Who this is for: anyone who wants to know how Pinpoint decides that something in a swing is worth mentioning — and how to change its mind. No technical background assumed, and no golf coaching background assumed either. Every term is explained the first time it appears.
How to read it: sections 1–4 explain how it works, ending with two worked examples. Section 5 says where everything lives in the app. Sections 6–8 are how to change things, what to be careful of, and what the app's own warnings mean. There is a glossary at the end.
⚠ What ships today, and what does not
Everything in this guide about how the numbers are stored, shown and edited is live in the app right now: the metric catalogue, the corridors, the editors, the health list.
The automatic per-shot findings — the app watching a swing and saying "this is scooping, and here is what usually causes it" — are not switched on yet. The knowledge is all in place and the screens that let you inspect and edit it work, but nothing is yet feeding real shots through it. So if you read Section 3 and go looking for a list of findings after a swing, you will not find one. That is the next piece of work, not a fault.
Pinpoint measures things about your swing, compares each measurement against what is normal for that kind of shot, and only says something when a measurement is genuinely unusual.
That is the whole idea. The rest of this section is the vocabulary, because each step in that sentence has a name in the app, and you need the names to find anything.
Here is the chain, and then each link explained:
a metric → a measure → a norm → a grade
(what we can (one specific (what normal (how unusual
measure) reading of it) looks like) this reading is)
│
▼
a characteristic ← a signal ←──────────────────── Watch or Action?
(something worth (a rule that
telling you) watches for it)
│
▼
its causes (what tends to produce it, and what it tends to produce)
A metric is a quantity the software can extract from a swing. "Ball position in the stance." "Lead wrist bow/cup." "Clubhead speed."
A metric on its own is just a name and a unit. It does not know what a good value is.
Golf term: "lead" and "trail". Your lead side is the side nearer the target — for a right-handed player, the left arm, hand and foot. Your trail side is the other one. The app always says lead/trail rather than left/right, so it reads correctly for both handednesses.
Here is the first thing that surprises people. A metric like "lead wrist bow/cup" is not one number — it is a number that changes continuously through the swing. So to compare it against anything, you first have to say exactly which reading you mean.
A measure is that decision. It names the metric plus how to reduce it to a single number:
| The measure says | Meaning |
|---|---|
| at P7 | the value at impact |
| change from P1 to P4 | how much it moved from address to the top of the backswing |
| highest between P5 and P6 | the peak somewhere in that stretch |
Golf term: P1 to P9. A standard way of naming eight or nine moments in a swing, so that everyone means the same instant. The ones you will meet most: P1 = address (set up, before you move), P4 = the top of the backswing, P7 = impact (the moment you strike the ball), P8/P9 = after impact.
This distinction matters more than it looks. "Lead wrist at impact" and "how much the lead wrist changed between address and impact" are two different numbers, about the same part of the body, and a good value for one is not a good value for the other. In the app they are two separate measures, each with its own idea of normal. Section 4 works through exactly this case.
A norm is the app's description of a normal range for one measure. It is also called a corridor, because of how it is drawn on screen: a band you would expect most people's readings to fall inside.
A norm holds:
- a centre — the typical value;
- a tolerance either side — how far from the centre is still unremarkable. The two sides can differ, and often should: for ball position, being a little forward is far more common and far less interesting than being a long way back.
A norm describes a population, never a person. It is "what golfers' readings look like", not "your target". There is no personal best hiding in here. That matters when you start editing them (Section 7).
Given a reading and a norm, the app grades it in four bands:
| Grade | Meaning in plain words |
|---|---|
| Ideal | right in the middle of normal |
| Good | comfortably normal |
| Watch | unusual enough to be worth noticing |
| Action | far enough out that it is very likely a real feature of your swing |
There is a fifth state, Not measured, which means the app could not get a reading at all — no camera, no sensor, or the swing could not be broken into positions confidently. Not measured never means "fine". It means we do not know, and the app is careful to keep those two apart.
A signal is a small rule attached to a measure. The commonest kind says: "tell me when this reading is unusual, on this particular side."
The side matters. Ball position being too far forward and too far back are different problems with different consequences, so they are two separate signals on the same measure, each watching one direction.
A signal fires on Watch or Action — not merely on leaving Ideal. This is deliberate and it is worth understanding: "Ideal" is the middle of normal, so a large fraction of perfectly ordinary golfers sit outside it on any given measurement. If the app spoke up every time, it would flag something on nearly every swing and you would stop reading it.
A characteristic is a named feature of a swing, in the language a coach would use: Scooping. Ball too far back. Coming out of it. When a signal fires, its characteristic becomes a finding for that swing.
A characteristic carries more than a name. It carries the consequence — why it matters. For example, for Scooping:
Adding loft through impact with the lead wrist means the club is rising at the ball, costing compression and distance control.
The library holds several different sorts of thing, and it treats them differently, because a coach does:
- a swing fault — a movement error, and what a lesson is usually about: over the top, early extension, casting
- a setup position — a static thing, before the swing starts: ball too far back, stance too wide. Often fixed inside one session, by being told
- a delivery number — what the club was doing at the ball: path, face angle, angle of attack, where the low point was. These are the scoreboard of the downswing rather than a fault in themselves, and they are what actually decide the shot
- an outcome — what the ball did: slice, thin, short carry. Where an explanation ends
- a physical capacity — what your body can and cannot do, like hip or shoulder range of motion. Settled by a physical test, never by a camera, and it constrains what is worth working on
- an intent — what you are trying to do or believe, like trying to lift the ball. Only you can confirm it, so it is always offered as a question and never concluded
That distinction is why the app can say "you are coming over the top, and limited lead-hip rotation may be why — here is a test that would settle it" rather than handing you a mobility screen and calling it your swing fault.
Some of them are marked as more common than others. That ranking is this library's own judgement, made to help sort what is worth looking at first — it is not measured prevalence, because nobody has counted swing faults across a population. Treat it as a coach's ordering, and expect it to change as the library learns.
Characteristics are linked to each other, because swings are not lists of independent faults. One thing produces another.
Scooping, for instance, is recorded as being caused by casting, by trying to lift the ball, and by hanging back — each link labelled with how strong the relationship is. Some causes are latent: they cannot be seen in the swing directly, only inferred from the things they explain.
This is what makes the model useful rather than just a list of complaints. Several visible characteristics often trace back to one underlying cause — and that one cause is the thing worth working on.
The same reading can be normal for one shot and odd for another. A driver is played well forward in the stance; a wedge is played near the middle. One "normal ball position" for both would be wrong for both.
So norms are organised by shot type, called a context in the app, arranged as a family tree:
Any shot ← the general case lives here
├── Full swing
│ ├── Driver
│ ├── Fairway wood
│ ├── Iron
│ ├── Wedge
│ ├── Bowed top (a style, not a club — see Section 4)
│ └── Cupped top
├── Partial swing
│ ├── Pitch
│ └── Chip
├── Bunker
└── Specialty
How a norm is found. The app starts at the shot's own context and walks upward until it finds a norm for that measure. The nearest one wins. So for an iron shot it looks at Iron first, then Full swing, then Any shot.
Two consequences worth holding on to:
- You only author what is different. If a measure behaves the same whatever club is in hand, one norm at Any shot covers everything. Add a Driver norm only when the driver genuinely differs.
- The general case belongs at Any shot, not at Full swing. A full swing is a type of shot, with no more claim to being the default than a pitch has. If you put a general norm on Full swing instead, a pitch or a bunker shot cannot see it — those branches sit beside Full swing, not under it — and every reading on such a shot comes back Not measured.
That second point is a real trap and the app now watches for it: the health list (Section 8) reports any context that nothing grades.
Shot type informs a diagnosis. It never silences one. Pinpoint will not decide that a characteristic is irrelevant to your bunker shot and quietly skip it. It measures, grades and reports; whether the result matters for the shot you were playing is your call, not the software's.
The simplest end-to-end case, and the one where shot type does real work.
Ball position is measured as a percentage across your stance:
- 0 % = level with your lead heel (the foot nearer the target)
- 100 % = level with your trail heel
- 50 % = the middle of your stance
A higher number means the ball is further back, toward the trail foot. (A driver is sometimes played slightly ahead of the lead heel, which reads as a small negative number. That is a real setup, not an error.)
This is read at P1 — address, before you move. The measure is therefore "Ball position in the stance, at P1".
| Context | Centre | Tolerance | Ideal band |
|---|---|---|---|
| Any shot | 30 % | ± 14 | 16 % – 44 % |
| Driver | 5 % | ± 8 | −3 % – 13 % |
| Fairway wood | 18 % | ± 9 | 9 % – 27 % |
| Iron | 33 % | ± 10 | 23 % – 43 % |
| Wedge | 50 % | ± 10 | 40 % – 60 % |
Read that table as golf, not arithmetic: the driver is played off the lead heel, the wedge in the middle, irons a little back of centre — and if the app does not know which club you used, it falls back to a wide general band.
Take an iron shot with the ball at 50 %.
The app finds the Iron norm: centre 33, tolerance 10. The reading is 17 above centre, which is 1.7 tolerances out. Under the standard setting that is Good — noticeably back of typical, but not unusual enough to be a finding. Nothing is reported.
Now the same 50 % on a wedge. Centre 50, tolerance 10 — dead centre. Ideal.
The identical measurement, graded differently, correctly, because the shot was different. That is the context tree earning its place.
Push the iron further back, to 60 %: now 2.7 tolerances out, which is Watch. The signal watching the "too far back" side fires, and the characteristic Ball too far back becomes a finding, with its consequence:
A ball played too far toward the trail foot is met early and steeply, which tends to close the shoulders at address to compensate.
Ball position has two characteristics, one per side — Ball too far forward and Ball too far back — because they are different problems. They are recorded as two ends of one axis, so the app knows they are a pair rather than two unrelated faults.
Each is also linked into the causal model:
- Ball too far back → tends to cause closed shoulder alignment (a moderate link). Setting up with the ball back encourages you to close your shoulders to compensate.
- An open aim bias → weakly tends to cause Ball too far forward.
So if the app reports both a ball-position finding and an alignment finding, the model already knows they may be one story rather than two.
Harder, and worth the effort — it shows why "which reading" matters, and why two golfers can both be right.
Your lead wrist can be bowed (flexed — the back of the hand angled toward the forearm, the strong "hands ahead" look) or cupped (extended — the back of the hand angled back). The app measures this in degrees, with positive = more bowed and negative = more cupped.
It matters because this angle is the biggest single influence on where the clubface points. Cupping tends to leave the face open; bowing tends to square or close it.
The wrist angle changes constantly through the swing, so the app carries two different measures of it — and this is the part to slow down on:
| Measure | What it is | Its norm |
|---|---|---|
| Lead wrist bow/cup — change from address at P4 | How much your wrist moved between address and the top of the backswing | centre +5°, tolerance ± 11 |
| Lead wrist flexion/extension, at P7 | The actual angle at impact, not a change | centre +15°, tolerance ± 12 |
The first is a change; the second is an absolute reading. +10° means two entirely different things depending on which one you are looking at. On the app's screens, a corridor that is a change from address is labelled as such, precisely so the two cannot be confused.
Impact norm: centre 15, tolerance 12, giving Ideal from 3° to 27°.
- A reading of +15° (nicely bowed at impact): Ideal.
- A reading of −10° (cupped at impact): 25 below centre, 2.1 tolerances out → Watch, on the low side.
That low side has a signal on it, and the characteristic it belongs to is Scooping:
Adding loft through impact with the lead wrist means the club is rising at the ball, costing compression and distance control.
Note that the signal watches only the low side. A wrist more bowed than typical at impact is not scooping — it is a different thing entirely, and it would be wrong for one signal to fire on both.
Some excellent players arrive at the top with a bowed wrist; others arrive cupped. Both win golf tournaments. If the app graded everyone against one corridor, it would flag one of those two groups as faulty on every single swing.
So Bowed top and Cupped top are contexts — shot "types" in the family tree, sitting under Full swing — each with its own corridor for this one axis:
| Context | Centre of the change-to-the-top norm |
|---|---|
| Any shot (neutral) | +5° |
| Bowed top | +15° |
| Cupped top | −5° |
A player identified as bowed is graded against the bowed corridor. Every other measurement is unaffected — only this axis shifts, because style only makes a difference here.
Scooping is recorded as being caused by:
- casting (strong) — releasing the angle in the wrists too early in the downswing;
- trying to lift the ball (strong) — the instinct to help the ball into the air;
- hanging back (moderate) — weight staying on the trail foot through impact.
This is the payoff of the causal model. Told only "you are scooping", you would work on your hands. The model points at three upstream candidates instead — and trying to lift the ball is an intention, not a movement, which is a very different conversation.
The library now names the shots themselves, not only the swings that produce them: pull, push, slice, hook, chunk, thin, top, sky, shank and the rest. They are ordinary characteristics, in a group called Ball flight, which is what lets an explanation run all the way from a physical restriction to the shot you actually saw — a tight lead hip causes early extension, early extension causes the club to come from under, and that causes the block right.
Two things are worth understanding about them.
Some need a launch monitor, and — for now — so do the rest. How the ball CURVES develops over a flight an indoor capture never sees, so slice, hook, spin and strike location need a launch monitor. Until one is connected those read "needs a launch monitor" rather than going quietly blank — the same way a metric says "needs a face-on camera" when a camera is missing. Launch-monitor integration is planned; when it arrives those readings simply start working, with no change to the library.
Where the ball STARTS — start direction, launch height, ball speed — is in principle camera work, and we do not have it yet either. The reason is worth stating plainly rather than leaving as a vague "coming soon": the app watches the ball to know when you hit it, not where it went. It locks onto the ball sitting at address and notices the instant it disappears, which is exactly what shot detection needs and exactly not what a launch reading needs. Following the ball through the air is a different piece of work, and until it is built those readings stay marked as planned.
Some the app cannot yet tell apart, and says so. A chunk is the ground struck before the ball AND a collapse in ball speed — two readings at once — and the app currently judges one reading at a time. So chunk, thin, top, sky and shank are marked as things only you or your coach can confirm. They are still in the library, still carry their definitions, and still sit at the end of the causal chains; the app simply does not claim to have seen one.
Draw and fade are deliberately absent. A draw is a hook you meant. The shape is identical and the app has no way to know your intention, so it reports the curvature and leaves whether it was wanted to you — the same reason nothing in this model calls a finding good or bad.
Every reading in the app comes from one camera, in front of you. That camera sees the plane you stand in — up and down, and side to side — with real precision. What it cannot see directly is depth: movement toward and away from it.
That single fact explains most of what is and is not in the library:
- Side-to-side and up-and-down readings are straightforward. How far your hips slide, whether your shoulders are level at the top, how high the trail elbow gets, whether you finish balanced over the lead foot — all of these are read directly, and this release added nine more of them.
- Rotation is not, but it can be estimated — and the app tells you so. How far your hips and chest have turned is rotation about your own axis, which a camera in front of you cannot see head-on. What it CAN see is that your shoulders look narrower as you turn away from it, and how much narrower says how far you have turned. That is a real reading, and it is a weaker one than a sensor strapped to your back would give. So the app labels it estimated rather than measured, and it is least certain when you are nearly square — right at address and right at impact. If body sensors are ever added, the same reading quietly becomes a measurement, with nothing else changing.
- Some things it genuinely cannot do, and will not pretend to. How far your hips push toward the ball, which way the club is travelling in-to-out or out-to-in, and how far you stand from the ball are all depth. They need a second camera down the line, and they are marked as such rather than guessed at.
The principle behind all three: give you the best reading the equipment can support, say which kind of reading it is, and never put a confident number on something that was not measured.
| What you want | Where to go |
|---|---|
| Every metric the app can measure, what it means, its corridors | Settings → Metrics |
| The corridors themselves, to inspect or edit | Settings → Diagnostics → Measures & norms |
| Characteristics, their consequences, their causes | Settings → Diagnostics → Characteristics |
| What a coaching term means, in plain language | Settings → Diagnostics → Glossary |
| The physical tests, and the drills that answer a fault | Settings → Diagnostics → Screens & drills |
| What is missing or wrong in the library | Settings → Diagnostics → Causes & health |
| What is not yet measurable, and why | Settings → Diagnostics → Roadmap |
Search by the word you were taught. Every characteristic carries the other names coaches use for it, so flip finds Scooping, OTT finds Over the top, and standing up finds Early extension. That works in the Characteristics list and in the Glossary, and the row tells you which term matched so you can see the library understood you rather than guessed.
A shortcut worth knowing. On any metric's page in Settings → Metrics, each corridor is drawn as a band with a coloured link underneath it naming the norm behind it — for example "Lead wrist bow / cup — change from address at P4 · Any shot →". Tapping it takes you straight to that norm, where you can edit it. There is one link per corridor, because (as Section 4 showed) different phases of one metric are different measures with different norms.
Settings → Diagnostics → Measures & norms, choose the measure, then tap the norm row for the context you want to change. The corridor editor opens.
It offers three routes:
- Drag the two handles to set the band by eye.
- Type the numbers if you have a figure in mind.
- Seat it from swings — let the app fit the corridor to a set of stored swings you choose.
Underneath, it draws a histogram: every stored swing's reading for this measure, with your corridor overlaid, and a running count of how many land in each grade. Watch that histogram. It is the single best check that a corridor is sane: a band that grades almost every swing in your library as Action is visibly wrong without needing any statistics, and so is a band so wide that nothing ever falls outside it.
The editor also tells you where a corridor came from — whether it is the shipped figure or one of yours — and offers the matching undo:
- Reset to shipped — when Pinpoint ships a corridor at exactly this context, this puts it back.
- Remove your override — when it does not. Your row is dropped and the corridor is inherited from further up the tree instead. Both restore the app's own behaviour; the wording differs because what you get back differs.
Open the characteristic in Settings → Diagnostics → Characteristics and edit the measure attached to it. Instead of asking you to pick "high" or "low", the app shows you the measure's own words — for ball position, "further back, toward the trail foot" versus the other end of that range — and you pick the sentence that describes what the characteristic actually is.
This is not decoration. Picking the wrong side is the single most damaging mistake available here, because the result looks completely correct: the app fires confidently, with sensible-sounding text, on exactly the wrong swings. Reading a sentence rather than choosing a direction is what prevents it.
From the same editor, "add a measure" opens a picker. You choose the metric, then how to reduce it (at a position, a change between two positions, a peak across a stretch).
The picker will refuse to create a measure that does not state what a high value means. That sentence is what the direction control shows later, so a measure without one cannot be attached safely.
The characteristic's page shows its consequence text, what detects it, and a map of its causes and effects — tap any box to re-centre the map on it, or long-press for options including adding or removing a link. Removing a link offers an undo in the same breath, which puts it back with its original strength.
Settings → Diagnostics → Measures & norms carries the grade policy, which applies to the whole library:
| Setting | Effect |
|---|---|
| Lenient | flags less; suits a wide range of styles |
| Standard | the shipped setting; ordinary variation is not a finding |
| Strict | flags more; suits a narrow, coached population |
It is one choice for everything rather than a slider per corridor, deliberately: if "Ideal" means one thing on one measurement and something else on another, nothing can be compared.
The honest list. Most of these are mistakes that produce confident, plausible, wrong results, which are far more dangerous than obvious breakage.
1. You are editing a population norm, not your target. Widening a corridor because your own swings fall outside it does not improve your swing — it removes the app's ability to tell you about it. If you want to track your own progress, that is a different feature; this is "what is normal for golfers".
2. The numbers that ship are starting points, not findings. Every corridor Pinpoint ships is an authored estimate — the app marks them as such, with no sample size behind them. They are expected to move as real data accumulates. Do not treat a shipped band as established fact, and do not be shy about correcting one you have good reason to doubt.
3. Never judge a corridor on one swing. A single reading outside a band tells you almost nothing — about the swing or the band. The histogram over your whole library is the evidence; one swing is an anecdote.
4. Check which reading you are editing. The most common confusion in the whole system. "At impact" and "change from address to impact" are different measures with different corridors, and they look almost identical in a list. Read the measure's full name before you drag anything.
5. Units can match while conventions do not. The app refuses a corridor whose unit differs from its measure's — degrees against a percentage is caught immediately. What it cannot catch is a corridor written against a different convention in the same unit: two things both labelled "% of shoulder width" where one means something else. If readings sit consistently at roughly double or half your corridor, suspect this before adjusting the numbers.
6. A corridor that never says anything is a broken corridor. It is tempting to widen a band until nothing gets flagged. The result cannot report a deviation at all. The health list watches for both failure modes — almost everything outside, and almost everything inside — and reports them differently, because they have opposite fixes.
7. Mind where you author, in the shot-type tree. Editing a corridor at Full swing creates a row that applies to full swings only — pitches, chips and bunker shots carry on using the general one. That may be exactly what you want. If you meant "change this for everything", author at Any shot.
8. Bowed and cupped are both correct. Before deciding the wrist corridor is wrong for you, check which style context your swings are being graded against. Two valid styles is the whole reason those contexts exist.
9. "Not measured" is not "fine". If a finding is absent because no camera saw the swing, or no sensor was attached, or the swing could not be segmented, that is silence — not a clean bill of health. The app keeps these apart everywhere and never collapses one into the other; be equally careful when reading it.
10. Editing a shared measure affects every characteristic using it. Before you change one, the app tells you how many characteristics ride on it and which ones. Read that. A measure feeding four characteristics changes all four.
Settings → Diagnostics → Causes & health is the app checking its own knowledge, grouped by kind. The messages are written in plain language, and here is what the common ones mean:
| What it says | What it means | What to do |
|---|---|---|
| Nothing to compare against — the signal cannot fire | A characteristic is watching a measurement that has no corridor anywhere. It can never report anything. | Author a corridor, or accept that this characteristic is undetectable for now |
| A context graded by nothing at all | Some shot type has no corridors anywhere up its branch, so every reading on such a shot comes back Not measured | Author corridors for it, or move it under a context that has them |
| A context that changes no grade | A shot type with no corridors of its own. Harmless — it grades exactly as its parent does | Nothing, unless you intended it to differ |
| Grades almost the whole library into one band | The corridor is in the wrong place, in the wrong unit, or so wide it can never speak | Open it and look at the histogram |
| Your corridor, seated on no swings | You typed a corridor rather than fitting it to data. Fine as a starting point | Seat it from swings when you have a sample |
| Club-dependent, but graded by one corridor | The metric's own description says the number depends on the club, and only one corridor exists | Add per-club corridors |
| Corridor and measure are in different units | A real error — degrees graded against percent, or similar | Fix the unit; the grade is meaningless until you do |
| Flags both sides at once | One characteristic is watching both ends of a range, so it fires whichever way the reading goes and cannot tell too much from too little | Remove one side; the two ends belong to two characteristics |
| The shipped corridor has been revised since you changed it | You overrode a corridor, and Pinpoint has since updated its own version | Compare, then keep yours or take theirs |
There is also a Check corridors button. It reads every stored swing, grades it against every corridor, and reports any corridor that puts nearly everything in one band. It needs a swing library configured (Settings → Storage), and it tells you how many swings it looked at — so "nothing found" is never confused with "nothing checked".
| Term | Meaning |
|---|---|
| Address | Set up to the ball, before the swing starts. Also P1 |
| Axis | A pair of characteristics describing the two ends of one range (too far forward / too far back) |
| Bowed | Lead wrist flexed, back of the hand angled toward the forearm. Positive degrees |
| Characteristic | A named feature of a swing, in a coach's language, with a consequence |
| Context | A kind of shot — a club, a style, or a shot type — that norms can be organised by |
| Corridor | Another word for a norm, from how it is drawn: a band of normal readings |
| Cupped | Lead wrist extended, back of the hand angled back. Negative degrees |
| Finding | What a characteristic becomes when a signal fires on a particular swing |
| Grade | Ideal / Good / Watch / Action — how unusual a reading is. Plus Not measured |
| Impact | The moment the club strikes the ball. Also P7 |
| Latent | A cause that cannot be seen directly, only inferred from what it explains |
| Lead / trail | Lead is the side nearer the target; trail is the other. Handedness-neutral |
| Measure | A metric plus exactly which reading of it is meant (at impact, change to the top, …) |
| Metric | A quantity the software can extract from a swing |
| Norm | The normal range for one measure — a centre and a tolerance either side |
| P1 … P9 | Standard names for moments in the swing. P1 address, P4 top, P7 impact |
| Signal | The rule that watches one measure on one side and fires when a reading is unusual |
| Tolerance | How far from the centre still counts as unremarkable. Can differ on each side |
| Top | The end of the backswing, where the club changes direction. Also P4 |