Research theme
The 2022 ISBS paper established that ball release angle is what distinguishes a yorker from a stock ball from a bouncer (η² = 0.78–0.79). The obvious next question for a coach: what is the body doing to produce that angle? This abstract reports the answer — separate stepwise regressions of release angle onto whole-body kinematics, run independently for each delivery type. Method: the same 21 county-level fast bowlers, 48 deliveries each (12 yorkers, 12 bouncers, 24 stock), 18-camera Vicon at 250 Hz, a 53-marker full-body model built specifically for fast bowling, plus two ball markers. One representative trial per bowler per delivery type (the successful trial closest to that bowler’s mean landing point for that type) fed the ANOVA and regressions. Correlational — regression on measured technique, not simulation and not intervention.
What they measured
(abstract only — the abstract names only the parameters that survived into the models; the thesis lists the full candidate set)
- The upward tilt of the ball’s path the instant it leaves the hand (ball release angle).
- How cocked-back or flexed the bowling wrist is at the moment of release (wrist angle at ball release).
- Which way the hand as a whole is pointing at release, seen side-on against the horizon (2D hand segment orientation at ball release).
- Other release parameters (speed, height) were compared across delivery types by ANOVA; effect sizes other than release angle are not reported in the source I could access.
Findings
- Release angle is again the standout. ANOVA showed significant differences between the three intended lengths, with release angle carrying the highest effect size of any release parameter, η² = 0.78. (The absolute release-angle means are not reported in this abstract; see the 2022 ISBS file and the thesis file — yorker ~2°, stock ~5°, bouncer ~13°.)
- For a YORKER, the wrist and hand explain almost everything. 86.6% of the variance in release angle was explained by wrist angle at ball release plus 2D hand orientation at ball release. (The thesis adds the split: wrist angle alone = 75.3%.)
- For a BOUNCER, nothing in the technique model predicted release angle. The abstract’s word is “unexpectedly” — no technique parameter correlated with bouncer release angle at all. No model could be built.
- For a STOCK ball, the link is weak. Only 31.2% of release-angle variance was explained, by 2D hand orientation at ball release alone.
- Implication. The fuller you try to bowl, the more the length is decided by a small, fast, late wrist-and-hand action — and the more precisely that action has to be repeated.
What a coach should look for on video
Cue: wrist position at the moment the ball leaves the hand, when bowling a yorker.
- Camera view + frame: Side-on, slow-motion, zoomed on the hand. Scrub to the exact frame of ball release and the 2–3 frames either side.
- What “good” looks like: for a yorker, a more extended (more cocked back) wrist at release than the same bowler shows on a bouncer. This single variable accounted for 75.3% of the yorker release-angle variance in the thesis version of this analysis; add the direction the hand is pointing and you reach 86.6%.
- What the fault looks like: a yorker attempt where the wrist snaps through into the same flexed position the bowler uses for a bouncer — the ball comes out with a good-length tilt and lands 3–4 m out, as a half-volley.
- Why it matters: this is the single most concrete, most predictive, most video-visible cue in the whole variations line. If a bowler keeps missing the yorker short, look at the wrist first.
Cue: where the hand is pointing at release (hand segment against the horizon).
- Camera view + frame: Side-on, freeze at ball release. Read the line through the forearm/hand relative to horizontal.
- What “good” looks like: a consistently lower, flatter hand line for full deliveries. It was a predictor for both yorkers and stock balls.
- What the fault looks like: the hand line varying ball to ball on stock deliveries — but note this only accounted for 31.2% of stock-ball release angle, so it is a partial cue, not the whole answer.
- Why it matters: it is the last thing acting on the ball, and it is coachable by feel.
Cue (a negative one): do NOT try to coach a bouncer through a technique checkpoint.
- Camera view + frame: n/a.
- What “good” looks like: leave it alone. In this dataset no measured technique parameter predicted bouncer release angle, and bouncers were landed successfully 98.7% of the time anyway.
- What the fault looks like: a coach imposing a wrist or trunk cue on a short ball that the bowler already lands reliably, and disturbing it.
- Why it matters: the authors’ own explanation is that the bouncer’s successful release-angle window is so wide (~16°) that many different techniques land in it. Nothing to fix.
Caveats and limits
- This is a single-page abstract. No methods detail beyond participants and hardware, no results table, no p-values for the regressions, no confidence intervals, no discussion. Treat the numbers as headline figures and use the thesis for anything load-bearing.
- n = 21, all male, all county level. Regressions used one trial per bowler, so n = 21 per model — small for a stepwise regression with multiple candidate predictors, which inflates R² and makes the models fragile. The thesis reports 95% CIs on the explained percentages that are enormous (e.g. 86.6% with a CI of “<1% – 97%”). Do not treat 86.6% as precise.
- Stepwise regression on a small n is a variable-selection method, not a causal test. It tells you which measured variable tracks release angle in this sample; it does not prove that changing the wrist changes the length.
- Lab, indoor, artificial surface, no batter (per the parent study).
- The abstract contains a garbled sentence — “Unexpectedly no technique parameter was correlated with the ball release angle” appears without naming the bouncer condition. The thesis makes clear this refers to the bouncer regression.
Relationship to other Felton work
- Direct follow-up to Manawadu et al. 2022, 40th ISBS, which it cites as its only reference. That paper found release angle mattered; this one finds what produces it.
- Published in full as Chapter 6 of Manawadu 2023 PhD thesis — same data, same models, with the full tables, the ball-release-height regressions and the discussion.
- Refines rather than contradicts the Felton speed line. Worthington, King & Ranson (2013) and Felton, Yeadon & King (2020) locate ball speed in the front leg and trunk. This locates ball length in the wrist and hand. Different levers for different outcomes — a genuinely useful division of labour for a coach.
No contradiction with other Felton work.