Radar or Camera Indoors? Launch Monitor Accuracy in a Garage

Your garage is a hostile environment for a launch monitor, and it is hostile in two completely different ways depending on which kind you bought. Radar units are starved of the thing they exist to measure: ball flight. Camera units are starved of light. Neither problem is a bug, neither gets fixed by a firmware update, and the two failure modes look nothing alike on screen. One quietly invents numbers that are plausible. The other refuses to give you numbers at all, or gives you one wild reading in twenty.

Search launch monitor accuracy indoor and you get a lot of forum arguments about which box is “closer to Trackman”. Wrong question for a UK garage. The right question is: given roughly five metres of length, two and a half metres of ceiling, one LED strip and a net, which measurements are still real, and which is your device guessing? Because you can practise productively on the real ones and you will ruin your golf swing chasing the guesses.

Radar needs metres of flight and your net is 2.4 of them away

Doppler radar works by watching a moving ball and fitting a trajectory to it. Garmin asks you to put the R10 six to eight feet behind the ball. FlightScope asks for roughly eight feet of ball flight before the net for Mevo and Mevo+. Add those together and a compliant indoor radar setup wants about five metres of clear depth, which is the entire internal length of an older UK single garage (2.4m by 4.8m is the classic spec) before you have stood in it.

Now do the timing. A 7-iron struck at 110 mph ball speed is travelling 49 metres per second. Eight feet is 2.44 metres. The radar therefore has about 50 milliseconds of ball to look at. With a driver at 138 mph the window shrinks to roughly 40 milliseconds. Everything the unit tells you about carry, peak height, descent angle and landing dispersion is extrapolated from that fragment, run through a ball-flight model, and printed to the nearest yard with no error bars.

Spin is worse. A driver at 2,800 rpm spins 46.7 times per second, so in that 40ms window the ball completes fewer than two full revolutions. Estimating spin rate and spin axis from under two rotations is genuinely hard, and it is why indoor radar spin drifts. A 7-iron at 6,500 rpm gets you about five and a half revolutions, which is better, and correspondingly 7-iron spin indoors is more trustworthy than driver spin indoors.

Here is what the drift costs you in yards. Take a mid-handicap 7-iron: 110 mph ball speed, 18° launch, 6,500 rpm, which carries about 150 yards.

Ball speed  Launch   Spin      Carry    Peak height
110 mph     18.0°    6,500     150 yd   26 yd
110 mph     18.0°    7,200     148 yd   28 yd     <- +700 rpm read error
110 mph     20.0°    6,500     153 yd   29 yd     <- +2° read error
110 mph     18.0°    5,800     151 yd   24 yd     <- -700 rpm read error

Two yards of carry for 700 rpm. That is the good news and the bad news together: indoor radar spin error barely moves iron carry numbers, so your gapping is roughly safe, but the spin figure itself is too noisy to coach from. If you are changing your wrist action to knock 400 rpm off your 7-iron based on an R10 reading in a garage, you are responding to noise.

Cameras don’t care about depth. They care enormously about light

Photometric units solve the depth problem outright. SkyTrak+ sits about a foot to the side of the ball and captures its images within roughly the first 12 to 18 inches of flight. Uneekor’s EYE XO2 hangs from the ceiling and does the same. Foresight’s GC3 and Bushnell Launch Pro measure over inches, not feet. Spin comes from comparing two stereo frames of a marked ball microseconds apart, so a full revolution is never required.

What they need instead is clean, controlled illumination and a clean read on the ball. The better units bring their own infrared flash, which is why a GCQuad works in a dark bay, and it is also why direct sunlight through a side window or an open garage door wrecks them: your daylight floods the sensor the flash was calibrated for. Uneekor wants marked balls for spin. Rapsodo’s MLM2PRO is a hybrid that leans on your phone’s camera for its Impact Vision clip and for shot detection, so a 60-lux garage gives you a video you cannot read and occasional missed shots. Garmin’s R50, with its three cameras, states plainly that it needs adequate lighting.

Adequate is a number, and most garages miss it badly:

Your phone footage lives or dies on the same table. A clubhead at 100 mph covers 4.47 cm per millisecond. Shoot iPhone slow-mo at 240fps and your maximum exposure per frame is 1/240s, or 4.17ms, during which the head smears 18.6 cm across the frame. Drop to 1/1000s and the smear is 4.5 cm. Get to 1/2000s and it is 2.2 cm, which is sharp enough for an AI model to find a shaft angle. But 1/2000s at f/1.8 in an 80-lux garage means ISO in the five figures and a frame made of noise. That is the whole reason your range footage looks crisp and your garage footage looks like a ghost story.

Which AI tools actually read garage footage

Sportsbox is the one people get wrong most often. It infers three-dimensional pelvis and thorax positions from a single 2D camera, and that inference degrades quietly rather than erroring out. Film with the up-and-over door open behind you and you get a silhouette: the model still returns numbers, still draws its skeleton, still tells you your pelvis has 4 cm of sway. Close the door, put 500 lux on yourself from the front and side, wear a top that contrasts with the wall, and the same swing produces different numbers. If two sessions disagree, check your lighting before you check your technique.

What to practise instead

Trust the two measurements every unit actually measures rather than models: ball speed and clubhead speed. Their ratio is smash factor, and smash factor is the single most honest number available to you indoors. A 6 to 24 handicap golfer should be chasing 1.33 or better with a 7-iron and 1.45 or better with driver. If your 7-iron smash sits at 1.26, you have a strike problem worth more shots than any spin figure on the screen. Our guide to launch monitor numbers and what they mean unpacks which of those readings hold up and which are downstream of a model.

Build gapping ladders in ball speed, not carry. Write down the ball speed for each club, fifteen shots, take the median and the spread. Indoors that ladder is measured; the carry column next to it is calculated. When you get outdoors, one range session with the same unit tells you what your measured ball speeds are worth in real yards, and you calibrate the model to yourself instead of the other way around.

Strike location is the other thing a garage does well. A £4 can of dry foot spray on the face gives you an impact mark that no lighting condition can corrupt, and toe strikes explain most of the spin weirdness people blame on their device. Face spray plus smash factor plus start line against a string across your net will find more shots in six weeks than an indoor spin readout ever will.

Ceiling height decides your club selection more than any of this. A 2.3m garage ceiling is 7ft 6in, and most golfers over 5ft 8in cannot swing a driver in that without decapitating a light fitting or shortening the swing into something that isn’t their swing. Work 8-iron and down indoors, and treat driver as an outdoor club that you rehearse indoors with a slow-motion drill.

Go and measure three things this week before you hit another ball into that net: the distance from your ball position to the net in feet, the ceiling height above your ball at its lowest point, and the lux at ball level with a free light meter app. Those three numbers tell you which half of your launch monitor’s output to believe.