Every club player who wants a bigger serve eventually runs the same experiment: swing harder. It buys a fortnight of louder contact, then the double faults arrive, then the shoulder starts talking on Sunday evenings. The plateau returns, and the player quietly files it under talent.
Three explanations compete for that plateau, and most instruction on serve mechanics commits to one while ignoring the other two. The first says the arm isn't fast enough. The second says the racquet is wrong — too light, too heavy, strung badly. The third says the sequence is broken: the legs, hips and trunk never load or release in order, so the arm is left doing a job that was never assigned to it.
We read what the published evidence says about each.
How we evaluated
We did not hit any serves for this piece and we did not measure anything. What follows is a synthesis of three things: peer-reviewed research on serve mechanics, manufacturer specifications and the claims made alongside them, and the consensus visible across independent tester and owner reviews.
We weighted those sources unevenly, on purpose. Direct kinematic measurement of real serves outranks modelling estimates. Modelling estimates outrank coaching consensus. Coaching consensus outranks anything a brand says about its own frame. We then judged the three explanations against four named criteria:
- Contribution — how much of racquet-head speed at impact the mechanism actually accounts for
- Causality — whether it sits upstream or downstream of the others
- Injury cost — what the shoulder and elbow pay for it
- Cost to change — money, time, or both
| Explanation | Contribution to racquet speed | Injury cost | Cost to change |
|---|---|---|---|
| Arm and wrist | Largest single direct contributor | Highest — the load lands on shoulder and elbow | Free, but the ceiling arrives fast |
| Racquet and string | Real but modest, and self-cancelling | Can lower arm load; rarely raises speed | The price of a frame, per attempt |
| Leg drive and trunk sequencing | Indirect — sets the speed the arm can reach | Lowest; associated with reduced upper-limb load | Free, slow, unglamorous |
The case for the arm
The strongest single piece of evidence in this argument belongs to the explanation we half expected to dismiss. Elliott, Marshall and Noffal (1995), analysing high-performance first serves, attributed the majority of racquet-head speed at impact to internal rotation of the upper arm, with wrist flexion the next largest contributor. Later work has revised the exact proportions, but the direction has held: at the moment of contact, the arm is where the speed is.
That finding gets misread as licence to muscle the ball, and it is worth being precise about why the reading fails. Internal rotation at those velocities is not a voluntary muscular effort — it happens too fast to be driven that way. It is a release, not a push. The study answers where the speed appears, not what produced it. Sample sizes in this literature are small and the players are elite, which is a genuine limit on how far a 3.5 should generalise from it.
The case for the racquet
Here the evidence is thinnest and the incentives are worst. The physics isn't in dispute: Howard Brody's work on racquet mechanics laid out the trade-off plainly — a higher-swingweight frame returns more ball speed for a given swing speed, while simultaneously reducing the swing speed you can generate. The two effects substantially cancel, which is why most players who switch frames report a different feel rather than a different serve.
Where equipment does earn its keep is comfort. The consensus among independent testers and long-term owners is consistent: a softer string, or the same string at lower tension, reduces impact harshness. That is an arm-health argument, and a good one. Reviewers who claim a string adds pace are usually reporting a change in launch angle or dwell time that they experienced as speed. Unattributed mph claims from anyone selling the product are marketing, not measurement.
The case for the chain
The lower-body evidence is less dramatic and more consistent. Girard, Micallef and Millet (2005) reported that higher-level players generate greater vertical ground reaction force during the serve than lower-level players — leg drive is not decoration, it scales with standard. Reid, Elliott and Alderson (2008) found that restricting leg drive altered shoulder and elbow loading, and Martin and colleagues (2014) reported that hip and knee flexion reduce upper-limb joint loads across the service motion.
One caveat, because it is repeated carelessly everywhere. The much-quoted claim that a 20% loss of kinetic energy from the hip and trunk demands a 34% increase in shoulder rotational velocity to deliver the same energy to the hand comes from Kibler (1995). It is a modelling estimate about energy transfer — not a measured mph penalty, and not something anyone clocked on a court. It illustrates a principle well. It is not a number you can hold a serve to.
The verdict
The arm and the chain were never competing answers. The arm is where serve speed appears; the legs and trunk decide how much speed the arm is permitted to produce, and who pays for it. That is why the sequencing explanation wins the comparison: it is the only one of the three sitting upstream of the other two. Fix the sequence and the arm gets faster as a consequence. Drive the arm directly and you have bought speed on credit, with the shoulder posted as collateral.
Equipment finishes a distant third as a speed intervention and a respectable second as a comfort one — which is a real result, just not the one it is sold as.
Evidence grade: Moderate. The lower-body findings run in the same direction across independent research groups, but the samples are small, largely elite or well-trained, and they measure joint loads and forces rather than mph gained by a recreational player over a season. Nobody has run the trial that 3.5s actually want.
Who this is for, and who it isn't
This is for the 3.0–5.0 who serves slower than they'd like and whose shoulder aches by the third set. It is not for the player whose toss lands somewhere new every time — no sequencing survives an unstable ball, and that is the prior fix. It is not for anyone whose real problem is a second serve landing short, where speed is the wrong target entirely. And if you have current shoulder or elbow pain, the load-reduction findings are a reason to see a physio, not a reason to add leg drive to an already-injured chain.
Try this week
Take one basket of balls. Serve ten with your feet flat and your knees deliberately locked — no bend, no push, arm only. Then serve ten normally. You are not measuring anything, and you shouldn't pretend you are. You are finding out whether the two sets feel meaningfully different. If they don't, your chain currently begins at your shoulder, and you have just located the thing worth practising.