ShotIQ Learn/⚙️ Inside ShotIQ

Where ShotIQ Says "I Don't Know"

There is a question you should ask any app that tells you things about your golf swing: what does it do when it doesn't know?

It's a better question than "is it accurate," because accuracy is easy to claim and hard to check. What an app does with a gap in its knowledge is visible, testable, and tells you almost everything about whether the rest of its numbers deserve your trust.

Most consumer golf analytics fill the gap. They infer which club you hit from how far the ball went. They average a bladed wedge in with eleven good ones. They score a swing metric the camera was never in a position to see. Each of those is small and invisible on its own, and they compound into a system you stop believing around week six — usually without being able to say exactly why.

This is a complete, specific list of the places ShotIQ refuses. If you find one it's missing, we'd genuinely like to hear about it.

1 · Your carry distance never tells us which swing you made

The Wedge Matrix is built around which swing a shot was — a 50%, a 75%, or a full one. It would be trivial to work that out from carry distance, and it would be wrong.

A wedge carry means nothing on its own. Fifty-six yards is a full swing for one golfer and a half swing for another, and the same golfer's "full" changes with the wind, the lie and how the day is going. So a shot arrives with its swing declared by you, captured before you hit it in a guided session, or read from the file's own metadata — and if none of those happened, it's filed as Unknown and the app asks you later.

Five things a shot's intent can be: user_labeled, guided_session, import_metadata, inferred, unknown. Distance is not on that list.

2 · A metric the camera couldn't see is not a metric we scored

The Swing Analyzer measures shoulder turn, hip turn, X-Factor and posture from phone video. Some of those are only visible from a specific camera angle.

Rather than producing a plausible-looking number from the wrong view, it tells you exactly what it couldn't measure and how to fix it:

"This one needs a face-on video. From down-the-line the camera is looking along the target line, so it can't see movement toward or away from the target — only toward or away from the ball. Film one swing facing the camera and this fills in."

We learned this the hard way. Hang-back was being scored from down-the-line video, where it is geometrically invisible — the axis it happens on runs straight at the lens. Every golfer filming from that angle had been graded on something the camera never saw. That's now gated, and the gate is the point.

3 · Mishits are judged against your own club, never a textbook

Thin, fat, off-centre and flier are all classified relative to that club's own median in that session — never against an absolute threshold.

"Apex under 40 feet is thin" is wrong for a 60° wedge and wrong again for a senior flex. "This shot peaked at a fifth of the height your other eleven did" is true for everyone. You calibrate yourself.

And below six shots with a club there's no trustworthy median to compare against, so every shot comes back unknown — deliberately not clean, because clean would be a claim we can't support.

4 · We remove fewer of your bad shots than we could

Stripping mishits out of a carry average is an obvious feature, and we deliberately under-apply it. Only thin and fat — the two unambiguous contact failures — come out by default. Off-centre strikes stay in.

Missing the middle is normal golf. Nobody finds the centre every time, and excluding those shots would quietly inflate every carry in your bag, which is precisely how you end up one club short on a par three. The first version of this raised a 7-iron from 132 to 135 yards. That was flattering, and wrong, so we changed it.

5 · Bad data is stopped at the door, with the reason stated

A club-fitting session where you hit four demo drivers, a warm-up block of cold swings, and a scored combine are indistinguishable from range practice by inspection — and all three will quietly corrupt your bag distances.

Uploads declare what they were (eleven session types, from range to fitting to warm-up), and where the data is allowed to go is enforced when your bag is read, not merely suggested in the interface. Each refusal explains itself:

"You may have been hitting clubs that aren't yours — this is the single fastest way to corrupt your bag."

A destination that's off because the file can't feed it ("no hole data") looks different from one that's off because the session type says so — and you can override the second but not the first.

6 · Your Golf DNA leaves fields blank

This one we got wrong for a while, and fixing it is instructive.

The Golf DNA card reports your shot shape, typical miss and play style. It used to print a confident word in every field — and for three of them, that word was the fallback, rendered identically to a real finding.

"Typical miss: Balanced" meant there are no logged on-course shots and no side bias — that is, we don't know. "Play style: Balanced" was the numeric midpoint that four of the nine player archetypes map to by default. Worst of all, "Shape: Straight" was computed from the average of your side-carry — so twenty shots twenty yards right and twenty shots twenty yards left averaged to zero and were reported as straight. The most flattering possible reading of the least controlled possible ball flight.

Now a shape is only claimed when your pattern genuinely sits on one side — the average has to clear its own scatter. Wide two-way scatter is reported as "Two-way miss", which is the actual finding. And where nothing is known, the card says so and tells you what would answer it.

7 · Distance comparisons need two facts, and won't run on one

When the app compares your carries to "typical", it uses published reference distances that are banded by handicap and split by men's and women's sets — because those differ by 40 to 60 yards on a driver.

If you haven't set a handicap, or haven't chosen which set to compare against, the comparison doesn't appear at all. It doesn't quietly assume you're a 15, and it doesn't guess which set applies to you. Two clubs — the 2-iron and 3-iron — have no published row, so they get no comparison rather than one borrowed from a neighbouring club. Where a women's row genuinely doesn't exist, the nearest club stands in and the app tells you it's standing in.

Comparisons are shown as a percentile against golfers of your own handicap and gender, not a raw yardage, because "+13 yards" is a number with no scale and "71st percentile" is a claim you can act on.

8 · Links to things that don't exist are never drawn

A small one, but it's the same principle at the level of a button.

Fifteen launch monitors have a written export guide. If you pick a device that doesn't have one, you get nothing — not the closest-looking guide, because sending you to another device's export steps is worse than sending you nowhere. Thirteen of seventeen launch-monitor metrics have a full written guide, and the four that don't show no link at all. When the AI coach references a guide, it can only reference one that exists; if it ever gets a name wrong, the reference silently disappears rather than becoming a dead link in the middle of a sentence.

Why this costs us something

Every one of these makes the app look less capable in a demo. A screen with three confident verdicts sells better than one with a verdict and two honest blanks. An app that infers your swing type from distance has a finished-looking wedge matrix on day one; ours asks you a question.

We think that trade is correct, for a reason that isn't really about principle: a golfer eventually catches an app being wrong about something they know for certain. They know they hit that one fat. They know their 9-iron doesn't go 150. The moment the app says otherwise with total confidence, everything else it has ever told them becomes suspect at once — including the parts that were right.

A number that declines to exist can't produce that moment.

How you can check

The rules above aren't a description of intent. They're executable, in 23 test suites covering the pure calculation engines, kept separate from all interface code so they can be run and read on their own. The refusals specifically are tested: that a shot with no declared intent stays Unknown, that below six shots the answer is unknown and not clean, that a device without a guide renders nothing, that a missing handicap or benchmark set produces no comparison at all.

Frequently asked questions

Does this mean ShotIQ shows me less than other apps? In a few specific places, yes — deliberately. You'll see fewer filled-in fields on day one, particularly before you've logged a round or set your handicap. What you get in exchange is that the fields which are filled in mean what they say.

Why does my Golf DNA have blank fields? Because those specific things aren't knowable from what you've uploaded yet. Typical miss needs logged on-course shots to tell short from long. Shot shape needs enough shots to separate a real bias from scatter. Each blank field says what would answer it.

Why can't I see how my distances compare to other golfers? That comparison needs both your handicap and which published benchmark set to use — men's and women's reference carries differ by 40 to 60 yards on a driver. Set both in Settings and the ◆ markers appear on every club in your bag.

Why is my wedge shot marked "Unknown" instead of 50% or 75%? Because nothing told us which swing it was. Distance alone can't — 56 yards is a full swing for one golfer and a half swing for another. Tag it yourself, or run a guided calibration session where the swing is declared before you hit.

My launch monitor gives me club path and face angle. Why doesn't ShotIQ show them? Check your device guide. Some units measure club delivery directly, some estimate it from ball data, and some don't capture it at all. Where the hardware doesn't measure it, ShotIQ won't display it — no app can honestly show you numbers the device never collected.

Is any of this actually different from other golf apps? The individual rules aren't exotic. Applying them consistently, in one place, where they're testable rather than remembered, is the part that's hard — and it's why we wrote this page instead of just claiming to be accurate.

Related guides

• How ShotIQ Works: the physics and the math — the calculations themselves, rather than the refusals
• Every ShotIQ Feature, Explained
• The device directory — what each launch monitor actually measures