The Kinetic Chain: Motion, Work, and Control in the Golf Swing

The central question of the kinetic chain is how a golfer’s coordinated motion produces a useful club state at impact. The answer involves muscle work, contact forces, changing geometry, inertia, elastic storage, and timing. None of these can replace the others. A pelvis that slows while the club accelerates is an observation worth explaining; it is not, by itself, a measurement of energy moving from pelvis to club. This chapter develops the balances needed to make that explanation testable.

The useful intuition behind the whip analogy is that a connected system can develop high distal speed without every part moving equally fast. Its limitation is equally important: a golfer has active joints, two hands on a club, two changing foot contacts, and a three-dimensional body. A prescribed sequence of peaks is not a mechanical law for that system. We will use simple rotors and a moving-grip club to isolate mechanisms, then reconnect those mechanisms to measurements and control. The examples are deliberately specified teaching models, not fitted swings or prescriptions for an athlete.

Define the Observation Before Explaining It

Segments, Joints, and Comparable Rates

A segment is a body with a defined mass, center of mass, and inertia. A joint is a connection with relative motion and transmitted force or moment. A wrist is a joint region, not an additional rigid body whose energy can simply be added to forearm and hand energy. Similarly, a measured shoulder-line orientation may describe the thorax; counting both as separate energy reservoirs can count the same mass twice. A whole-body energy calculation requires disjoint segment definitions.

For a segment, the angular velocity vector describes rotation relative to a declared frame. A joint rate describes relative motion between adjacent segments. Neither is clubhead linear speed. The relation between clubhead velocity and generalized velocity is a Jacobian relation, $\mathbf v_C=J_C(q)\dot q$, including base translation when the base moves. Changing posture changes this map. Comparing a pelvic rate in degrees per second with clubhead speed in miles per hour does not establish amplification without geometry and a unit conversion.

Three-dimensional orientation also requires care. Euler-angle derivatives generally are not the Cartesian components of angular velocity. If $\boldsymbol\omega=E(\theta)\dot\theta$, a physical moment $\boldsymbol\tau$ has generalized force $E^T\boldsymbol\tau$, so that power is preserved. Angular-speed magnitude, a selected angular-velocity component, and the derivative of a projected turn angle can have different peak times.

A Sequence Is a Measurement Result

Define downswing onset, the instant immediately before ball contact, the measured quantity, its sign convention, smoothing, and the method for finding peaks. If $t_0$ and $t_I$ are those events, normalized time is $s=(t-t_0)/(t_I-t_0)$. A peak at a percentage of downswing time is meaningful only under that event definition. Finite sampling and filtering create timing uncertainty; a broad plateau should not be represented as a precisely ordered instant.

Peak segment speed and peak acceleration are distinct events. For a smooth signed velocity, an interior velocity extremum has zero acceleration; an acceleration extremum concerns its derivative. Nothing makes the peak speed of one segment equivalent to the peak acceleration of its neighbor. A speed magnitude introduces another distinction: $d\|\boldsymbol\omega\|/dt=\boldsymbol\omega\cdot\dot{\boldsymbol\omega}/\|\boldsymbol\omega\|$ away from zero speed.

Vena and colleagues’ five-subject study, using instantaneous screw-axis angular velocities, reported support for a proximal-to-distal kinematic sequence in two subjects, rather than a universal order (Vena et al. 2011). That result does not say sequencing is useless. It says a proposed invariant sequence must survive explicit measurement choices and between-player variation. We therefore do not assign universal pelvic, thoracic, arm, or wrist peak percentages here. Ball contact also changes the dynamics: pre-impact club speed should not be inferred from a peak several milliseconds into a collision or follow-through.

Force and Power Balances for a Connected Body

Start at Each Segment’s Center of Mass

For a rigid segment of mass $m$, central inertia tensor $I_G$, and center $G$, Newton–Euler equations are \[ m\mathbf a_G=\sum_k\mathbf F_k+m\mathbf g,\qquad \left.\frac{d\mathbf H_G}{dt}\right|_{\rm inertial}=\sum_k\bigl[(\mathbf r_k-\mathbf r_G)\times\mathbf F_k+\boldsymbol\tau_k\bigr]. \] In body-fixed principal coordinates the left angular term is $I_G\dot{\boldsymbol\omega}+\boldsymbol\omega\times(I_G\boldsymbol\omega)$. All vector components must be represented consistently. A scalar $I\ddot\theta=\tau$ is a special fixed-axis equation, not the general balance about a moving anatomical joint. Forces at both proximal and distal boundaries contribute moments about the center of mass.

The corresponding kinetic energy and rate are \[ T=\tfrac12m\|\mathbf v_G\|^2+\tfrac12\boldsymbol\omega^TI_G\boldsymbol\omega,\qquad \dot T=\sum_k(\mathbf F_k\cdot\mathbf v_k+\boldsymbol\tau_k\cdot\boldsymbol\omega)+m\mathbf g\cdot\mathbf v_G. \] Here $\mathbf v_k$ is the material-point velocity at the application point. A force can supply power to a translating segment even if the joint has no relative translation. Moving the point at which a wrench is represented changes its moment and the associated velocity together; the total wrench power stays the same. This is why a club calculation must retain grip force power.

Transfer, Generation, Storage, and Dissipation

At a coincident ideal joint, the two segments share the same point velocity. Equal and opposite joint forces then make powers that cancel when both bodies are included. They can nevertheless transfer energy across the segment boundary. For an applied joint couple $\boldsymbol\tau$ acting on the distal body and $-\boldsymbol\tau$ on the proximal body, the combined power is \[ P_j=\boldsymbol\tau\cdot(\boldsymbol\omega_d-\boldsymbol\omega_p). \] An ideal constraint couple does zero power along the allowed relative motion. An actuator couple need not. In a planar hinge $P_j=\tau_j\dot\phi_j$, where $\phi_j=\theta_d-\theta_p$. Calling every joint moment a “constraint torque” hides precisely the work we need to account for.

A spring and damper are another distinct case. Let $\delta=\theta_p-\theta_d-\delta_0$, and let $k\delta+c\dot\delta$ act positively on the distal rotor. The combined spring/damper power is \[ P_{p}+P_{d}=-(k\delta+c\dot\delta)\dot\delta =-\frac{d}{dt}\left(\tfrac12k\delta^2\right)-c\dot\delta^2. \] Thus proximal power loss need not equal distal power gain at that instant. With $k=100$ Nm/rad, $c=2$ Nm s/rad, $\delta=0.1$ rad, and angular speeds $5$ and $3$ rad/s, the two powers are $-70$ and $42$ W. The remaining $28$ W is $20$ W of storage and $8$ W of dissipation. This single example exposes the ambiguity in a claim that all proximal braking becomes distal acceleration.

Whole-System Equations Do Not Identify Individual Muscles

In local generalized coordinates with $v=\dot q$, a constrained mechanical model can be written \[ M(q)\dot v+h(q,v)=B(q)u+J(q)^T\lambda,\qquad J(q)v=0. \] The term $h$ includes the chosen velocity-dependent, gravitational, elastic, and dissipative contributions. The multipliers $\lambda$ are solved together with acceleration and constraint consistency. They are not an arbitrary derivative of a separate “constraint Lagrangian.” Multiplying a generalized-force residual by $M^{-1}$ yields acceleration, not torque.

Joint-coordinate mass matrices are generally coupled. A block-diagonal inertia representation is possible for separate bodies before connectivity is imposed; it does not make the assembled joint dynamics independent. An inverse-dynamics result is a model-dependent net generalized force. Many muscle-force combinations, including different amounts of co-contraction, may produce it. Segment kinematics alone do not select one biological realization.

What Coupled Oscillators Actually Predict

A Solvable Two-Rotor Counterexample

Consider two fixed-axis rotors with absolute angles $\theta_1,\theta_2$, inertias $I_1,I_2$, one torsional spring $k$, no gravity or damping, and a constant external torque $\tau$ on rotor 1. Set both angles and speeds to zero at $t=0$. These are not anatomical hip and shoulder equations; they isolate elastic coupling: \[ I_1\ddot\theta_1=\tau-k(\theta_1-\theta_2),\qquad I_2\ddot\theta_2=k(\theta_1-\theta_2). \] Define the inertia-weighted mean $\bar\theta=(I_1\theta_1+I_2\theta_2)/(I_1+I_2)$ and separation $\delta=\theta_1-\theta_2$. Then \[ \ddot{\bar\theta}=\frac{\tau}{I_1+I_2},\qquad \ddot\delta+\Omega^2\delta=\frac{\tau}{I_1},\qquad \Omega^2=k\left(\frac1{I_1}+\frac1{I_2}\right). \] The response combines an accelerating rigid mode and an oscillatory relative mode. It is not two independent oscillators with frequencies $\sqrt{k/I_i}$. For $I_1=2$, $I_2=0.5$ kg m$^2$, $k=100$ Nm/rad, and $\tau=50$ Nm, \[ \Omega=\sqrt{250}\ \mathrm{rad/s},\quad \ddot\theta_1=20+5\cos(\Omega t),\quad \ddot\theta_2=20[1-\cos(\Omega t)]. \] Rotor 1 accelerates positively throughout: its acceleration never falls below $15$ rad/s$^2$. It therefore has no interior speed peak, although rotor 2 has its first acceleration maximum at $t=\pi/\Omega\approx0.19869$ s. The first proximal speed maximum on $[0,1]$ s is simply the endpoint. Coupling alone has not produced the proposed universal peak order.

Integrating gives $\bar\theta=10t^2$ and $\delta=0.1[1-\cos(\Omega t)]$. The individual angles are $\theta_1=\bar\theta+0.2\delta$ and $\theta_2=\bar\theta-0.8\delta$. Their total kinetic-plus-spring energy equals $50\theta_1$ joules, the input work from rest. This is a useful independent check on signs and numerical integration. A different torque history, damping, initial preload, or terminal objective can yield different ordering; that is a result to compute, not an exception to hide.

Normal Modes Belong to the Coupled System

For a linear chain $M\ddot\theta+C\dot\theta+K\theta=\tau(t)$, undamped modes solve $K\psi=\omega^2M\psi$. Free overall rotation gives a zero-frequency mode when there is no spring to ground. With $M=\operatorname{diag}(2,1.2,0.4)$ kg m$^2$ and neighboring springs $120$ and $100$ Nm/rad, \[ K=\begin{bmatrix}120&-120&0\\-120&220&-100\\0&-100&100\end{bmatrix}. \] There is one rigid mode and two elastic modes. Each mode involves the chain, not a single segment. Relative-joint damping gives a similarly assembled matrix; terms $-c_i\dot\theta_i$ instead describe damping to a stationary reference. Neither model automatically imposes a quarter-period intersegment delay. Even a single oscillator’s phase depends on whether one compares force, displacement, velocity, or acceleration and whether the forcing is transient or periodic.

For golf, this means that effective inertia, stiffness, posture, muscle activation, and boundary conditions jointly shape the response. A smaller distal inertia alone does not prove a particular timing sequence. Linear modes are local approximations; large rotations, changing contacts, and activation-dependent stiffness can make those modes vary through a swing.

X-Factor: Relative Motion Before Biological Interpretation

Use a Declared Angular Convention

One workable scalar convention defines $\alpha_X=\theta_T-\theta_P$, with thorax and pelvis turn angles measured from address and positive toward the backswing. Then $\dot\alpha_X=\dot\theta_T-\dot\theta_P$. If the pelvis turns toward the target while the thorax is still turning away, the separation increases in this convention. At a smooth interior maximum the two turn rates are equal; no particular number of degrees of pelvic motion is required.

These turn angles must come from a stated projection or orientation decomposition. The angle between two rotation axes is a different quantity. For example, thorax and pelvis could rotate about parallel vertical axes while their turn angles differ substantially. A full relative orientation $R_P^TR_T$ contains three-dimensional information that a scalar X-factor does not. Marker placement, tilt, side bending, angle unwrapping, and handedness affect interpretation.

Define the reference event $t_0$ and an early-downswing window $[t_0,t_1]$ before computing stretch: \[ \Delta\alpha_X=\max_{t\in[t_0,t_1]}\alpha_X(t)-\alpha_X(t_0). \] Including $t_0$ makes this number nonnegative by definition; zero means no increase in that window. It does not prove that every golfer exhibits a positive stretch. A signed change between two specified events can instead be negative. Maximum pelvic and thoracic backswing positions need not occur simultaneously, and two zero angular velocities do not imply zero relative angle or zero elastic preload.

What the Classic Study Measured

Cheetham and colleagues compared 10 highly skilled and 9 less-skilled right-handed golfers using 5-irons and electromagnetic tracking. Their reported 30 Hz sampling and pelvis/T3 rotation measurements concern the transition region. They found a greater early-downswing X-factor increase in the higher-skill group, but no statistically significant between-group difference at the top of the backswing (Cheetham et al. 2001). This group comparison does not establish that increasing one player’s separation will increase distance.

The observed angle change also does not measure tissue strain energy, spindle response, or muscle-fiber shortening. For an assumed linear torsional spring, $U=\tfrac12k(\alpha_X-\alpha_0)^2$ is a model with a specified rest angle and stiffness. In a living trunk, distributed joints, muscle routing, tendon compliance, activation, and passive tissues must justify that approximation. An increased angle can alter storage and muscle operating conditions while also changing loads and the future club path. The measured association is a reason to investigate that interaction, not a warrant to maximize separation.

Wrist Motion and Work at the Grip

Lag Is Geometry, Not a Muscle State

In a planar approximation, define an absolute arm angle $\theta_a$ and absolute club angle $\theta_c$ from the same reference; their signed relative angle is $\phi=\theta_c-\theta_a$. An included angle from directed forearm and shaft vectors is another convention and can differ by a supplement. State the directions and zero position. Three-dimensional wrist motion includes more than a single lag coordinate, and a flexible shaft adds deformation. A projected angle does not uniquely specify the clubface orientation.

Neither hand speed nor clubhead speed determines muscle shortening speed. A muscle-tendon length $\ell(q)$ has rate $\dot\ell=(\partial\ell/\partial q)\dot q$; fiber speed additionally depends on tendon compliance and pennation. Classical force–velocity behavior concerns the contractile element under declared activation and length conditions. It cannot be applied by converting an absolute club speed directly into a universal wrist torque limit. Active lengthening and isometric force are also different operating conditions from concentric shortening.

Sprigings and Neal’s planar three-segment simulation found that suitably timed active wrist torque could increase clubhead speed by about 9 percent in its modeled conditions (Sprigings and Neal 2000). That is a simulation result with particular torque generators and constraints, not an experimentally established instruction to add a late wrist action. It is nevertheless enough to reject the universal claim that wrist torque cannot contribute at high club speed. Conversely, a measured small net wrist moment does not prove that all wrist muscles are inactive.

The Club Receives a Wrench, Not Only a Torque

For a rigid club, with grip point $H$, external grip force $\mathbf F_H$, and grip couple $\boldsymbol\tau_H$, its pre-impact power balance is \[ \frac{d}{dt}(T_c+U_{g,c})= \mathbf F_H\cdot\mathbf v_H+\boldsymbol\tau_H\cdot\boldsymbol\omega_c+P_{\rm other}. \] The term $P_{\rm other}$ accounts for effects such as aerodynamic loading. For a flexible club, include shaft strain energy and dissipation and use the grip section’s angular velocity for its applied couple. The power entering the club is not the same as the combined power of an anatomical wrist actuator acting between two moving bodies.

Even when $\boldsymbol\tau_H=0$, a translating grip can do work through $\mathbf F_H\cdot\mathbf v_H$. The two hands may also produce a resultant couple through their separated forces. A zero net couple at one chosen port does not remove force transmission or make the club an isolated body. Nesbit and Serrano’s analysis explicitly modeled both force and torque work at the club interface (Nesbit and Serrano 2005). The balance here follows rigid-body mechanics independently of any biological interpretation of their reconstructed loads.

A Release Model That Retains the Moving Grip

Derive the Unactuated Equation

Use an inertial vertical plane with horizontal $x$ and upward $y$. The rigid club’s center is $\mathbf r_G=\mathbf r_H+\ell\mathbf e$, where $\mathbf e=(\cos\theta,\sin\theta)$ points from grip to center and $\mathbf e_\perp=(-\sin\theta,\cos\theta)$. Let $m$ be mass, $I_G$ central inertia, and $I_H=I_G+m\ell^2$. Eliminating the grip force between translation and rotation gives \[ I_H\ddot\theta=\tau_H+m\ell(\mathbf g-\mathbf a_H)\cdot\mathbf e_\perp. \] To see the origin of the forcing, substitute $\mathbf a_G=\mathbf a_H+\ell\ddot\theta\mathbf e_\perp-\ell\dot\theta^2\mathbf e$ into $m\mathbf a_G=\mathbf F_H+m\mathbf g$, then take moments about $G$. The radial $\dot\theta^2$ term has no moment about the grip; the acceleration of the grip has a tangential projection that does. Omitting $\mathbf a_H$ changes the problem after “release.”

For an illustrative instantaneous state, set $m=0.2$ kg, $\ell=1$ m, $I_G=0.02$ kg m$^2$, $\theta=0$, $\dot\theta=2$ rad/s, $\mathbf v_H=(0,-3)$ m/s, $\mathbf a_H=(0,-20)$ m/s$^2$, and $\tau_H=0$. Then $I_H=0.22$ kg m$^2$ and $\ddot\theta\approx9.2636$ rad/s$^2$. The grip force is approximately $(-0.8,-0.18527)$ N and supplies $0.55582$ W. Directly differentiating $T_c+U_{g,c}$ gives the same rate. Angular acceleration is present with no applied grip couple because the grip trajectory is still imposed and its force still acts.

Changing only grip acceleration to free fall, $\mathbf a_H=\mathbf g$, gives zero angular acceleration. A stationary grip instead gives $\ddot\theta\approx-8.9182$ rad/s$^2$ at the same horizontal orientation. These are three different boundary conditions, not three explanations of the same swing. Centrifugal or Coriolis terms in other coordinates reorganize the dynamics; they do not introduce an additional energy source.

Holding, Releasing, and Feasibility

Suppose an idealized holding condition imposes $\theta=\theta_a(t)+\phi_0$. The required grip couple is \[ \tau_{\rm hold}=I_H\ddot\theta_a-m\ell(\mathbf g-\mathbf a_H)\cdot\mathbf e_\perp. \] A bound $|\tau_{\rm hold}|\le\tau_{\max}$ checks whether this imposed trajectory is feasible in the toy model. It does not assert that a human wrist locks perfectly until its torque limit is reached. A controller could change its command earlier; anatomical joints need not behave as binary latches. For a smooth release without an impulse, initialize the released motion with the angle and angular velocity immediately before release and continue the specified grip trajectory. If a constraint change produces an impulse, its momentum and energy effects require a separate event model.

A claim that later release improves speed must specify which intervention changes, which upstream controls or trajectories remain fixed, the admissible torque/force and joint-range limits, and the endpoint. Compare states at a common impact surface with correct clubface orientation, not just at an arbitrary common clock time. If grip motion is prescribed, its force and work generally change between alternatives. Giving both alternatives the same torque history does not make their input work equal. There is no universal percentage benefit or optimal delay that follows from this pendulum equation alone.

Ground Reaction: Momentum Channel and Contact Constraint

Compute the Full Moment Vector

Choose a right-handed frame: $x$ along the target line, $y$ horizontally across it, and $z$ upward. About origin $O$, a force applied at displacement $\mathbf r=(r_x,r_y,r_z)$ and a free moment $\mathbf m$ produce \[ \boldsymbol\tau_O=\mathbf r\times\mathbf F+\mathbf m =\begin{bmatrix}r_yF_z-r_zF_y\\r_zF_x-r_xF_z\\r_xF_y-r_yF_x\end{bmatrix}+\mathbf m. \] Vertical force at a horizontal offset produces horizontal-axis moments, not a vertical-axis yaw moment. Horizontal force at a vertical offset contributes horizontal-axis moments too. Summing those components as if all were pelvic yaw torque is invalid.

For a numerical example about a stated origin, let $\mathbf r=(0.4,0.2,-1)$ m and $F_z=1.5(80)(9.81)=1177.2$ N. Vertical force alone gives $(235.44,-470.88,0)$ Nm. Add $F_x=100$ N, $F_y=200$ N, and a free yaw moment $m_z=10$ Nm: the complete moment, in Nm, is \[ \boldsymbol\tau_O=(435.44,-570.88,70)^T. \] If that net yaw moment stays constant for $0.05$ s, its yaw angular impulse is $3.5$ N m s. These are declared illustrative inputs; the calculation is not a prediction of hip angular speed.

Whole-Body Angular Momentum Is Not Pelvic Rotation

For the complete golfer-plus-club system about a fixed inertial origin, \[ \dot{\mathbf H}_O=\sum\boldsymbol\tau_{O,\rm external}. \] Joint forces and couples are internal to that system. Ground wrenches, gravity, aerodynamic loads, and ball contact where present are external. About the total center of mass, uniform gravity has zero resultant moment, and the external-moment balance also holds for angular momentum defined relative to that center. Other moving reference points require the appropriate transport terms. The force example above gives a change in whole-system angular momentum only when all other relevant external moments are accounted for.

One cannot divide that moment by an isolated pelvic inertia to obtain pelvic acceleration in a coupled swing. Nor does zero external yaw moment prevent the pelvis from rotating: internal actions can produce counter-rotation elsewhere while conserving total angular momentum. Ground contact permits exchange of momentum with the environment and changes feasible motion; it is not the sole possible cause of every observed segment rotation.

Center of Pressure and Ground Work

A force plate’s resultant force, moment, and center of pressure are related but not interchangeable. The center of pressure describes a pressure-resultant location on the contact plane under its adopted convention. It does not by itself determine horizontal shear force or the residual free yaw moment. Near vanishing normal force, center-of-pressure estimates can become ill-conditioned. Combining two plates also requires a common origin and frame.

At a perfectly stationary, non-slipping contact, ideal reaction forces do zero mechanical work because the material contact points have zero velocity. Yet they can provide a nonzero moment and angular impulse. Muscles supply mechanical work internally from chemical energy while contact constrains the motion. Real shoes, ground compliance, slipping, and foot deformation can add storage or work terms. A peak ground force before impact does not mean zero force at impact, and it does not determine when the center-of-mass speed peaks. Timing force, impulse, work, and segment velocity answers four different questions.

Close the Energy Budget Before Naming Efficiency

Choose the System and the Time Interval

For a pre-impact mechanical model including the body, club, gravity, and chosen elastic elements, write \[ E=T_{\rm body}+T_{\rm club}+U_g+U_e,\qquad E(t_I)-E(t_0)=W_{\rm act}+W_{\rm ext}-D. \] Here $W_{\rm act}$ is signed actuator work at the declared joint ports, $W_{\rm ext}$ is work from the other external interactions not already represented by potential energy, and $D\ge0$ is the modeled dissipation. Avoid counting an elastic element both in $U_e$ and again as an independent actuator energy input. An ideal frictionless model with ongoing positive actuator work does not conserve mechanical energy; it obeys this balance with increasing $E$.

Backswing history enters through initial position, velocity, and stored energy. The top of a backswing is not generally a state of complete rest. Gravity needs a scale check: raising an 80 kg whole-body center of mass by $0.1$ m changes potential energy by $78.48$ J. Individual segment and club height changes must also be handled consistently. Small-looking displacement is not a sufficient argument to delete the term.

Energy of a segment depends on both translation and rotation. A decrease in angular speed about one measured axis can coexist with increasing segment kinetic energy. Even a verified decrease in total segment energy does not identify its destination without boundary powers and storage. Negative net joint power also does not uniquely identify eccentric action in every spanning muscle. Tendon energy changes and simultaneous muscle actions can yield the same net value.

Signed Work, Positive Work, and Metabolism

For an actuator power $P_j$, distinguish signed work $W_j=\int P_jdt$ from positive work $W_j^+=\int\max(P_j,0)dt$ and absorbed-work magnitude $W_j^-=\int\max(-P_j,0)dt$, so $W_j=W_j^+-W_j^-$. Summing positive work over joints is not the same as taking the positive part after summing their powers. State which denominator a reported ratio uses.

Nesbit and Serrano analyzed one selected swing from each of four amateur golfers with reconstructed body and club models. Their “swing efficiency” was club work divided by total body joint work; their table reports about 20.2–26.8 percent (Nesbit and Serrano 2005). Those values are not a measured elite population range, a direct muscle efficiency, or a fraction of metabolic energy. A larger ratio also need not imply a faster club: numerator and denominator can change together.

For a transparent illustrative budget, suppose initial recoverable mechanical energy is $100$ J, positive actuator work is $500$ J, absorbed actuator work is $80$ J, and other dissipation is $20$ J, with zero other external work. Final mechanical energy is $500$ J. It could comprise $250$ J in the club and $250$ J in body motion and retained potentials. Club energy divided by positive actuator work is $0.50$, while club energy divided by signed actuator work is about $0.595$. Neither ratio alone identifies where the rest became heat or establishes metabolic efficiency. Initial energy can even make a club-energy/work ratio exceed one without violating conservation.

Collision adds another boundary and another interval. Pre-impact club kinetic energy is not all transferred to the ball; club motion, deformation, losses, and ball spin remain relevant. A numerical impact-transfer fraction must follow an explicit collision model or measurement. It cannot be appended as a universal percentage to close an incomplete downswing budget.

How Drift and Control Connect to the Chain

Declare What Zero Input Removes

The full local state must at least include configuration and velocity, $x=(q,v)$. With a direct torque input and consistently eliminated regular constraints, it may admit \[ \dot x=f(x)+G(x)u,\qquad f(x)=\begin{bmatrix}v\\a_0(q,v)\end{bmatrix}. \] The first component is configuration rate, not the complete dynamics. The drift is the zero-input field for the declared input port. Its contents depend on which gravity, elastic, dissipative, actuator, and contact effects the model retains. A change of coordinates transforms both the drift and the input fields; numerical component magnitudes need a declared scaling or metric.

If input means neural excitation, zero input need not mean zero activation or zero muscle force. Activation and tendon states must then be included, and a simple affine torque model may no longer represent the same intervention. If input means a net joint torque, removing it also differs from making all muscles inactive. In a moving-grip club model, setting its grip couple to zero retains the imposed grip path and its force. These distinctions are essential to the zero-torque counterfactual framework in the zero-torque counterfactual chapter.

Attribution and Later Consequences

At a fixed state in a genuinely affine model, $G(x)u$ is the instantaneous difference between actual and zero-input state rates. It does not follow that the difference between two finite trajectories equals its integral along the original trajectory. Once states diverge, gravity, geometry, constraint reactions, and future control effectiveness change too. A counterfactual that crosses a contact boundary requires an event-consistent model.

The first-order endpoint sensitivity along a nominal smooth trajectory is \[ \delta x(t_I)=\Phi(t_I,t_0)\delta x(t_0) +\int_{t_0}^{t_I}\Phi(t_I,s)G(x(s))\delta u(s)\,ds, \] where $\Phi$ is the state-transition matrix of the full linearized dynamics under the specified control policy. For a terminal output $y=h(x)$ at a fixed time, premultiply by $Dh$. If impact time is determined by a contact event, include its timing sensitivity and any impact map rather than silently treating $t_I$ as fixed.

This gives the site’s central argument a precise form: early action changes the state and geometry from which later action operates, and the existing state continues to evolve between inputs. The relevant question is how much admissible future action can change the required impact outputs, with what uncertainty and cost. A large instantaneous drift-control ratio does not measure neural latency, prove an optimal policy, or establish absence of useful distal control. It needs a declared model, coordinate metric, input port, and output horizon before it can inform that question.

Test the Explanation

Link Each Claim to an Observable

To test sequencing, collect synchronized segment kinematics with declared coordinate frames and timing uncertainty. To test power flow, add external wrenches, body/club inertial estimates, and an inverse-dynamics model whose residuals and filtering sensitivity are reported. To infer individual muscle or tissue action, additional measurements and physiological assumptions are necessary. Electromyography measures an electrical signal related to activation; it does not directly measure muscle force, joint work, or tendon energy.

Check integrated boundary powers against the change in mechanical energy, allowing for measurement uncertainty. Large unexplained residuals can arise from differentiation noise, mismatched time bases, inaccurate inertias, omitted foot or shaft deformation, or inconsistent moment references. Reproducing measured positions in a kinematically driven model is not an independent validation of all its internal forces.

Test Interventions Against Competing Explanations

Suppose a player gains club speed after changing transition timing. Candidate explanations include different active work, a changed hand path, different segment geometry, altered elastic storage, a different impact time, or changed clubface/contact conditions. Compare repeated trials and quantify the outputs that matter: speed, path, face orientation, strike location, ball result, and consistency. A faster isolated segment peak is not enough.

Use a model intervention to separate hypotheses only after declaring the preserved quantities. Holding the grip path fixed isolates one question; holding actuator effort fixed isolates another. Sensitivity to initial state and model parameters determines whether the predicted advantage survives measurement uncertainty. A simulation that matches an observed sequence can demonstrate compatibility with a mechanism, but competing mechanisms may match the same sequence. Successful explanation requires discriminating observations or interventions.

What the Reader Can Use in Practice

The mechanics support an investigation organized around the desired club state and the player’s repeatability. They do not justify a universal instruction to maximize pelvic speed, freeze the wrists, or seek the largest possible X-factor. A change should be evaluated by its measured effect on the complete shot, including dispersion and strike, rather than by conformity to an attractive sequence diagram.

Slow practice can help a player explore positions and coordination, but it is not automatically dynamically similar to a full-speed swing. Under a time scaling, inertial terms scale differently from gravity, and activation, elastic, and contact effects need not scale together. The motion should be reassessed at the intended speed. Likewise, smoother commanded torque does not guarantee small club jerk near rapidly changing geometry or contact events.

Force-plate feedback can inform contact forces, moments, impulse, and their timing when measured correctly. It does not directly calibrate a drift-control ratio. Feeling “effortless” does not prove that muscles stopped working, and muscle activation does not identify concentric action; active muscle can shorten, remain approximately isometric, or lengthen. These distinctions allow useful coaching descriptions to coexist with technically honest explanations.

The connections are the important result. Ground contact shapes feasible momentum exchange; muscle action changes energy and state; geometry maps that state to club motion; elastic elements store and return work; and the remaining time constrains what later control can accomplish. An account that quantifies those connections is more informative than either a list of peak timings or a claim that one part of the chain supplies all the speed.

Reproduce the Mechanical Checks

The following Python functions use NumPy and SciPy to evaluate the stated teaching cases. Times are in seconds and other inputs use SI units. The rotor solution can be checked against independent numerical integration; the modal vectors are mass-normalized. The moving-grip calculation checks instantaneous force, moment, and total mechanical-energy rates, not a complete golf trajectory. The same program is printed in both editions.

# Kinetic-chain mechanical checks
import numpy as np
from scipy.linalg import eigh


def rotor_case(time: np.ndarray) -> dict[str, np.ndarray]:
    """Two fixed-axis rotors, zero initial state, constant 50 Nm input."""
    frequency = np.sqrt(250.0)
    phase = frequency * time
    separation = 0.1 * (1 - np.cos(phase))
    relative_speed = 0.1 * frequency * np.sin(phase)
    mean_angle = 10 * time**2
    mean_speed = 20 * time
    angle_one = mean_angle + 0.2 * separation
    angle_two = mean_angle - 0.8 * separation
    speed_one = mean_speed + 0.2 * relative_speed
    speed_two = mean_speed - 0.8 * relative_speed
    energy = speed_one**2 + 0.25 * speed_two**2
    energy += 50 * separation**2
    return {
        "state": np.array([angle_one, angle_two, speed_one, speed_two]),
        "acceleration": np.array([20 + 5 * np.cos(phase), 20 * (1 - np.cos(phase))]),
        "energy": energy,
    }


def mode_case() -> dict[str, np.ndarray]:
    """Mass-normalized modes of the stated three-rotor model."""
    mass = np.diag([2.0, 1.2, 0.4])
    stiffness = np.array([[120, -120, 0], [-120, 220, -100], [0, -100, 100]])
    squared, modes = eigh(stiffness, mass)
    return {"squared_frequencies": squared, "modes": modes}


def grip_case(acceleration: tuple[float, float]) -> dict:
    """Horizontal rod, zero grip couple, translating/accelerating grip."""
    mass, length, central_inertia = 0.2, 1.0, 0.02
    gravity = np.array([0.0, -9.81])
    grip_velocity = np.array([0.0, -3.0])
    angular_speed = 2.0
    polar_inertia = central_inertia + mass * length**2
    alpha = mass * length * (gravity[1] - acceleration[1]) / polar_inertia
    center_acceleration = np.array(acceleration) + length * np.array([-4.0, alpha])
    force = mass * (center_acceleration - gravity)
    center_velocity = grip_velocity + np.array([0.0, length * angular_speed])
    kinetic_rate = mass * center_velocity @ center_acceleration
    kinetic_rate += central_inertia * angular_speed * alpha
    return {
        "alpha": alpha, "force": force,
        "grip_power": force @ grip_velocity,
        "energy_rate": kinetic_rate - mass * gravity @ center_velocity,
    }


def ground_case() -> dict:
    """Moments about the declared origin; all inputs are illustrative SI values."""
    arm = np.array([0.4, 0.2, -1.0])
    vertical_force = np.array([0.0, 0.0, 1.5 * 80 * 9.81])
    force = vertical_force + np.array([100.0, 200.0, 0.0])
    moment = np.cross(arm, force) + np.array([0.0, 0.0, 10.0])
    return {
        "vertical_moment": np.cross(arm, vertical_force),
        "full_moment": moment, "yaw_impulse": moment[2] * 0.05,
    }


def joint_case() -> dict[str, float]:
    """A spring/damper stores and dissipates work while transferring power."""
    separation, relative_speed = 0.1, 2.0
    stiffness, damping = 100.0, 2.0
    torque = stiffness * separation + damping * relative_speed
    return {
        "proximal_power": -torque * 5.0,
        "distal_power": torque * 3.0,
        "storage_rate": stiffness * separation * relative_speed,
        "dissipation": damping * relative_speed**2,
    }

For example, evaluate the rotor function at a time grid from zero to one second and compare its energy with $50\theta_1$. For the modal case, verify $K\Psi=M\Psi\operatorname{diag}(\omega^2)$ and $\Psi^TM\Psi=I$; the smallest squared frequency is zero up to floating-point roundoff. Evaluate the grip case with accelerations $(0,-20)$, $(0,-9.81)$, and $(0,0)$ to reproduce the three boundary conditions. Ground and joint functions reproduce the moment and power examples directly. Small rounding differences should not be interpreted as physical residuals.

Exercises

Sequence Counterexample. Starting from the two-rotor equations, derive the mean and relative solutions and reconstruct both angular velocities. On $[0,1]$ s, locate rotor 1’s maximum velocity and rotor 2’s first acceleration maximum. Explain why their order refutes a universal proximal-speed/distal-acceleration rule without refuting every possible sequencing mechanism. Verify the input-work identity independently.
X-Factor and Storage. In the declared backswing-positive convention, let thorax and pelvis angles at the reference event be $90$ and $45$ degrees. At a later event let them be $88$ and $38$ degrees. Compute separation at each event and its change. For an assumed torsional spring with rest separation $30$ degrees and stiffness $100$ Nm/rad, compute the storage change in joules. State why this result is not a measurement of human trunk energy and why both turn rates becoming zero would not remove preload.
A Moving-Grip Release. Derive the club equation by eliminating grip force from Newton–Euler balances. Reproduce the three acceleration cases and check energy power. Then prescribe a smooth grip trajectory and an arm angle history of your choosing, state them explicitly, and compute the holding couple. Explain the additional endpoint, feasibility, and input-work checks needed before comparing two release times. Zero grip couple alone is not a complete boundary condition.
Ground Wrench and Reference Point. Reproduce all three moment components in the ground example. Change the origin by a known displacement and transform the moment consistently. Integrate the original yaw moment over the stated interval. Identify the additional information needed to infer pelvic acceleration from whole-system angular impulse, and explain why center of pressure alone cannot supply it.
Energy and Identifiability. Close the illustrative $100$ J initial-energy budget using the stated positive work, absorbed work, and dissipation. Compute both club-energy ratios. Construct a second consistent budget with the same final club energy but different body energy and actuator work. Explain why club speed alone cannot distinguish the budgets or determine individual muscle action. Add a separately declared collision model only if estimating ball energy.
Modes and Experimental Discrimination. Solve the three-rotor generalized eigenproblem, identify its rigid mode, and verify mass normalization. Apply a finite torque pulse to one rotor and calculate the response from rest. Repeat with relative-joint damping and changed initial preload. Compare the resulting peak definitions explicitly. Propose a synchronized measurement that could distinguish a change in active work from a change in energy redistribution in a golfer; report which required quantities remain model-dependent.

On This Site

Proximal-to-Distal Energy Transfer in the Golf Swing — Sequential timing, power ports, and chain coordination
A Journey Through the Swing — Narrative companion to kinetic-chain mechanics

Cite this page

References

Cheetham, P. J., P. E. Martin, R. E. Mottram, and B. F. St Laurent. 2001. “The Importance of Stretching the X-Factor in the Downswing of Golf: The X-Factor Stretch.” In Optimising Performance in Golf, edited by P. R. Thomas. Australian Academic Press. https://www.philcheetham.com/wp-content/uploads/2011/11/Stretching-the-X-Factor-Paper.pdf.

Nesbit, Steven M., and Monika Serrano. 2005. “Work and Power Analysis of the Golf Swing.” Journal of Sports Science and Medicine 4: 520–33. https://www.jssm.org/volume04/iss4/cap/jssm-04-520.pdf.

Sprigings, Eric J., and Robert J. Neal. 2000. “An Insight into the Importance of Wrist Torque in Driving the Golfball: A Simulation Study.” Journal of Applied Biomechanics 16 (4): 356–66. https://doi.org/10.1123/jab.16.4.356.

Vena, Alessandro, Dave Budney, Tom Forest, and Jason P. Carey. 2011. “Three-Dimensional Kinematic Analysis of the Golf Swing Using Instantaneous Screw Axis Theory, Part 2: Golf Swing Kinematic Sequence.” Sports Engineering 13 (3): 125–33. https://doi.org/10.1007/s12283-010-0059-7.

Citation

BibTeX citation: Download .bib

@online{untitled,
  author = {},
  url = {https://affinedrift.com/articles/The_Physics_of_Golf/quarto/ch30_kinetic_chain.html},
  langid = {en}
}

For attribution, please cite this work as:

n.d. https://affinedrift.com/articles/The_Physics_of_Golf/quarto/ch30_kinetic_chain.html.

--- pagetitle: "The Kinetic Chain: Motion, Work, and Control in the Golf Swing" description: "A rigorous account of sequencing, joint work, moving-grip release, ground reactions, and finite-horizon control in the golf swing." css: ../../../css/kinetic-chain.css --- # The Kinetic Chain: Motion, Work, and Control in the Golf Swing {#sec-kinetic_chain} The central question of the kinetic chain is how a golfer's coordinated motion produces a useful club state at impact. The answer involves muscle work, contact forces, changing geometry, inertia, elastic storage, and timing. None of these can replace the others. A pelvis that slows while the club accelerates is an observation worth explaining; it is not, by itself, a measurement of energy moving from pelvis to club. This chapter develops the balances needed to make that explanation testable. The useful intuition behind the whip analogy is that a connected system can develop high distal speed without every part moving equally fast. Its limitation is equally important: a golfer has active joints, two hands on a club, two changing foot contacts, and a three-dimensional body. A prescribed sequence of peaks is not a mechanical law for that system. We will use simple rotors and a moving-grip club to isolate mechanisms, then reconnect those mechanisms to measurements and control. The examples are deliberately specified teaching models, not fitted swings or prescriptions for an athlete. ## Define the Observation Before Explaining It ### Segments, Joints, and Comparable Rates A segment is a body with a defined mass, center of mass, and inertia. A joint is a connection with relative motion and transmitted force or moment. A wrist is a joint region, not an additional rigid body whose energy can simply be added to forearm and hand energy. Similarly, a measured shoulder-line orientation may describe the thorax; counting both as separate energy reservoirs can count the same mass twice. A whole-body energy calculation requires disjoint segment definitions. For a segment, the angular velocity vector describes rotation relative to a declared frame. A joint rate describes relative motion between adjacent segments. Neither is clubhead linear speed. The relation between clubhead velocity and generalized velocity is a Jacobian relation, $\mathbf v_C=J_C(q)\dot q$, including base translation when the base moves. Changing posture changes this map. Comparing a pelvic rate in degrees per second with clubhead speed in miles per hour does not establish amplification without geometry and a unit conversion. Three-dimensional orientation also requires care. Euler-angle derivatives generally are not the Cartesian components of angular velocity. If $\boldsymbol\omega=E(\theta)\dot\theta$, a physical moment $\boldsymbol\tau$ has generalized force $E^T\boldsymbol\tau$, so that power is preserved. Angular-speed magnitude, a selected angular-velocity component, and the derivative of a projected turn angle can have different peak times. ### A Sequence Is a Measurement Result Define downswing onset, the instant immediately before ball contact, the measured quantity, its sign convention, smoothing, and the method for finding peaks. If $t_0$ and $t_I$ are those events, normalized time is $s=(t-t_0)/(t_I-t_0)$. A peak at a percentage of downswing time is meaningful only under that event definition. Finite sampling and filtering create timing uncertainty; a broad plateau should not be represented as a precisely ordered instant. Peak segment speed and peak acceleration are distinct events. For a smooth signed velocity, an interior velocity extremum has zero acceleration; an acceleration extremum concerns its derivative. Nothing makes the peak speed of one segment equivalent to the peak acceleration of its neighbor. A speed magnitude introduces another distinction: $d\|\boldsymbol\omega\|/dt=\boldsymbol\omega\cdot\dot{\boldsymbol\omega}/\|\boldsymbol\omega\|$ away from zero speed. Vena and colleagues' five-subject study, using instantaneous screw-axis angular velocities, reported support for a proximal-to-distal kinematic sequence in two subjects, rather than a universal order [@vena2011kinematic]. That result does not say sequencing is useless. It says a proposed invariant sequence must survive explicit measurement choices and between-player variation. We therefore do not assign universal pelvic, thoracic, arm, or wrist peak percentages here. Ball contact also changes the dynamics: pre-impact club speed should not be inferred from a peak several milliseconds into a collision or follow-through. ## Force and Power Balances for a Connected Body ### Start at Each Segment's Center of Mass For a rigid segment of mass $m$, central inertia tensor $I_G$, and center $G$, Newton–Euler equations are $$ m\mathbf a_G=\sum_k\mathbf F_k+m\mathbf g,\qquad \left.\frac{d\mathbf H_G}{dt}\right|_{\rm inertial}=\sum_k\bigl[(\mathbf r_k-\mathbf r_G)\times\mathbf F_k+\boldsymbol\tau_k\bigr]. $$ In body-fixed principal coordinates the left angular term is $I_G\dot{\boldsymbol\omega}+\boldsymbol\omega\times(I_G\boldsymbol\omega)$. All vector components must be represented consistently. A scalar $I\ddot\theta=\tau$ is a special fixed-axis equation, not the general balance about a moving anatomical joint. Forces at both proximal and distal boundaries contribute moments about the center of mass. The corresponding kinetic energy and rate are $$ T=\tfrac12m\|\mathbf v_G\|^2+\tfrac12\boldsymbol\omega^TI_G\boldsymbol\omega,\qquad \dot T=\sum_k(\mathbf F_k\cdot\mathbf v_k+\boldsymbol\tau_k\cdot\boldsymbol\omega)+m\mathbf g\cdot\mathbf v_G. $$ Here $\mathbf v_k$ is the material-point velocity at the application point. A force can supply power to a translating segment even if the joint has no relative translation. Moving the point at which a wrench is represented changes its moment and the associated velocity together; the total wrench power stays the same. This is why a club calculation must retain grip force power. ### Transfer, Generation, Storage, and Dissipation At a coincident ideal joint, the two segments share the same point velocity. Equal and opposite joint forces then make powers that cancel when both bodies are included. They can nevertheless transfer energy across the segment boundary. For an applied joint couple $\boldsymbol\tau$ acting on the distal body and $-\boldsymbol\tau$ on the proximal body, the combined power is $$ P_j=\boldsymbol\tau\cdot(\boldsymbol\omega_d-\boldsymbol\omega_p). $$ An ideal constraint couple does zero power along the allowed relative motion. An actuator couple need not. In a planar hinge $P_j=\tau_j\dot\phi_j$, where $\phi_j=\theta_d-\theta_p$. Calling every joint moment a “constraint torque” hides precisely the work we need to account for. A spring and damper are another distinct case. Let $\delta=\theta_p-\theta_d-\delta_0$, and let $k\delta+c\dot\delta$ act positively on the distal rotor. The combined spring/damper power is $$ P_{p}+P_{d}=-(k\delta+c\dot\delta)\dot\delta =-\frac{d}{dt}\left(\tfrac12k\delta^2\right)-c\dot\delta^2. $$ Thus proximal power loss need not equal distal power gain at that instant. With $k=100$ Nm/rad, $c=2$ Nm s/rad, $\delta=0.1$ rad, and angular speeds $5$ and $3$ rad/s, the two powers are $-70$ and $42$ W. The remaining $28$ W is $20$ W of storage and $8$ W of dissipation. This single example exposes the ambiguity in a claim that all proximal braking becomes distal acceleration. ### Whole-System Equations Do Not Identify Individual Muscles In local generalized coordinates with $v=\dot q$, a constrained mechanical model can be written $$ M(q)\dot v+h(q,v)=B(q)u+J(q)^T\lambda,\qquad J(q)v=0. $$ The term $h$ includes the chosen velocity-dependent, gravitational, elastic, and dissipative contributions. The multipliers $\lambda$ are solved together with acceleration and constraint consistency. They are not an arbitrary derivative of a separate “constraint Lagrangian.” Multiplying a generalized-force residual by $M^{-1}$ yields acceleration, not torque. Joint-coordinate mass matrices are generally coupled. A block-diagonal inertia representation is possible for separate bodies before connectivity is imposed; it does not make the assembled joint dynamics independent. An inverse-dynamics result is a model-dependent net generalized force. Many muscle-force combinations, including different amounts of co-contraction, may produce it. Segment kinematics alone do not select one biological realization. ## What Coupled Oscillators Actually Predict ### A Solvable Two-Rotor Counterexample Consider two fixed-axis rotors with absolute angles $\theta_1,\theta_2$, inertias $I_1,I_2$, one torsional spring $k$, no gravity or damping, and a constant external torque $\tau$ on rotor 1. Set both angles and speeds to zero at $t=0$. These are not anatomical hip and shoulder equations; they isolate elastic coupling: $$ I_1\ddot\theta_1=\tau-k(\theta_1-\theta_2),\qquad I_2\ddot\theta_2=k(\theta_1-\theta_2). $$ Define the inertia-weighted mean $\bar\theta=(I_1\theta_1+I_2\theta_2)/(I_1+I_2)$ and separation $\delta=\theta_1-\theta_2$. Then $$ \ddot{\bar\theta}=\frac{\tau}{I_1+I_2},\qquad \ddot\delta+\Omega^2\delta=\frac{\tau}{I_1},\qquad \Omega^2=k\left(\frac1{I_1}+\frac1{I_2}\right). $$ The response combines an accelerating rigid mode and an oscillatory relative mode. It is not two independent oscillators with frequencies $\sqrt{k/I_i}$. For $I_1=2$, $I_2=0.5$ kg m$^2$, $k=100$ Nm/rad, and $\tau=50$ Nm, $$ \Omega=\sqrt{250}\ \mathrm{rad/s},\quad \ddot\theta_1=20+5\cos(\Omega t),\quad \ddot\theta_2=20[1-\cos(\Omega t)]. $$ Rotor 1 accelerates positively throughout: its acceleration never falls below $15$ rad/s$^2$. It therefore has no interior speed peak, although rotor 2 has its first acceleration maximum at $t=\pi/\Omega\approx0.19869$ s. The first proximal speed maximum on $[0,1]$ s is simply the endpoint. Coupling alone has not produced the proposed universal peak order. Integrating gives $\bar\theta=10t^2$ and $\delta=0.1[1-\cos(\Omega t)]$. The individual angles are $\theta_1=\bar\theta+0.2\delta$ and $\theta_2=\bar\theta-0.8\delta$. Their total kinetic-plus-spring energy equals $50\theta_1$ joules, the input work from rest. This is a useful independent check on signs and numerical integration. A different torque history, damping, initial preload, or terminal objective can yield different ordering; that is a result to compute, not an exception to hide. ### Normal Modes Belong to the Coupled System For a linear chain $M\ddot\theta+C\dot\theta+K\theta=\tau(t)$, undamped modes solve $K\psi=\omega^2M\psi$. Free overall rotation gives a zero-frequency mode when there is no spring to ground. With $M=\operatorname{diag}(2,1.2,0.4)$ kg m$^2$ and neighboring springs $120$ and $100$ Nm/rad, $$ K=\begin{bmatrix}120&-120&0\\-120&220&-100\\0&-100&100\end{bmatrix}. $$ There is one rigid mode and two elastic modes. Each mode involves the chain, not a single segment. Relative-joint damping gives a similarly assembled matrix; terms $-c_i\dot\theta_i$ instead describe damping to a stationary reference. Neither model automatically imposes a quarter-period intersegment delay. Even a single oscillator's phase depends on whether one compares force, displacement, velocity, or acceleration and whether the forcing is transient or periodic. For golf, this means that effective inertia, stiffness, posture, muscle activation, and boundary conditions jointly shape the response. A smaller distal inertia alone does not prove a particular timing sequence. Linear modes are local approximations; large rotations, changing contacts, and activation-dependent stiffness can make those modes vary through a swing. ## X-Factor: Relative Motion Before Biological Interpretation ### Use a Declared Angular Convention One workable scalar convention defines $\alpha_X=\theta_T-\theta_P$, with thorax and pelvis turn angles measured from address and positive toward the backswing. Then $\dot\alpha_X=\dot\theta_T-\dot\theta_P$. If the pelvis turns toward the target while the thorax is still turning away, the separation increases in this convention. At a smooth interior maximum the two turn rates are equal; no particular number of degrees of pelvic motion is required. These turn angles must come from a stated projection or orientation decomposition. The angle between two rotation axes is a different quantity. For example, thorax and pelvis could rotate about parallel vertical axes while their turn angles differ substantially. A full relative orientation $R_P^TR_T$ contains three-dimensional information that a scalar X-factor does not. Marker placement, tilt, side bending, angle unwrapping, and handedness affect interpretation. Define the reference event $t_0$ and an early-downswing window $[t_0,t_1]$ before computing stretch: $$ \Delta\alpha_X=\max_{t\in[t_0,t_1]}\alpha_X(t)-\alpha_X(t_0). $$ Including $t_0$ makes this number nonnegative by definition; zero means no increase in that window. It does not prove that every golfer exhibits a positive stretch. A signed change between two specified events can instead be negative. Maximum pelvic and thoracic backswing positions need not occur simultaneously, and two zero angular velocities do not imply zero relative angle or zero elastic preload. ### What the Classic Study Measured Cheetham and colleagues compared 10 highly skilled and 9 less-skilled right-handed golfers using 5-irons and electromagnetic tracking. Their reported 30 Hz sampling and pelvis/T3 rotation measurements concern the transition region. They found a greater early-downswing X-factor increase in the higher-skill group, but no statistically significant between-group difference at the top of the backswing [@Cheetham2001]. This group comparison does not establish that increasing one player's separation will increase distance. The observed angle change also does not measure tissue strain energy, spindle response, or muscle-fiber shortening. For an assumed linear torsional spring, $U=\tfrac12k(\alpha_X-\alpha_0)^2$ is a model with a specified rest angle and stiffness. In a living trunk, distributed joints, muscle routing, tendon compliance, activation, and passive tissues must justify that approximation. An increased angle can alter storage and muscle operating conditions while also changing loads and the future club path. The measured association is a reason to investigate that interaction, not a warrant to maximize separation. ## Wrist Motion and Work at the Grip ### Lag Is Geometry, Not a Muscle State In a planar approximation, define an absolute arm angle $\theta_a$ and absolute club angle $\theta_c$ from the same reference; their signed relative angle is $\phi=\theta_c-\theta_a$. An included angle from directed forearm and shaft vectors is another convention and can differ by a supplement. State the directions and zero position. Three-dimensional wrist motion includes more than a single lag coordinate, and a flexible shaft adds deformation. A projected angle does not uniquely specify the clubface orientation. Neither hand speed nor clubhead speed determines muscle shortening speed. A muscle-tendon length $\ell(q)$ has rate $\dot\ell=(\partial\ell/\partial q)\dot q$; fiber speed additionally depends on tendon compliance and pennation. Classical force–velocity behavior concerns the contractile element under declared activation and length conditions. It cannot be applied by converting an absolute club speed directly into a universal wrist torque limit. Active lengthening and isometric force are also different operating conditions from concentric shortening. Sprigings and Neal's planar three-segment simulation found that suitably timed active wrist torque could increase clubhead speed by about 9 percent in its modeled conditions [@Sprigings2005]. That is a simulation result with particular torque generators and constraints, not an experimentally established instruction to add a late wrist action. It is nevertheless enough to reject the universal claim that wrist torque cannot contribute at high club speed. Conversely, a measured small net wrist moment does not prove that all wrist muscles are inactive. ### The Club Receives a Wrench, Not Only a Torque For a rigid club, with grip point $H$, external grip force $\mathbf F_H$, and grip couple $\boldsymbol\tau_H$, its pre-impact power balance is $$ \frac{d}{dt}(T_c+U_{g,c})= \mathbf F_H\cdot\mathbf v_H+\boldsymbol\tau_H\cdot\boldsymbol\omega_c+P_{\rm other}. $$ The term $P_{\rm other}$ accounts for effects such as aerodynamic loading. For a flexible club, include shaft strain energy and dissipation and use the grip section's angular velocity for its applied couple. The power entering the club is not the same as the combined power of an anatomical wrist actuator acting between two moving bodies. Even when $\boldsymbol\tau_H=0$, a translating grip can do work through $\mathbf F_H\cdot\mathbf v_H$. The two hands may also produce a resultant couple through their separated forces. A zero net couple at one chosen port does not remove force transmission or make the club an isolated body. Nesbit and Serrano's analysis explicitly modeled both force and torque work at the club interface [@Nesbit2005b]. The balance here follows rigid-body mechanics independently of any biological interpretation of their reconstructed loads. ## A Release Model That Retains the Moving Grip ### Derive the Unactuated Equation Use an inertial vertical plane with horizontal $x$ and upward $y$. The rigid club's center is $\mathbf r_G=\mathbf r_H+\ell\mathbf e$, where $\mathbf e=(\cos\theta,\sin\theta)$ points from grip to center and $\mathbf e_\perp=(-\sin\theta,\cos\theta)$. Let $m$ be mass, $I_G$ central inertia, and $I_H=I_G+m\ell^2$. Eliminating the grip force between translation and rotation gives $$ I_H\ddot\theta=\tau_H+m\ell(\mathbf g-\mathbf a_H)\cdot\mathbf e_\perp. $$ To see the origin of the forcing, substitute $\mathbf a_G=\mathbf a_H+\ell\ddot\theta\mathbf e_\perp-\ell\dot\theta^2\mathbf e$ into $m\mathbf a_G=\mathbf F_H+m\mathbf g$, then take moments about $G$. The radial $\dot\theta^2$ term has no moment about the grip; the acceleration of the grip has a tangential projection that does. Omitting $\mathbf a_H$ changes the problem after “release.” For an illustrative instantaneous state, set $m=0.2$ kg, $\ell=1$ m, $I_G=0.02$ kg m$^2$, $\theta=0$, $\dot\theta=2$ rad/s, $\mathbf v_H=(0,-3)$ m/s, $\mathbf a_H=(0,-20)$ m/s$^2$, and $\tau_H=0$. Then $I_H=0.22$ kg m$^2$ and $\ddot\theta\approx9.2636$ rad/s$^2$. The grip force is approximately $(-0.8,-0.18527)$ N and supplies $0.55582$ W. Directly differentiating $T_c+U_{g,c}$ gives the same rate. Angular acceleration is present with no applied grip couple because the grip trajectory is still imposed and its force still acts. Changing only grip acceleration to free fall, $\mathbf a_H=\mathbf g$, gives zero angular acceleration. A stationary grip instead gives $\ddot\theta\approx-8.9182$ rad/s$^2$ at the same horizontal orientation. These are three different boundary conditions, not three explanations of the same swing. Centrifugal or Coriolis terms in other coordinates reorganize the dynamics; they do not introduce an additional energy source. ### Holding, Releasing, and Feasibility Suppose an idealized holding condition imposes $\theta=\theta_a(t)+\phi_0$. The required grip couple is $$ \tau_{\rm hold}=I_H\ddot\theta_a-m\ell(\mathbf g-\mathbf a_H)\cdot\mathbf e_\perp. $$ A bound $|\tau_{\rm hold}|\le\tau_{\max}$ checks whether this imposed trajectory is feasible in the toy model. It does not assert that a human wrist locks perfectly until its torque limit is reached. A controller could change its command earlier; anatomical joints need not behave as binary latches. For a smooth release without an impulse, initialize the released motion with the angle and angular velocity immediately before release and continue the specified grip trajectory. If a constraint change produces an impulse, its momentum and energy effects require a separate event model. A claim that later release improves speed must specify which intervention changes, which upstream controls or trajectories remain fixed, the admissible torque/force and joint-range limits, and the endpoint. Compare states at a common impact surface with correct clubface orientation, not just at an arbitrary common clock time. If grip motion is prescribed, its force and work generally change between alternatives. Giving both alternatives the same torque history does not make their input work equal. There is no universal percentage benefit or optimal delay that follows from this pendulum equation alone. ## Ground Reaction: Momentum Channel and Contact Constraint ### Compute the Full Moment Vector Choose a right-handed frame: $x$ along the target line, $y$ horizontally across it, and $z$ upward. About origin $O$, a force applied at displacement $\mathbf r=(r_x,r_y,r_z)$ and a free moment $\mathbf m$ produce $$ \boldsymbol\tau_O=\mathbf r\times\mathbf F+\mathbf m =\begin{bmatrix}r_yF_z-r_zF_y\\r_zF_x-r_xF_z\\r_xF_y-r_yF_x\end{bmatrix}+\mathbf m. $$ Vertical force at a horizontal offset produces horizontal-axis moments, not a vertical-axis yaw moment. Horizontal force at a vertical offset contributes horizontal-axis moments too. Summing those components as if all were pelvic yaw torque is invalid. For a numerical example about a stated origin, let $\mathbf r=(0.4,0.2,-1)$ m and $F_z=1.5(80)(9.81)=1177.2$ N. Vertical force alone gives $(235.44,-470.88,0)$ Nm. Add $F_x=100$ N, $F_y=200$ N, and a free yaw moment $m_z=10$ Nm: the complete moment, in Nm, is $$ \boldsymbol\tau_O=(435.44,-570.88,70)^T. $$ If that net yaw moment stays constant for $0.05$ s, its yaw angular impulse is $3.5$ N m s. These are declared illustrative inputs; the calculation is not a prediction of hip angular speed. ### Whole-Body Angular Momentum Is Not Pelvic Rotation For the complete golfer-plus-club system about a fixed inertial origin, $$ \dot{\mathbf H}_O=\sum\boldsymbol\tau_{O,\rm external}. $$ Joint forces and couples are internal to that system. Ground wrenches, gravity, aerodynamic loads, and ball contact where present are external. About the total center of mass, uniform gravity has zero resultant moment, and the external-moment balance also holds for angular momentum defined relative to that center. Other moving reference points require the appropriate transport terms. The force example above gives a change in whole-system angular momentum only when all other relevant external moments are accounted for. One cannot divide that moment by an isolated pelvic inertia to obtain pelvic acceleration in a coupled swing. Nor does zero external yaw moment prevent the pelvis from rotating: internal actions can produce counter-rotation elsewhere while conserving total angular momentum. Ground contact permits exchange of momentum with the environment and changes feasible motion; it is not the sole possible cause of every observed segment rotation. ### Center of Pressure and Ground Work A force plate's resultant force, moment, and center of pressure are related but not interchangeable. The center of pressure describes a pressure-resultant location on the contact plane under its adopted convention. It does not by itself determine horizontal shear force or the residual free yaw moment. Near vanishing normal force, center-of-pressure estimates can become ill-conditioned. Combining two plates also requires a common origin and frame. At a perfectly stationary, non-slipping contact, ideal reaction forces do zero mechanical work because the material contact points have zero velocity. Yet they can provide a nonzero moment and angular impulse. Muscles supply mechanical work internally from chemical energy while contact constrains the motion. Real shoes, ground compliance, slipping, and foot deformation can add storage or work terms. A peak ground force before impact does not mean zero force at impact, and it does not determine when the center-of-mass speed peaks. Timing force, impulse, work, and segment velocity answers four different questions. ## Close the Energy Budget Before Naming Efficiency ### Choose the System and the Time Interval For a pre-impact mechanical model including the body, club, gravity, and chosen elastic elements, write $$ E=T_{\rm body}+T_{\rm club}+U_g+U_e,\qquad E(t_I)-E(t_0)=W_{\rm act}+W_{\rm ext}-D. $$ Here $W_{\rm act}$ is signed actuator work at the declared joint ports, $W_{\rm ext}$ is work from the other external interactions not already represented by potential energy, and $D\ge0$ is the modeled dissipation. Avoid counting an elastic element both in $U_e$ and again as an independent actuator energy input. An ideal frictionless model with ongoing positive actuator work does not conserve mechanical energy; it obeys this balance with increasing $E$. Backswing history enters through initial position, velocity, and stored energy. The top of a backswing is not generally a state of complete rest. Gravity needs a scale check: raising an 80 kg whole-body center of mass by $0.1$ m changes potential energy by $78.48$ J. Individual segment and club height changes must also be handled consistently. Small-looking displacement is not a sufficient argument to delete the term. Energy of a segment depends on both translation and rotation. A decrease in angular speed about one measured axis can coexist with increasing segment kinetic energy. Even a verified decrease in total segment energy does not identify its destination without boundary powers and storage. Negative net joint power also does not uniquely identify eccentric action in every spanning muscle. Tendon energy changes and simultaneous muscle actions can yield the same net value. ### Signed Work, Positive Work, and Metabolism For an actuator power $P_j$, distinguish signed work $W_j=\int P_jdt$ from positive work $W_j^+=\int\max(P_j,0)dt$ and absorbed-work magnitude $W_j^-=\int\max(-P_j,0)dt$, so $W_j=W_j^+-W_j^-$. Summing positive work over joints is not the same as taking the positive part after summing their powers. State which denominator a reported ratio uses. Nesbit and Serrano analyzed one selected swing from each of four amateur golfers with reconstructed body and club models. Their “swing efficiency” was club work divided by total body joint work; their table reports about 20.2–26.8 percent [@Nesbit2005b]. Those values are not a measured elite population range, a direct muscle efficiency, or a fraction of metabolic energy. A larger ratio also need not imply a faster club: numerator and denominator can change together. For a transparent illustrative budget, suppose initial recoverable mechanical energy is $100$ J, positive actuator work is $500$ J, absorbed actuator work is $80$ J, and other dissipation is $20$ J, with zero other external work. Final mechanical energy is $500$ J. It could comprise $250$ J in the club and $250$ J in body motion and retained potentials. Club energy divided by positive actuator work is $0.50$, while club energy divided by signed actuator work is about $0.595$. Neither ratio alone identifies where the rest became heat or establishes metabolic efficiency. Initial energy can even make a club-energy/work ratio exceed one without violating conservation. Collision adds another boundary and another interval. Pre-impact club kinetic energy is not all transferred to the ball; club motion, deformation, losses, and ball spin remain relevant. A numerical impact-transfer fraction must follow an explicit collision model or measurement. It cannot be appended as a universal percentage to close an incomplete downswing budget. ## How Drift and Control Connect to the Chain ### Declare What Zero Input Removes The full local state must at least include configuration and velocity, $x=(q,v)$. With a direct torque input and consistently eliminated regular constraints, it may admit $$ \dot x=f(x)+G(x)u,\qquad f(x)=\begin{bmatrix}v\\a_0(q,v)\end{bmatrix}. $$ The first component is configuration rate, not the complete dynamics. The drift is the zero-input field for the declared input port. Its contents depend on which gravity, elastic, dissipative, actuator, and contact effects the model retains. A change of coordinates transforms both the drift and the input fields; numerical component magnitudes need a declared scaling or metric. If input means neural excitation, zero input need not mean zero activation or zero muscle force. Activation and tendon states must then be included, and a simple affine torque model may no longer represent the same intervention. If input means a net joint torque, removing it also differs from making all muscles inactive. In a moving-grip club model, setting its grip couple to zero retains the imposed grip path and its force. These distinctions are essential to the zero-torque counterfactual framework in [the zero-torque counterfactual chapter](./ch06_zero_torque_counterfactual.qmd#sec-zero_torque_counterfactual). ### Attribution and Later Consequences At a fixed state in a genuinely affine model, $G(x)u$ is the instantaneous difference between actual and zero-input state rates. It does not follow that the difference between two finite trajectories equals its integral along the original trajectory. Once states diverge, gravity, geometry, constraint reactions, and future control effectiveness change too. A counterfactual that crosses a contact boundary requires an event-consistent model. The first-order endpoint sensitivity along a nominal smooth trajectory is $$ \delta x(t_I)=\Phi(t_I,t_0)\delta x(t_0) +\int_{t_0}^{t_I}\Phi(t_I,s)G(x(s))\delta u(s)\,ds, $$ where $\Phi$ is the state-transition matrix of the full linearized dynamics under the specified control policy. For a terminal output $y=h(x)$ at a fixed time, premultiply by $Dh$. If impact time is determined by a contact event, include its timing sensitivity and any impact map rather than silently treating $t_I$ as fixed. This gives the site's central argument a precise form: early action changes the state and geometry from which later action operates, and the existing state continues to evolve between inputs. The relevant question is how much admissible future action can change the required impact outputs, with what uncertainty and cost. A large instantaneous drift-control ratio does not measure neural latency, prove an optimal policy, or establish absence of useful distal control. It needs a declared model, coordinate metric, input port, and output horizon before it can inform that question. ## Test the Explanation ### Link Each Claim to an Observable To test sequencing, collect synchronized segment kinematics with declared coordinate frames and timing uncertainty. To test power flow, add external wrenches, body/club inertial estimates, and an inverse-dynamics model whose residuals and filtering sensitivity are reported. To infer individual muscle or tissue action, additional measurements and physiological assumptions are necessary. Electromyography measures an electrical signal related to activation; it does not directly measure muscle force, joint work, or tendon energy. Check integrated boundary powers against the change in mechanical energy, allowing for measurement uncertainty. Large unexplained residuals can arise from differentiation noise, mismatched time bases, inaccurate inertias, omitted foot or shaft deformation, or inconsistent moment references. Reproducing measured positions in a kinematically driven model is not an independent validation of all its internal forces. ### Test Interventions Against Competing Explanations Suppose a player gains club speed after changing transition timing. Candidate explanations include different active work, a changed hand path, different segment geometry, altered elastic storage, a different impact time, or changed clubface/contact conditions. Compare repeated trials and quantify the outputs that matter: speed, path, face orientation, strike location, ball result, and consistency. A faster isolated segment peak is not enough. Use a model intervention to separate hypotheses only after declaring the preserved quantities. Holding the grip path fixed isolates one question; holding actuator effort fixed isolates another. Sensitivity to initial state and model parameters determines whether the predicted advantage survives measurement uncertainty. A simulation that matches an observed sequence can demonstrate compatibility with a mechanism, but competing mechanisms may match the same sequence. Successful explanation requires discriminating observations or interventions. ## What the Reader Can Use in Practice The mechanics support an investigation organized around the desired club state and the player's repeatability. They do not justify a universal instruction to maximize pelvic speed, freeze the wrists, or seek the largest possible X-factor. A change should be evaluated by its measured effect on the complete shot, including dispersion and strike, rather than by conformity to an attractive sequence diagram. Slow practice can help a player explore positions and coordination, but it is not automatically dynamically similar to a full-speed swing. Under a time scaling, inertial terms scale differently from gravity, and activation, elastic, and contact effects need not scale together. The motion should be reassessed at the intended speed. Likewise, smoother commanded torque does not guarantee small club jerk near rapidly changing geometry or contact events. Force-plate feedback can inform contact forces, moments, impulse, and their timing when measured correctly. It does not directly calibrate a drift-control ratio. Feeling “effortless” does not prove that muscles stopped working, and muscle activation does not identify concentric action; active muscle can shorten, remain approximately isometric, or lengthen. These distinctions allow useful coaching descriptions to coexist with technically honest explanations. The connections are the important result. Ground contact shapes feasible momentum exchange; muscle action changes energy and state; geometry maps that state to club motion; elastic elements store and return work; and the remaining time constrains what later control can accomplish. An account that quantifies those connections is more informative than either a list of peak timings or a claim that one part of the chain supplies all the speed. ## Reproduce the Mechanical Checks The following Python functions use NumPy and SciPy to evaluate the stated teaching cases. Times are in seconds and other inputs use SI units. The rotor solution can be checked against independent numerical integration; the modal vectors are mass-normalized. The moving-grip calculation checks instantaneous force, moment, and total mechanical-energy rates, not a complete golf trajectory. The same program is printed in both editions. ```python # Kinetic-chain mechanical checks import numpy as np from scipy.linalg import eigh def rotor_case(time: np.ndarray) -> dict[str, np.ndarray]: """Two fixed-axis rotors, zero initial state, constant 50 Nm input.""" frequency = np.sqrt(250.0) phase = frequency * time separation = 0.1 * (1 - np.cos(phase)) relative_speed = 0.1 * frequency * np.sin(phase) mean_angle = 10 * time**2 mean_speed = 20 * time angle_one = mean_angle + 0.2 * separation angle_two = mean_angle - 0.8 * separation speed_one = mean_speed + 0.2 * relative_speed speed_two = mean_speed - 0.8 * relative_speed energy = speed_one**2 + 0.25 * speed_two**2 energy += 50 * separation**2 return { "state": np.array([angle_one, angle_two, speed_one, speed_two]), "acceleration": np.array([20 + 5 * np.cos(phase), 20 * (1 - np.cos(phase))]), "energy": energy, } def mode_case() -> dict[str, np.ndarray]: """Mass-normalized modes of the stated three-rotor model.""" mass = np.diag([2.0, 1.2, 0.4]) stiffness = np.array([[120, -120, 0], [-120, 220, -100], [0, -100, 100]]) squared, modes = eigh(stiffness, mass) return {"squared_frequencies": squared, "modes": modes} def grip_case(acceleration: tuple[float, float]) -> dict: """Horizontal rod, zero grip couple, translating/accelerating grip.""" mass, length, central_inertia = 0.2, 1.0, 0.02 gravity = np.array([0.0, -9.81]) grip_velocity = np.array([0.0, -3.0]) angular_speed = 2.0 polar_inertia = central_inertia + mass * length**2 alpha = mass * length * (gravity[1] - acceleration[1]) / polar_inertia center_acceleration = np.array(acceleration) + length * np.array([-4.0, alpha]) force = mass * (center_acceleration - gravity) center_velocity = grip_velocity + np.array([0.0, length * angular_speed]) kinetic_rate = mass * center_velocity @ center_acceleration kinetic_rate += central_inertia * angular_speed * alpha return { "alpha": alpha, "force": force, "grip_power": force @ grip_velocity, "energy_rate": kinetic_rate - mass * gravity @ center_velocity, } def ground_case() -> dict: """Moments about the declared origin; all inputs are illustrative SI values.""" arm = np.array([0.4, 0.2, -1.0]) vertical_force = np.array([0.0, 0.0, 1.5 * 80 * 9.81]) force = vertical_force + np.array([100.0, 200.0, 0.0]) moment = np.cross(arm, force) + np.array([0.0, 0.0, 10.0]) return { "vertical_moment": np.cross(arm, vertical_force), "full_moment": moment, "yaw_impulse": moment[2] * 0.05, } def joint_case() -> dict[str, float]: """A spring/damper stores and dissipates work while transferring power.""" separation, relative_speed = 0.1, 2.0 stiffness, damping = 100.0, 2.0 torque = stiffness * separation + damping * relative_speed return { "proximal_power": -torque * 5.0, "distal_power": torque * 3.0, "storage_rate": stiffness * separation * relative_speed, "dissipation": damping * relative_speed**2, } ``` For example, evaluate the rotor function at a time grid from zero to one second and compare its energy with $50\theta_1$. For the modal case, verify $K\Psi=M\Psi\operatorname{diag}(\omega^2)$ and $\Psi^TM\Psi=I$; the smallest squared frequency is zero up to floating-point roundoff. Evaluate the grip case with accelerations $(0,-20)$, $(0,-9.81)$, and $(0,0)$ to reproduce the three boundary conditions. Ground and joint functions reproduce the moment and power examples directly. Small rounding differences should not be interpreted as physical residuals. ## Exercises 1. **Sequence Counterexample.** Starting from the two-rotor equations, derive the mean and relative solutions and reconstruct both angular velocities. On $[0,1]$ s, locate rotor 1's maximum velocity and rotor 2's first acceleration maximum. Explain why their order refutes a universal proximal-speed/distal-acceleration rule without refuting every possible sequencing mechanism. Verify the input-work identity independently. 1. **X-Factor and Storage.** In the declared backswing-positive convention, let thorax and pelvis angles at the reference event be $90$ and $45$ degrees. At a later event let them be $88$ and $38$ degrees. Compute separation at each event and its change. For an assumed torsional spring with rest separation $30$ degrees and stiffness $100$ Nm/rad, compute the storage change in joules. State why this result is not a measurement of human trunk energy and why both turn rates becoming zero would not remove preload. 1. **A Moving-Grip Release.** Derive the club equation by eliminating grip force from Newton–Euler balances. Reproduce the three acceleration cases and check energy power. Then prescribe a smooth grip trajectory and an arm angle history of your choosing, state them explicitly, and compute the holding couple. Explain the additional endpoint, feasibility, and input-work checks needed before comparing two release times. Zero grip couple alone is not a complete boundary condition. 1. **Ground Wrench and Reference Point.** Reproduce all three moment components in the ground example. Change the origin by a known displacement and transform the moment consistently. Integrate the original yaw moment over the stated interval. Identify the additional information needed to infer pelvic acceleration from whole-system angular impulse, and explain why center of pressure alone cannot supply it. 1. **Energy and Identifiability.** Close the illustrative $100$ J initial-energy budget using the stated positive work, absorbed work, and dissipation. Compute both club-energy ratios. Construct a second consistent budget with the same final club energy but different body energy and actuator work. Explain why club speed alone cannot distinguish the budgets or determine individual muscle action. Add a separately declared collision model only if estimating ball energy. 1. **Modes and Experimental Discrimination.** Solve the three-rotor generalized eigenproblem, identify its rigid mode, and verify mass normalization. Apply a finite torque pulse to one rotor and calculate the response from rest. Repeat with relative-joint damping and changed initial preload. Compare the resulting peak definitions explicitly. Propose a synchronized measurement that could distinguish a change in active work from a change in energy redistribution in a golfer; report which required quantities remain model-dependent. ::: {.callout-note} ## On This Site - **[Proximal-to-Distal Energy Transfer in the Golf Swing](../../proximal-distal-energy-transfer.html)** — Sequential timing, power ports, and chain coordination - **[A Journey Through the Swing](../../proximal-distal-a-journey-through-the-swing.html)** — Narrative companion to kinetic-chain mechanics :::