improving aspects associated with performance … verse safety”.
• The 43.1% / 42.9% labelling muddle. The 2018 journal text says pelvis at ball release = 43.1% and pelvis at FFC = 42.9%; its own Table 3 has them swapped. The conference abstract follows the text. The thesis abstract follows the table; the thesis research-question summary follows the text. Difference is 0.2 percentage points and changes no coaching advice, but do not quote either as unambiguous.
• Thesis internal: elbow extension. Ch. 4 reports 0–15° (10.6 ± 6.1°) for UAH→BR on the 23 finger spinners; Ch. 5 and the journal paper report 0–10.5° (3.65 ± 3.24°) on the same bowlers. Irreconcilable, and it matters — 10.6° sits close to the 15° legality limit, 3.65° does not.
• Thesis internal: deviation angle. §7.5.4 says the lowest averages and economies came at large deviations (>±7°, average 11.95, 2.69 rpo); the Ch. 9 summary says the lowest averages came at small deviations (±2°). Probably a U-shaped curve, but the thesis never says so. Do not read either as “the optimal amount of turn”.
• Thesis internal: participant count. Abstract says 36 bowlers in the Hawk-eye study; Ch. 7 says 34.
• ROM paper internal: abstract gives rear-hip internal rotation as r = 0.466, p = 0.059; Table 3 gives r = 0.485, p = 0.057. Quote the table.
• ROM paper: adaptation runs against performance. The bowling shoulder adapts toward less internal rotation (p = 0.089), but more internal rotation is what correlates with spin (r = 0.476). The authors’ unfalsified speculation is that the adaptation comes from fielding throws, not bowling — implying throwing workload in spinners may need monitoring.
• Power hitting: downswing duration. Conference reports females having longer downswings (“strong” evidence). The journal version’s 26-parameter table does not contain the variable at all.
• Power hitting: methods drift, conference vs journal. 400 Hz vs 250 Hz; 51 markers on participant plus bat vs 46 body + 5 bat; 28 parameters vs 26; 7 females flexing the lead elbow (−9 to −34°) vs 8 (−7 to −34°); Bayesian t-tests vs GLM with height and body mass as covariates. The covariate change is the likely driver of the substantive disagreements above.
• Front-leg bracing — unsupported everywhere in this cluster. Finger spin: no relationship with spin (r = 0.156, p = 0.477), and the average front knee flexes (−9.5 ± 17.0°). Power hitting: lead knee angle at impact and lead knee extension both non-significant, and both sexes’ mean knee extension is negative — the front knee flexes into contact. A durable coaching instruction that no study here supports.
• Lab ≠ match. Elite finger spinners release at 52.12 ± 1.39 mph in test cricket but 44.95 ± 1.12 mph in labs across six studies (Δ = 13.76%, p = 0.001) — while the four bowlers in Ch. 5 who had actually played tests released at 53.34 mph. Every spin rate and release speed in the lab studies should be read with that ~14% gap in mind.