Research theme
This is the conference version of the elbow work that became the 2016 Journal of Sports Sciences paper. It answers a question with direct ICC/legality relevance: a hyperextending elbow is exempt from the 15° extension limit (the ICC treats hyperextension as involuntary, caused by load on the joint), so a bowler born with lax elbows is allowed something a bowler with tight elbows is not. Does that actually buy ball speed, and how much?
Method, plainly: a two-segment planar computer simulation model of the bowling arm only (upper arm; forearm+hand with the ball as a point mass), customised to one elite fast bowler. A constant torque at the shoulder drives the arm over; the elbow is modelled as a damped linear torsional spring that only acts once the elbow passes straight — i.e. hyperextension is treated as a passive spring, exactly as the ICC’s “involuntary” characterisation implies. The shoulder joint centre is dragged horizontally using the bowler’s real data. A separate one-segment model (upper arm + forearm + hand fused, no elbow) gives the counterfactual “what would this same bowler have bowled with a straight arm”.
What they measured
- How fast the ball leaves the hand (ball release speed, mph and m/s).
- How far past straight the elbow bends back (peak elbow hyperextension angle, degrees).
- How much the elbow springs back towards straight before the ball leaves (“recoil” — the amount and percentage of recovery from peak hyperextension by release).
- The elbow angle at the moment of release (elbow angle at ball release, degrees).
- How long the delivery period lasts (total simulation time).
- How fast the arm is rotating (upper arm and forearm angular velocities).
Findings
All causal within the model, single bowler.
The model matched real performances well. One common parameter set (shoulder torque magnitude, elbow spring stiffness and damping) matched three deliveries with an average RMS difference of 3.8% (individual trials 4.5%, 4.4%, 2.5%). A fourth, held-out delivery was simulated with the fixed parameters and matched to 4.4% — a genuine evaluation, not just a fit. Per-component differences (Table 1): peak elbow angle 0.0–0.5°, elbow angle at release 0.4–2.5°, ball velocity 1.1–5.0%, total time 4.3–8.6%.
This bowler’s hyperextension was worth 4%. His fastest delivery was 85.8 mph (38.1 m/s). The one-segment straight-arm model, given the same inputs, predicted 82.5 mph (36.7 m/s). So the elbow hyperextension he already had was generating an extra 4% of ball speed relative to bowling with a straight arm.
Optimised hyperextension was worth 5%. Letting the elbow spring parameters vary to maximise ball speed (with a penalty preventing peak hyperextension exceeding a 25° upper bound) gave 86.6 mph (38.5 m/s) — +5% vs a straight arm. The optimum hit the 25° ceiling and recoiled 5° before release. The optimal damping was zero (unsurprising: damping removes energy).
Hyperextension always helps — but the amount depends on two things, not one. Perturbing the spring stiffness produced elbow histories in three categories: “Recoiling” (elbow springing back at release), “At peak” (elbow at maximum hyperextension at release), and “Extending” (still hyperextending at release — dismissed as mechanically unrealistic under bowling loads). In both realistic categories, ball speed was always faster than with a straight arm. The gain is governed by peak hyperextension magnitude and how much it recoils, so an optimal recoil percentage exists for every peak angle.
Two competing mechanisms explain it. (a) To land the ball in the same place, a more hyperextended elbow forces a larger shoulder release angle, so the shoulder torque acts over a longer arc and does more work → faster arm. (b) As the elbow recoils, the wrist’s angular velocity about the elbow acts in the same direction as the shoulder torque, adding to ball speed. But recoiling reduces the benefit from (a). The trade-off is why an optimum recoil exists.
What a coach should look for on video
Honest framing first: this is not a coachable technique change. Elbow laxity is an anatomical gift. A bowler cannot train hyperextension into their elbow, and trying to would be both futile and an injury risk. The value of this paper to a coach is in talent identification and in interpreting why one bowler is quicker than an apparently identical one — which is exactly what the paper claims (“this knowledge can help inform talent identification protocols and coaching practice”).
Cue 1 — Does this bowler hyperextend at all?
- The cue: Whether the bowling elbow goes past straight during the delivery arc.
- Camera view + frame: Behind-the-arm or side-on, high frame rate, scrubbing the window from upper arm horizontal to ball release. This is a fast window — you need genuine slow motion (240 fps phone slo-mo is about the minimum), and the elbow must be clearly visible.
- What “good” looks like: The forearm passing beyond the line of the upper arm — a visible backwards bow at the elbow. Angles above 20° are documented in the literature (King & Yeadon, 2012) and are visible on video. A simple static screen: ask the bowler to straighten the arm fully with the palm up and see whether the forearm goes past straight.
- What the fault looks like: Not a fault — a bowler with a locked, tight elbow simply does not have this speed source. Do not try to create it.
- Why it matters: Worth ~4% to the bowler in this study, and up to ~5% at the modelled 25° ceiling. At 85 mph that is roughly 3–4 mph.
Cue 2 — Does the elbow spring back before release?
- The cue: After the elbow reaches its maximum bend-back, does it start recovering towards straight before the ball is gone?
- Camera view + frame: Same view; compare the frame of peak hyperextension with the release frame.
- What “good” looks like: Peak hyperextension occurring slightly before release, with the elbow already recoiling at release. In the optimum, peak was 25° with 5° of recoil by release.
- What the fault looks like: The elbow still at (or still heading towards) peak hyperextension at the moment of release — this is the “At peak” category, still faster than a straight arm but leaving speed on the table.
- Why it matters: The paper’s headline conclusion is that the gain is not governed by peak hyperextension alone — recoil matters too. Two bowlers with identical 20° hyperextension can extract different amounts of speed.
Do not use this to judge legality. The ICC’s 15° limit is on elbow extension between upper arm horizontal and release, and hyperextension is explicitly exempt as involuntary. Nothing in this paper changes that, and phone footage is not adequate to assess a bowling action’s legality — that requires the formal ICC testing protocol.
Caveats and limits
- n = 1, one elite male fast bowler (age/height/mass are given in the 2016 journal version: 19 years, 1.80 m, 82.4 kg), four deliveries.
- Two segments, planar. This is an arm model, not a bowler. The trunk, legs and run-up do not exist in it; the shoulder is driven along a horizontal path taken from real data and the shoulder torque is a single constant. It cannot tell you how hyperextension interacts with the rest of the action.
- Hyperextension is modelled as a passive spring — consistent with the ICC’s “involuntary” framing, but it means the model cannot represent a bowler actively flexing or extending the elbow. Active elbow torque was not included.
- Simulation, not intervention. No bowler was made to change anything.
- The 25° hyperextension ceiling is an imposed penalty bound, taken from King & Yeadon (2012), not a discovered limit. The optimum sat exactly on it, so the “5%” figure is a ceiling artefact as much as a result — with a higher bound the model would presumably have gone further.
- Time-match was the weakest component (4.3–8.6% differences on total duration).
Relationship to other Felton work
- This is the conference precursor of Felton 2016 — the effect of elbow hyperextension on ball speed, which reports the same 4% / 5% headline numbers and the same match/evaluation values (4.5%, 4.4%, 2.5%, 4.4%) but adds ~7,000 perturbation simulations, an inverse-hyperbolic optimal-recoil relationship, and the practically useful "~0.2% of ball speed per additional degree of hyperextension" rate. The journal version is the one to cite.
- Listed in the PhD thesis front matter as a conference output. It uses a completely different, much simpler model from the 16-segment whole-body model that carries the rest of this cluster.
TENSION: (softening at journal stage) This conference version states flatly that “the fast bowler’s ball speed was increased by 4% due to elbow hyperextension” and that the limiting characteristics “were found to be” peak hyperextension and recoil. The 2016 journal version hedges more carefully, notes that a non-hyperextending bowler may still be faster for other technique or strength reasons, and explicitly frames the claim as a comparison between two otherwise similar bowlers. See the 2016 file.