Launch Monitor Vendor Reference

technology
launch-monitors
reference
golf
A catalogue of publicly available vendor documentation for golf launch monitors and club-tracking systems, with what each vendor actually measures, how each defines its parameters, and where the definitions are mutually incompatible.
Author

Dieter Olson

Published

August 3, 2026

What This Page Is

Launch monitor data is the primary evidence base for almost everything written about the modern golf swing, and the documentation behind it is scattered across vendor blogs, help-centre articles, sticker-guide PDFs, user manuals and patent filings — much of it undated, some of it contradictory, and a fair amount of it marketing wearing a technical vocabulary.

This page is an attempt at a single catalogue: what each vendor publishes, where it lives, what it actually says, and — most importantly — where two vendors use the same word for different quantities.

ImportantThe Headline Finding: Club Numbers Are Not Portable Between Systems

There is no industry-standard reference point for club measurements, and no standard definition of closure rate. Two systems can report different numbers for the same swing and both be correct under their own conventions.

TrackMan says so itself, in its own documentation: “it is very important to be aware of and understand the differences in measurement methodologies applied (Center of Gravity vs Center of club face, for example) when tracking the collision interval between the club and the ball. Often, the numbers measured (or calculated) aren’t comparable.”

That sentence should be printed on the wall of every fitting bay.

For the physics behind these measurements, see Impact Mechanics and Ball Flight; for the measurement geometry, Launch Monitor Technology.


Part I: The Two Questions That Decide Everything

Before the catalogue, two axes on which vendors genuinely differ.

I.1 Where on the Clubhead Is “the Club”?

A clubhead is a rigid body of finite size rotating as it moves. Different points on it travel in different directions at the same instant. So “club path” is meaningless until you say which point.

Vendor Club path referenced at Source
TrackMan Geometric centre of the clubhead (within ~6 mm of CG) Club data definitions
FlightScope “The geometric center of the club” Data parameters
Uneekor “The club head’s geometric center at the time of impact” in older sticker guides; current Club AI material does not restate the reference point EYE XO / QED sticker guides; Club AI announcement
Full Swing “The club head’s center prior to impact” Data-points blog
Foresight “The club head face center” — stated only in a putting glossary PDF, not on any product page Putting guide & glossary
Garmin R50 A sticker on the club face (R10 measures at the head) R50 vs R10
GEARS Not stated for path; does report speed at both face centre and impact point Club/ball metrics

Two things in that table deserve comment.

TrackMan quantifies the discrepancy, and it is large. From its own page:

For a driver, the difference between the path of the club head’s center of gravity (CoG) and the path of the center of the club face is approximately 3 degrees (center of club face path being more outside-in). This is because the CoG is located approximately 25-50 mm behind the club face. The same goes for Attack Angle of CoG and the center face, where the center of the face moves more upwards compared to CoG for a driver (typically 1 degree).

And it explains why optical systems land elsewhere: “Optical-based launch monitors almost always measure Club Speed, Attack Angle and Club Path relative to the center of the club face, since this is where the reflective markers are positioned.”

For a driver, two systems using those different reference points can therefore disagree by roughly the 3° example TrackMan gives, with the face-centre path reading more out-to-in. That component of disagreement is definitional rather than an accuracy error, although the realised difference depends on club geometry and motion.

The published Uneekor and Full Swing definitions differ on the instant. Older Uneekor guides say “at the time of impact”; Full Swing says “prior to impact.” Those phrases are not metrologically interchangeable, but the documents do not define the sampling window closely enough to predict a fixed inter-device offset.

WarningTrackMan’s Face-to-Path Is a Hybrid Quantity

TrackMan references path and attack angle at the geometric centre, but face angle and dynamic loft at the contact point:

For Face Angle and Dynamic Loft, Trackman uses the orientation of the club face at the center point of where the ball is in contact with the club face, meaning “at the impact location”. […] Trackman has chosen to use the Face Angle and Dynamic Loft at the impact location, since this is the only orientation of club face that influences the ball.

Each choice is individually well-motivated. But their difference, face-to-path, is therefore an orientation at one point minus a velocity direction at a different point on the same body. It is the golf industry’s central diagnostic number and it is not a physically clean angle under either convention.

TrackMan also quantifies the face-side sensitivity: “For standard drivers, a 10 mm impact towards the toe will cause the Face Angle to be 2 degrees more open at the impact location when compared to the center of the club face.”

Does this build a slice into the numbers? The mechanism is real: if a player is coached to a zero reported path on a geometric-centre system, the contact point — which is what actually strikes the ball — is travelling roughly 3° more out-to-in than the number they are chasing. But no vendor makes that claim, and it should not be attributed to one. Treat it as an inference from TrackMan’s own quantities, with the caveat that players calibrate to ball flight rather than to numbers, and that a competing draw-side mechanism (the face continuing to close through contact) partly offsets it. The full treatment is in Impact Mechanics, Part VI.

I.2 Is Club Data Measured, or Inferred?

This is the distinction most obscured by marketing, and it separates the market more sharply than price does.

System Club data source Requires markers?
Uneekor EYE series Optical club imaging; marker requirement depends on model and software Older EYE/QED guides require stickers; 2026 Club AI is advertised as sticker-free
Foresight GCQuad Imaged from clubface markers Yes for club data (1 or 4 markers)
GEARS Full marker-based optical motion capture Yes
TrackMan Radar-tracked 3D silhouette of the head No
Full Swing KIT Radar + deep-learning trajectory estimation No
Rapsodo Stereo cameras + radar, reconstructed from a scanned 3D club model No
Sportsbox AI Monocular video; shaft line only — no face orientation at all No

Uneekor’s older sticker guide is blunt about its dependency: “CLUB DATA CANNOT BE READ IF THE BALL IS COVERING ONE OF THE TWO STICKERS. CLUB DATA WILL BE ONLY SHOWN IF BOTH STICKERS ARE VISIBLE.” That statement should not be generalised to the current range. Uneekor’s January 2026 Club AI announcement advertises sticker-free club data, while the current EYE XO2 product page still says reflective club stickers are required. The public documentation therefore conflicts as of 2026-08-03; marker requirements must be checked for the exact hardware, software package and data set. Marked-ball requirements for spin are a separate issue.

At the other end, Full Swing’s patent describes a radar-only device whose processor estimates trajectory “using deep-learning algorithms rather than predefined mathematical relationships.” Doppler radar returns range and radial velocity; it does not resolve the angular attitude of a clubface. A learned mapping from radar returns to a face angle may be empirically excellent, but it is inference. Notably, Full Swing’s own marketing claims measurement for the ball and makes no equivalent claim for the club.

A useful tell: on FlightScope, club path is unlocked by model and software tier rather than being a property of the sensor — which is itself evidence of derivation.


Part II: Vendor Catalogue

TrackMan

The most technically forthcoming vendor by a wide margin, though its material is scattered and several official URLs have rotted.

Document Link Notes
Tech insights: club data definitions trackman.com The single most important vendor document in golf. Reference points, the 3° figure, the 25–50 mm CG offset, bulge/roll sensitivity, in-contact face rotation
40+ TrackMan parameters explained trackman.com The full parameter dictionary
Spin axis trackman.com Definition, and yardage-per-degree tables
Spin loft trackman.com Defines spin loft as a 3D vector angle, calling the loft-minus-attack-angle subtraction “a close approximation”
Face to path trackman.com Curvature-per-degree figures by club and player class
Help Centre: Club Path support.trackmangolf.com “…geometric center at the time of maximum compression”
Help Centre: Face Angle support.trackmangolf.com “…at the center-point of contact… at the time of maximum compression”
Help Centre: D Plane Tilt support.trackmangolf.com The geometry-only spin axis, explicitly excluding gear effect

The TrackMan Newsletter Archive

Every official TrackMan newsletter URL now 404s, and the Wayback captures are redirect stubs. But a complete live third-party mirror of issues #1–#10 exists, and it is the single richest source of TrackMan-published numbers anywhere:

Issue Date Link
#1 Nov 2007 newsletter1.pdf
#2 Jan 2008 newsletter2.pdf
#3 May 2008 newsletter3.pdf
#4 Jan 2009 newsletter4.pdf
#5 Jul 2009 newsletter5.pdf
#6 Jan 2010 newsletter6.pdf
#7 Oct 2010 newsletter7.pdf
#8 Jun 2011 newsletter8.pdf
#9 Jan 2013 newsletter9.pdf
#10 Jan 2014 TrackMan_Newsletter_2014.pdf

The series ends at #10; the filename convention changed for the last issue, which is why sequential guesses fail. Highlights:

#7 (Oct 2010), “Ten Fundamentals” — the origin of most TrackMan rules of thumb. Contains the 85/15 claim in both planes (see Impact Mechanics, §IV.3 for why only the vertical one survives measurement); “SPIN LOFT = DYNAMIC LOFT − ANGLE OF ATTACK” as a definition rather than the approximation TrackMan’s current blog calls it; spin decay “typically 4% for each second”; and the attack-angle distance study — at 90 mph club speed, −5° attack angle optimises to 191 yards carry versus 214 yards at +5°.

#9 (Jan 2013) — the canonical club-delivery definitions, all specifying maximum compression, and the only published TrackMan accuracy specification we located:

Parameter Absolute Repeatability
Club speed ±1.5 mph ±0.4 mph
Attack angle ±1.0° ±0.6°
Club path ±1.0° ±0.6°
Dynamic loft ±0.8° ±0.6°
Face angle ±0.6° ±0.5°

(For TrackMan III/IIIe at 95% confidence. These are the figures Leach et al. tested against — and club path met ±1° on only 45% of shots.)

#8 (Jun 2011) — bulge quantified: “if the ball is impacted ½ inch (12.7 mm) towards the heel, the face angle will at this point on the club face be 2 deg. closed relative to the center of the club face… This rule-of-thumb actually works for all drivers on the market.”

Two details from the club-data page worth extracting, because they appear nowhere else in vendor documentation:

  • Face rotation during the collision. “For a toe impact with an iron, for instance, the club face can open 1-2 degrees during the collision interval (from first touch to separation of the ball from the club face).”
  • What the reported numbers exclude. “Trackman’s Club Speed, Attack Angle and Club Path data solely reflects player-controlled movement — meaning it reports pre-impact data only.”

The second is important for anyone comparing systems: TrackMan deliberately excludes collision-induced motion from its club numbers. A system reporting through the collision interval is measuring a different thing.

NoteDead and Login-Walled TrackMan Material

trackman.com/blog/golf/ball-flight-laws 404s. The 2009 newsletters survive only as third-party-hosted PDFs. TrackMan University is login-walled and cannot be cited.

Consequently, several TrackMan claims that circulate widely — including tour-versus-amateur club path and face angle averages — could not be verified against a primary TrackMan source. “Amateurs average an out-to-in path” is asserted constantly by coaching sites; we have not found TrackMan publishing it. Treat it as unverified.

FlightScope

Publishes clear parameter definitions and is explicit about its reference point.

Document Link
Data parameters (Mevo / Mevo+ / X3) flightscope.com
Club path explained flightscope.com
Face to path explained flightscope.com

Club Path: The direction of the geometric center of the club relative to the radar’s target line, measured at impact.

Angle of Attack: The angle at which the geometric center of the club approaches the ball, measured just before impact.

Technology is marketed as “Fusion Tracking” — radar combined with camera. Club path availability is gated by model and package tier, which is itself evidence of derivation rather than sensing.

Three documented inconsistencies are worth recording, because FlightScope’s documentation is otherwise among the better ones:

  • It contradicts itself on the reference point. Horizontal swing plane uses “the club’s geometric center”; club speed uses “the club’s centre of mass.” Those are different points on a clubhead, used interchangeably without acknowledgement. Face angle is given no clubhead reference point at all — only the external target line.
  • “Measured, not calculated” is contradicted by its own parameter pages. Marketing says “actual club and ball measurements - no calculations.” The club path page says “it calculates the club path”; angle of attack, dynamic loft, face-to-path, swing planes and spin axis all say “calculates.”
  • “Lighting-agnostic” in the blog, against a FAQ requiring “a minimum of 300 lux to capture accurate Face Impact Location data.”

Its cited patent, US10338209B2, discloses a continuous-wave signal around 10.5 GHz with multi-antenna phase interferometry — and is a ball-tracking patent. It discloses nothing about clubhead measurement, so there is no public disclosure of how FlightScope determines face angle.

Foresight Sports (And Bushnell Launch Pro)

Photometric, marker-based. Notable for taking the opposite side of the reference-point argument from FlightScope, in public:

Unlike radar-based technologies that analyze ball and club head performance results from the worst possible position (behind the club head) and then largely calculate performance based on the geometric center of the club head, Foresight Sports launch monitors use exact photometric measurements taken with a clear, direct view of the impact window. […] Our launch monitors measure – not calculate – club head data based on the club’s face plane.

That is a marketing claim, but it is a technical marketing claim and it is squarely at odds with FlightScope’s published definition.

Foresight does name its reference point, but in an unlikely place — a putting instructional glossary rather than any product page or the GCQuad manual:

Club path is the measured angle (in degrees) to the left or right of the target line that the club head face center is traveling at the moment of impact.

The same document is unusually careful elsewhere, and worth reading for its caveats. It notes that impact loft is “the inclination of this flat plane, not the actual loft at the impact point for curved face clubs such as a driver” — i.e. it does not correct for bulge and roll — and that the face plane itself “is a 2D plane described by the placement of the fiducials on the club face.” The coordinate frame is defined by where the operator sticks the markers, and no positional tolerance is published for that placement.

Document Link Notes
What we measure foresightsports.com Parameter definitions incl. closure rate
GCQuad user manual PDF Marker placement geometry; no glossary, no path definition
Ball & club data (EU) foresightsports.eu Slightly different wording of the same definitions

Marker dependency: four markers yield clubhead speed, smash factor, attack angle, swing path, face angle, loft, lie, closure rate and impact location; a single marker drops the output to speed, smash, attack angle and path. The expanded four-marker list is documented in Foresight’s club-marker guide.

GEARS Golf

Full marker-based optical motion capture — the highest-fidelity club tracking in common use, and the only system that ships the reference-point distinction as a user-visible feature:

Speed at Impact Point is the speed of the club head measured at the location that the ball made impact with the club at impact time. […] Speed at Face Center is the speed of the club head measured at the center of the club face. […] At the moment of impact, the speed at different locations of the club face varies.

GEARS therefore demonstrably understands that reference point matters — but ships both variants only for speed, not for path, and publishes no reference point for club path at all.

Uneekor

Uneekor’s overhead systems use high-speed infrared cameras. Older parameter definitions are buried in sticker-guide PDFs, while current product and software pages describe newer markerless capabilities without publishing a replacement metrological glossary.

Document Link
Support resources index uneekor.com
EYE XO2 manual PDF
EYE XO club sticker guide PDF
QED club sticker guide PDF
Club AI announcement (January 2026) uneekor.com
Current EYE XO2 product page uneekor.com
Patent marking page uneekor.com

The two current Uneekor pages are inconsistent: Club AI says its camera-based club data is sticker-free, while the EYE XO2 page says reflective club stickers are required for club data. This reference preserves both claims rather than selecting one without model-and-software-specific confirmation.

Worth knowing: Uneekor does not own its core measurement patents. Every patent on its own virtual-marking page resolves to Creatz Inc (Korea) — including US12008770 (the markerless “Dimple Optix” spin method, 2021 priority) and the earlier marked-ball patents US10776929 and US10587797 (2017). The progression from marked-ball to markerless dimple tracking is exactly the QED-to-EYE transition.

Full Swing

Marketing-grade documentation; substantive patents. The 37-page KIT user guide contains no specifications, no parameter definitions and no accuracy figures.

The patent, US11844990B2 (Full Swing Golf Inc), is the real technical document: four or more receive antennas in a deliberately non-uniform array, with one antenna placed at least 1.5× farther out to extend the angular-resolution baseline; sequential CW radar for Doppler velocity and FMCW for range; trajectory estimated by deep learning. Radar only — no cameras in the claims.

WarningA Commonly Miscited Patent

US7959517B2, “Infrared sensing launch monitor,” is assigned to Acushnet, not Full Swing, despite ranking highly in searches for Full Swing IP. Its method — deliberately heating or cooling golf equipment to create a thermal pattern, then imaging that pattern as a tracking marker — is unrelated to anything Full Swing sells.

Rapsodo

Holds a genuine technical patent, US11583746B2, “Measurement and reconstruction of the golf launching scene in 3D.” Stereo high-speed cameras plus radar; club parameters are reconstructed against a scanned 3D club model rather than measured directly. The word “reconstruction” in the title is the honest summary.

Sportsbox AI

Included because it is frequently discussed alongside launch monitors and should not be. It is a body-kinematics product from monocular video. Its published metric glossary contains no club path, face angle, closure rate, dynamic loft or attack angle. Its own help centre states plainly that shaft “droop, torque (twist) and lag of the club head will affect the relationship. We do not measure those characteristics.”

The geometric reason is decisive: Sportsbox produces a shaft line from two points (hands and clubhead). A two-point line cannot yield face angle or closure rate — orientation about the shaft axis requires a third, off-axis point. That is consistent with what it publishes.

Its accuracy page reports mean absolute differences of ~2° against an electromagnetic reference system, from 30 golfers at one swing each, with no dispersion statistic, no ICC, no limits of agreement, and zero club parameters validated. Searches of PubMed and Europe PMC for “Sportsbox” return zero results. For context, the published markerless literature reports transverse-plane RMSE up to ~57° against marker-based systems — and transverse rotation is precisely what Sportsbox’s headline metrics measure.

Garmin (Approach R10, R50)

Garmin provides one of the clearest public disclosure pages in this market: an explicit measured-versus-calculated split with manufacturer-stated tolerances (accuracy page):

Measured by radar Calculated by algorithm
Club head speed ± 3 mph Club face angle ± 2°
Ball speed ± 1 mph Apex height ± 5 ft
Launch angle ± 1° Carry distance ± 5 yd
Launch direction ± 1°

Two things follow. Club face angle is explicitly algorithmic. And club path and attack angle appear in neither column, though Garmin reports both: the cited page gives them no tolerance and does not classify them as measured or calculated. The omission is visible because Garmin publishes the surrounding detail.

Garmin also ships the only in-product measured/estimated indicator we found: “Spin rate is measured directly when possible. If it is displayed in italics, that means it was estimated based on other available data points.” Spin is not measured below 90 mph ball speed or under 20 m of observed flight.

The R50 is camera-only (2.4 GHz for connectivity, no radar band in its spec table) and requires club-face stickers for club data. Garmin states the reference-point consequence plainly: “Approach R50 requires a sticker and measures the speed and path of that point on the club face, while Approach R10 measures these at the club’s head.” Two devices from the same vendor, two different reference points, disclosed.

Rapsodo (MLM2PRO)

Tags eight of fifteen parameters (measured) on its FAQ — club path, attack angle, ball speed, club speed, launch angle, launch direction, spin rate and spin axis — leaving the rest calculated by omission. Club speed comes from radar, club path and attack angle from a 240 fps camera. It reports no face angle and no spin loft at all.

Its accuracy blog is the most creditable vendor-published validation in the consumer tier: n = 1,021 shots against a GCQuad, giving attack angle MAE 1.05° (r = 0.92) and club path MAE 1.19° (r = 0.86). Vendor-run and vendor-published, but with a real sample size and real error statistics — which is more than the premium tier offers.

SkyTrak / SkyTrak+

Photometric and ball-only on the original unit. The admission is arithmetic rather than stated: the FAQ lists roughly fifteen parameters it “provides,” then refers to testing of “all 5 measured data parameters” — and the only tolerance table SkyTrak has ever published covers exactly five, all ball metrics (ball speed ±1 mph, launch angle ±1°, back spin ±250 rpm, side spin ±250 rpm, side angle ±2°). No club metric appears in it. Club head speed and smash factor are derived, with no disclaimer anywhere.

SkyTrak+ adds radar and claims club path and face angle, but never drops the hedge: “proprietary machine-learning algorithms.” That is a model fitted to radar returns, not direct geometric measurement.

Worth flagging as a transparency regression: the 2014 product shipped real ± tolerances; the current flagship publishes no hardware specifications at all — no camera count, no frame rate, no radar band, no tolerances. The widely repeated “97–99% accurate” figure appears on no SkyTrak page.

Ernest Sports

Documentation has been deleted. The site was rebuilt and the entire manuals library, FAQ and Tour Plus manual now 404; the manual PDF was never captured by the Wayback Machine and is permanently lost. No measured-versus-calculated table has ever existed, and no Ernest Sports document anywhere defines club path or face angle, though the products advertise both.

Marketing asserts “zero estimation,” but two archived statements undercut it — an explanation that outdoor failures lose “the rotation of the ball and club data” because “the background is the same color as the golf ball” (tying spin and club data to the cameras, not the radar), and a 2014 FAQ conceding that a radar-only model “does not measure launch and spin however… reacts to various launch and spin conditions,” which is a calibration table, not a measurement.

NoteSimilar Portable Radars, Opposite Advice on Marked Balls

FlightScope requires aluminium stickers on the ball for indoor spin: “Aluminum stickers are needed in an indoor environment in order for the X3C, Mevo+, Mevo Gen2, and Mevo to accurately measure ball spin.”

Garmin warns against them: “Intentionally marked balls (e.g. metallic stickers) may degrade system performance.”

Garmin R10 and FlightScope Mevo/Mevo+ use 24 GHz radar, but that does not make their signal chains interchangeable; the FlightScope statement also covers X3C, whose architecture should not be inferred from the portable models. The opposite advice is evidence of model-specific antenna, signal-processing and estimation choices, not contradictory physics.

Garmin discloses a related trade-off for radar-reflective balls that nobody else does: with a Titleist RCT ball, “it is possible that the signal from the RCT ball is interfering with the signal from the club. You may need to use a standard golf ball if club metrics are what you are most interested in.” Good spin or good club angles, not both.

Swing Catalyst

Included for the same reason as Sportsbox: it appears in club-delivery discussions and measures no club data at all. Its own hardware is ground-only — pressure matrices and force plates — and its vision system is 2D markerless body pose built on off-the-shelf open-source models (RTMDet + RTMPose), detecting 26 body keypoints and no club keypoints.

Every club number displayed in Swing Catalyst is imported over a vendor SDK from a third-party launch monitor. Its own support documentation makes this explicit: “If you have a HMT unit, Swing Catalyst will pick up club data as well” — and when club data goes missing, “it may be the club is missing markers or the markers are no longer working properly.” The markers are Foresight’s. Notably, its supported-launch-monitor list records TrackMan integration as “limited to existing customers only.”

Two things are worth taking from Swing Catalyst regardless. First, one of the clearest vendor statements anywhere about the limits of pressure data:

Since these pressure measuring technologies can only measure forces perpendicular to the surface, they are only capable of measuring force in one dimension (1D). […] Making claims about the linear (towards/away from the target) or rotational components of the swing based on pressure measurements are at best an educated guess and must be interpreted with GREAT CAUTION.

That is a vendor arguing against the over-reading of its own cheaper product line, and it deserves to be quoted more often. It is also exactly the dimensional argument developed in Force Measurement Technology.

Second, a documentation gap worth naming: Swing Catalyst publishes sampling rate (1000 Hz for force plates, 150 Hz for the Balance Plate) and load limits, but no load-cell count, linearity, crosstalk, hysteresis, natural frequency, or accuracy figure. Bertec, AMTI and Kistler all publish those as a matter of course. Note also that the one indexed study using “Swing Catalyst” for force analysis in fact used Bertec plates with Swing Catalyst as the acquisition front end — which validates the software as a user interface, not the hardware as an instrument.


Part III: Where the Definitions Collide

III.1 Closure Rate — Four Incompatible Conventions

The worst interoperability trap in golf measurement, because the same phrase names quantities that differ by factors of two to five.

Convention What rotates, about what Typical range
Handle twist velocity The grip, about its own long axis 650–2,430 °/s
Foresight “The rotation of the club head heel to toe measured about the shaft”
GEARS Face rotation relative to the club path 360–620 °/s
Per unit distance Degrees of face rotation per foot of head travel 13–25 °/ft

Foresight’s shaft-relative definition and GEARS’ path-relative definition are not the same quantity, and neither vendor flags the discrepancy. Handle twist and clubhead closing velocity are related through the lie angle by \(\text{CCV} = \text{HTV}\sin(\text{lie}) + \text{SPV}\cos(\text{lie})\), with a measured ratio around 0.62 — so a player quoted 1,300 °/s on one system and 2,100 °/s on another may be perfectly consistent.

TipPrefer Degrees per Foot

Of the four, only °/ft is speed-invariant by construction. It is \(\omega/v\) — the reciprocal of the distance from the instantaneous screw axis — which is the dimensionally correct predictor of how much the reference-point choice matters. It is the only convention under which two players of different clubhead speeds can be compared meaningfully.

III.2 Instant of Measurement — And Why It Is Worth 3° per Millisecond

  • TrackMan: explicitly at maximum compression of the ball, for both path and face angle
  • Uneekor: “at the time of impact”
  • Full Swing: “prior to impact”
  • FlightScope: “measured at impact” — never disambiguated
  • Foresight: never states an instant for angular parameters. Its impact point is defined as first contact; closure rate as “just before the ball is impacted.”
ImportantOnly TrackMan Specifies Maximum Compression Precisely

Leach et al. put a number on it: “The face angle of a driver closes at a rate of 2.9°/ms immediately prior to impact… Therefore, the point in time at which the measurement is taken will significantly affect the output.”

Contact lasts roughly 0.45 ms. At the cited pre-impact closure rate, the difference between reporting face angle at first contact and at maximum compression can therefore be on the order of a degree. The precise effect on ball flight depends on speed, spin, launch conditions and aerodynamic modelling.

Of the documents reviewed, TrackMan is the only vendor that ties the reported club orientation to maximum compression while also limiting the measurement to pre-impact data. The other phrases identify the impact interval but do not locate a reproducible phase within the collision.

III.3 Nobody Actually Measures the Centre of Gravity

A limit that applies to every system, and which Leach et al. state plainly: CG location is specific to each clubhead model, and without it “neither Foresight nor TrackMan can measure the velocity of the COG directly, regardless of their definitions. Instead, velocity of a point on the surface of the clubhead is measured.”

So the reference-point taxonomy above describes what vendors report, not always what they sense. TrackMan reconstructs a geometric centre from the head’s radar silhouette; optical systems track markers on the face. Both then map to a declared reference point through a model.

III.4 An Inversion Worth Sitting With

The natural assumption is that optically measuring the face beats inferring it from radar. Leach’s data says otherwise: TrackMan calculated face angle and dynamic loft from other parameters rather than measuring them, and still agreed better with the gold standard than the optical system that measures them directly.

Foresight’s face angle landed within ±1° on 26% of shots and within ±2° on 46% — the worst parameter in the study, with an interquartile range of nearly 5°, against a then-published claim of ±0.5°.

There is a lesson in that for anyone building measurement systems, and it is not “cameras are worse than radar.” It is that a well-conditioned inference from robustly observed quantities can beat a direct measurement whose error budget is dominated by something mundane — here, plausibly, operator marker placement and the absence of any formal target-line alignment procedure.


Part IV: The Independent Validation Literature

Accuracy disclosure is sparse and uneven. Garmin publishes manufacturer tolerances, and historical tolerance tables exist for some products; independent validation is much rarer and is the stronger evidence for deciding how much confidence to place in club data.

IV.1 The Traceable Benchmark Study Located

Among the studies located in searches through 2026-08-03, Leach, Forrester, Mears & Roberts (2017), Measurement 112, 125–136, is the traceable validation study of consumer launch monitors rather than a comparison against another launch monitor. TrackMan Pro IIIe and Foresight GC2+HMT were tested simultaneously against a GOM Inspect reference reconstructed from four Photron cameras at 5,400 Hz, with calibration traceable to PTB Germany, over 240 shots from 8 golfers using driver, 7-iron and wedge.

The reference itself was characterised first: 0.06 mph in ball velocity, 0.01° in launch angle, 14 rpm in spin. So the disagreements below are real.

Parameter TrackMan bias Foresight bias
Ball velocity +0.2 mph +0.2 mph
Launch angle +0.1° +0.1°
Launch direction 0.0° −1.6°
Total spin −47 rpm −20 rpm
Clubhead velocity −1.1 mph +2.8 mph
Attack angle −1.4° (−3.5° for driver) +0.5°
Face angle 0.0° −0.8° (interquartile range −3.0° to +1.8°)
Dynamic loft −0.9° +2.2° (up to +5° for driver)

Judged against the authors’ “research grade” threshold of ±1 mph or ±1°, ball parameters pass on more than 80% of shots for both devices. Clubhead velocity passes on 54% of shots for TrackMan and 29% for Foresight.

ImportantThe Finding Nobody Quotes: Club Data Is Often Simply Absent

Leach also reports tracking success rates, and they are more alarming than the biases.

TrackMan tracked the ball on 98% of shots and clubhead velocity on 98% — but returned the remaining clubhead parameters on only 62% of shots, and on just 19% of utility-wedge shots, forcing the authors to exclude wedge club data entirely. Foresight tracked the ball on 90% and the clubhead on 75%.

A device that declines to report on a third of shots is doing something quite different from a device that reports on all of them. The authors also cautioned that the simultaneous multi-system setup, including tripods, lighting and possible interference, was not necessarily optimal for either commercial device, so these non-return rates should not be treated as universal field rates.

The authors’ own conclusion is measured and worth quoting in full, because it is neither a dismissal nor an endorsement:

For the scientific researcher, a high-level of confidence can be had in the ball parameters measured by both Trackman and Foresight; however, caution needs to be exercised in the use of clubhead parameters. Coaches, golfers and club-fitters should find the data to be of sufficient quality for most of their needs.

IV.2 Spin Rate Is the Consistent Weak Point

The reliability studies reviewed here identify spin or other reconstructed angular quantities as recurring weak points:

  • TrackMan 4 (Bishop et al., journal volume dated 2023; published online 2024): clubhead speed ICC 0.99, ball speed 0.97–0.99, carry 0.91–0.97 — but spin rate ICC 0.02–0.60, with significant between-session drift.
  • TrackMan (Shaw et al. 2023, 21 golfers): clubhead speed, ball speed and distances reliable; club path, attack angle and spin axis failed the reliability thresholds entirely.
  • Mevo+ (Brennan et al. 2024, 29 golfers): clubhead and ball speed r ≥ 0.92, but spin-rate 95% limits of agreement span roughly 7,700 rpm — wider than the entire physiological range.

Note that Leach’s bias for spin was modest (−47 rpm on TrackMan). So the spin problem is not systematic offset; it is shot-to-shot random error and estimation failure, and it worsens sharply on units that infer spin rather than measure it.

IV.3 A Circularity Problem

Both modern Mevo+ studies use TrackMan 4 as the criterion. That is device-versus-device agreement, not validation — and TrackMan 4 has no published gold-standard accuracy validation of its own, only reliability studies. Leach’s work was on the Pro IIIe, two generations earlier.

Bliss & Langdown (2026; available online 2025) illustrates what device-versus-device disagreement looks like between two premium units: attack angle ICC of 0.02–0.06 on irons, and dynamic loft differing by 7.8° to 9.7°. Their sample is a single golfer, which limits generalisation but not the basic point — two respected devices disagree about iron attack angle essentially completely.

IV.4 Devices Without Traceable Metrological Validation Located

Searches of OpenAlex, Crossref and Europe PMC through 2026-08-03 located no peer-reviewed study validating the following systems against a traceable metrological reference:

Foresight GCQuad, GCHawk, GC3 (only the older GC2+HMT, in Leach 2017) · Uneekor (all models) · Garmin Approach R10/R50 · Rapsodo MLM/MLM2PRO · SkyTrak/SkyTrak+ · Bushnell Launch Pro · Full Swing KIT · GEARS Golf · Sportsbox AI

This does not mean the devices are inaccurate, and it does not exclude studies of usability, repeatability or agreement against another launch monitor. Garmin’s ±1 mph ball-speed and ±3 mph club-speed figures, for example, are manufacturer-stated tolerances whose underlying reference, sample and protocol are not disclosed on the cited support page. GEARS’ “<0.2 mm research-grade accuracy” is likewise a manufacturer claim for which no peer-reviewed traceability study was located.

For contrast, the R&A’s bespoke club-tracking system (Corke et al. 2018) reports robot repeatability of ≤0.2° in face angle and club path and ≤0.1° in attack angle. Repeatability is not accuracy or traceability, but those figures illustrate the precision attainable in a purpose-built, controlled system.


Part V: On Slice, and What TrackMan Does and Does Not Publish

A specific question worth answering, because the answer is mostly a set of negatives.

There is no TrackMan article about slicing. The sitemap contains 122 blog URLs and none contains “slic”; a Help Centre search returns nothing on the topic. The closest published material is:

  • What is Face to Path? — the best vendor quantification of how much a given face-to-path curves the ball. For a PGA Tour driver at 275 yards carry, −2° of face-to-path ≈ 19 yards of left curvature; +5° ≈ 44 yards right. For a 6-iron at 183 yards, 2° ≈ 8 yards. A quoted TrackMan Master adds that “one degree Face to Path difference at 300 yards will produce a shot with 12 yards of curve” against a 30–32 yard fairway.
  • What is Spin Axis? — where TrackMan formally defines the shot shapes: “Large positive Spin Axis can be described as a slice”, with −2° to +2° counting as straight.
  • Six TrackMan numbers all amateur golfers should know — the only article naming the slice directly: “If you hit a fade or a slice, one of the reasons could be that you are swinging out to in.” It gives definitions only, no amateur values.
  • Is Zeroing Out Hurting Your Scorecard? — argues against coaching a player to a zero path, which is the closest any vendor comes to acknowledging the concern in Part I.1.
ImportantTrackMan Publishes No Club Path or Face Angle Averages, for Anyone

This is worth stating flatly because the opposite is asserted constantly.

TrackMan’s Tour Averages tables — the source of nearly every “tour average” figure in golf — contain exactly nine columns: club speed, attack angle, ball speed, smash factor, launch angle, spin rate, max height, land angle, carry. There is no club path column and no face angle column. Nor does TrackMan publish amateur averages of any kind; the only averages they publish are PGA and LPGA Tour.

So the widely repeated claim that “amateurs average an out-to-in path” is not traceable to TrackMan, and any amateur club-path or face-angle benchmark attributed to them is a third-party construction. The claim may well be true. It is not TrackMan’s.

For reference, the 2023 PGA Tour driver line is: 115 mph club speed, −0.9° attack angle, 171 mph ball speed, 1.49 smash, 10.4° launch, 2,545 rpm, 39° land angle, 282 yards carry. Methodology: 40+ events, 200+ players, competition and range shots combined.


Part VI: What Remains Under-Published

Stated plainly, because these absences shape what can honestly be claimed:

  • No current premium vendor reviewed here combines independent traceable validation with a comprehensive current tolerance table. Garmin publishes manufacturer tolerances, and historical TrackMan and SkyTrak figures remain available, but their evidence and scope differ.
  • Most vendors do not provide a complete measured-versus-calculated map. Garmin and Rapsodo are partial exceptions; Full Swing lists 16 data points without distinguishing them.
  • The vendor documents reviewed do not disclose per-parameter non-return rates comparable to independent studies. Leach et al. found club orientation data returned on 62% of shots for one premium device (19% for a utility wedge), subject to the study-setup caveat above.
  • No vendor quantifies the bias introduced by its own reference-point choice, TrackMan’s 3° figure excepted — and TrackMan frames that as a comparison, not as a bias in its own numbers.
  • No primary vendor source was located for amateur-population averages of club path or face angle.

Vendor documentation is undated more often than not, and is revised without notice. Links verified as resolving on 2026-08-03; quoted wording is verbatim from the pages as they stood on that date. Where a URL is dead or login-walled, that is stated rather than worked around.