Home Practice Drills You Can Run From Your Own Swing Data

Most golfers filming themselves in a garage have the same problem: plenty of footage, no idea which numbers on the screen are real. An app tells you your pelvis slides 6 inches toward the target, or your club path is 4° in-to-out, and you nod along without knowing whether the software actually measured that or inferred it from a blurry 30fps clip shot from chest height. Getting golf practice drills at home to work depends almost entirely on solving that first problem, because a drill aimed at a metric the software got wrong is worse than no drill at all.

This page assumes you already have the kit: a phone, a net, a mat, maybe a Garmin R10 or a Rapsodo MLM2PRO sitting in a cupboard since Christmas. What follows is the capture protocol that makes the AI readings trustworthy, an honest account of which tools handle phone footage properly, and five drills that map directly onto specific numbers those tools produce. For how this fits into a longer training block across a season, the broader overview lives in the pillar on turning analysis into a practice plan.

What a phone camera actually hands the model

Pose-estimation software finds joint centres in a 2D image and then, in the 3D tools, reconstructs depth using a learned model of human proportions. It is genuinely clever and genuinely fragile. Four things wreck it, in rough order of how often they wreck it in a British garage:

Low light forces the phone to drop shutter speed, which smears the club and the hands into a grey streak. At 1/250s or slower your hands at transition are three inches wide on the sensor. Sportsbox, Onform and every other tool will still hand you a number, and it will be soft by several degrees. Fix: two 5000K work lights, roughly £25 each from Screwfix, pointed at the ceiling rather than at you.

Baggy clothing is the second killer. A fleece hides the pelvis entirely, and pelvis sway is exactly the metric most 6-to-24 handicappers need. Wear something fitted for filming even if it feels daft in an unheated garage in February.

Camera height matters more than distance. Down-the-line footage should be shot with the lens at hand height at address, not at eye height, and on the hand line rather than the toe line or the ball. Raise the phone two feet and your shaft plane changes by around 5° on screen without a thing changing in your swing. Face-on goes at sternum height, square to your stance line, about 8 to 10 feet away with your head and the clubhead both inside the frame at every point in the swing.

Frame rate is the one people obsess over and it is the least of the four. 60fps is the floor for 3D pose tools. 120fps is comfortably enough for club delivery. iPhone 13 and later will do 1080p at 240fps, which is lovely for looking at impact but produces files big enough to choke an upload, and most of the AI tools downsample it anyway.

Which tools read the footage correctly

The honest headline: Sportsbox is the only widely available consumer tool that turns a phone clip into body numbers you can act on, and its own published validation work against marker-based motion capture is much stronger for large segments like the pelvis and thorax than for the hands and wrists. Treat its pelvis and thorax outputs as accurate to a couple of degrees when your capture is clean, and treat its wrist data as a rough indicator only. HackMotion exists precisely because wrists need a sensor.

Reading the session without fooling yourself

Here is a real-shape R10 session, twelve 7-irons before a change and twelve after, indoors into a net:

Set  Ball(mph)  Club(mph)  Smash  AoA    Path    Face   F-to-P  Carry(y)
A    108.4      78.1       1.39   -1.2   +5.8    +2.4   -3.4    148
B    110.9      78.4       1.41   -3.6   +3.2    +1.0   -2.2    153

Club speed barely moved. Everything that improved came from strike and delivery, and that is normal. The five-yard carry gain is worth roughly nothing as an absolute number indoors, because the R10 calculates spin rather than measuring it and indoor 7-iron carry is routinely overstated. The gain is worth a great deal as a relative number: same ball, same mat, same session, so the comparison is fair even when the absolute is not.

Two rules keep this honest. First, never compare a number from today’s session to a number from three weeks ago on a different mat. Second, ignore any single shot. Work in blocks of ten to twelve and look at the median, not the best one.

Drill one: the sway gate

Trigger metric: Sportsbox pelvis sway greater than about 4 inches away from target at P4, or lead-side sway over roughly 4 inches at P7 where tour players sit closer to 2 to 3.

Set an alignment stick vertically in the ground (or a broom against a wall) just outside your trail hip at address, close enough that a 2-inch slide touches it. Make ten slow half-swings to P4 without contact, then ten at three-quarter pace hitting balls. Film every fifth one. You are looking for the sway figure to come down without the turn number collapsing with it, which is the usual trade: golfers stop sliding by stopping turning, and pelvis turn at P4 drops from 45° to 30°. If that happens, the drill is failing and you should stop.

Dose: 3 sets of 10, twice a week, six weeks. Re-film on week 3 and week 6 only.

Drill two: hands in front of the chest

Trigger metric: hand depth at P6 significantly behind the address hand line, usually paired with thorax bend at P7 that has lost 8° or more from address. That combination is early extension, and it is the single most common finding in a 6-to-24 handicap sample.

Place a folding chair or a stack of two cushions just behind your backside at address, touching your glutes lightly. Swing at 50% and keep contact with the chair until well past impact. Most people fail this within three swings, which is the point. Progress to hitting balls at 70% once you can make ten consecutive contactless-free reps.

The measurable you are chasing is thorax bend at impact holding within 3° of its address value. Sportsbox will show this; a 2D app will only show you a shape you can argue about.

Drill three: the low-point ladder

Trigger metric: R10 or MLM2PRO angle of attack with a 7-iron shallower than about -2°, or wildly variable shot to shot (say, a spread from +1° to -6° across twelve balls).

Draw a line across your mat with a whiteboard marker, or lay a strip of masking tape, and place the ball an inch behind it. Hit ten shots aiming to bruise the mat on the target side of the tape. Then move the ball forward half an inch and repeat. Then back an inch. The ladder trains low point control under a changing constraint rather than grooving one position.

Watch AoA consistency, not its average. A player whose standard deviation across twelve shots falls from 2.4° to 1.1° has genuinely improved, even if the mean barely shifted.

Drill four: the face-to-path gate

Trigger metric: face-to-path outside ±2° with a mid-iron, or a face-to-path number that flips sign between shots.

Build a gate from two tees or two headcovers about 18 inches in front of the ball, set roughly a clubhead and a half apart. Choose a target shape before each shot and call it out loud: “small draw.” Hit, then check the face-to-path on your monitor before the next ball. Calling the shape first turns it into a skill test rather than a mechanical one, and skill tests transfer to the course far better.

Score it: 12 shots, one point for each where your called shape matched the sign of the face-to-path reading. Below 7 out of 12, you have a control problem worth a lesson. Above 10, move on to a different drill; this one has nothing left to teach you.

Drill five: speed blocks that don’t cost you your swing

Club speed is the one thing on an R10 you can trust in absolute terms, give or take a mile or two per hour. That makes overspeed work the easiest thing to run at home, and the easiest thing to overdo.

Run three sets of three swings with a light stick, a normal stick and a heavy stick, on non-consecutive days, and log the fastest reading in each set. A typical 14-handicap man starting at 88mph might reasonably reach 92 to 94mph in eight weeks. Progress stalls if you do it more than three times a week, and strike quality reliably falls apart if you run speed work immediately before any of the four drills above. Put it at the end of the session, never the start.

The re-test, and why it matters more than the drills

Book the re-test into your calendar the day you start, six weeks out, before you have any emotional stake in the result. Same garage, same lights, same clothes, same ball, same two camera positions marked on the floor with tape so you can put the tripod back in exactly the spot. Film twelve 7-irons and run the same three Sportsbox captures.

If pelvis sway has come down by 1.5 inches and AoA variance has halved, you have a training method that works for you and you should keep it and change only the target metric. If nothing has moved, the problem was almost certainly never the drill. Go back and check the capture first: the lighting, the camera height, the fleece.