How accurate are golf simulators? The evidence, tier by tier

Spending four figures on a launch monitor means trusting numbers you have no way to check. Every page answering how accurate are golf simulators quotes a percentage with nothing behind it, and the metric being quoted at you is usually the one that holds up worst indoors. Here is what published lab validation, independent testing and manufacturer documentation actually show, metric by metric.

The short answer, metric by metric

Accuracy is not one number, and asking how accurate a given simulator is will always produce a useless answer. Ask instead which metric you need. Ball speed and club head speed are the sturdiest readings at every price tier. Carry sits close behind, because it is computed from those readings. Spin rate is the number that falls apart, and indoors on a radar unit it falls apart badly.

That ordering matters before you buy rather than after. Accuracy is not a detail to settle once the room and the screen are done; it is bound up in the tracking choice, which sits near the front of the order these decisions come in.

Measured data and modeled data are not the same thing

A launch monitor observes a small number of physical quantities in the few milliseconds around impact. Everything else on the screen is arithmetic. Carry distance is almost never observed by a home unit. It is computed from ball speed, launch angle, spin rate and spin axis, run through an atmospheric model, and drawn as an arc that looks exactly as convincing whether the inputs were right or wrong.

The consequence is worth stating plainly, because no marketing page will state it for you. An error in an observed input propagates into every modeled output downstream. A spin reading 1,000 rpm off does not hand you a slightly wrong spin number. It hands you a wrong apex, a wrong carry and a wrong landing position. Which inputs get observed and which get inferred is decided by the sensor, which is why camera and radar tracking diverge most on exactly the numbers golfers argue about.

What a camera unit observes

Foresight is specific about what its hardware reads rather than what its software produces, which makes the list useful. Among the parameters Foresight measures on the ball side are ball speed, launch angle, azimuth, total spin and spin tilt axis, and on the club side club speed, impact point, angle of attack, club path, impact loft, delivered lie angle and closure rate. A camera unit reads the ball surface and the face directly, so most of that list is observation.

The caveat is that a list is only a list. Foresight publishes no tolerance figure alongside it: no degrees, no rpm, no mph, no yards. Neither does any other maker in this category that we could find, which is why the accuracy case here is built from academic and independent testing rather than spec sheets. Any claim about a camera unit being good to some precise fraction of a degree did not come from the manufacturer.

What a radar unit observes indoors

Radar does not read the ball at impact. It reads a segment of flight and works backwards to impact conditions. Outdoors that segment is as long as the shot. Indoors it is whatever fits between tee and screen, which is why placement is specified rather than suggested.

FlightScope's published indoor guidance puts the sensor 7 to 9 feet behind the tee and sets a floor of 8 feet of ball flight in limited flight mode. The shorter the observed arc, the larger the share of what you see that is model rather than measurement.

Spin is the number that breaks first

The strongest evidence available on this question is peer reviewed, and it is blunt. In a validation of the FlightScope Mevo+ published in the Journal of Strength and Conditioning Research, Brennan and colleagues had 29 youth golfers, each with at least three years of playing experience, hit 10 shots with a driver and 10 with a 6-iron indoors, with a TrackMan 4 as the reference device.

Correlations across the measured set ran from 0.254 to 0.985, with club head speed and ball speed at the top of that range at 0.92 or better. Spin rate showed the poorest agreement of any metric, a mean bias of up to 1,238 rpm with 95 percent limits running from minus 2,628 to 5,103 rpm. Reliability told the same story: the Mevo+ returned coefficients of variation from 1.46 percent at the tight end, on the speed and distance metrics, out to 72.70 percent at the loose end. Spin rate differed from the reference by a large effect size on both clubs, d of 1.27 with the driver and 0.90 with the 6-iron.

Read that as a practical statement. On a single shot, ball speed can be inside a couple of percent while the spin figure printed beside it is nowhere near the truth. The authors concluded that the unit is valid and reliable for speed, distance and smash factor, and that practitioners should be cautious with spin rate, launch angle, attack angle and spin axis. That is an independent finding about a $1,099 radar unit, and its limits are worth stating: one radar model, one indoor session, 29 youth players, and a reference device that is an instrument rather than ground truth. What it does not depend on is the brand, because the short indoor observation window that produces the spin result is a property of the room.

The sticker and the RCT ball

The workaround is itself an admission. FlightScope says in its own support answers that aluminum stickers are needed in an indoor environment for the X3C, Mevo+, Mevo Gen2 and Mevo to accurately measure ball spin. Titleist built a ball around the same problem: RCT, or Radar Capture Technology, puts a radar reflective mark on the casing layer so that radar monitors can measure rather than estimate spin on shots hit indoors, and Titleist reports that precise indoor spin capture rises to over 99 percent with it.

The sentence that settles the argument is the next one on that page: RCT is not required outdoors. The indoor spin problem is a property of the room, not a defect in the hardware. In practice a sticker has to face the sensor on every shot, which costs real time across a session.

What independent testing shows across price tiers

Lab validation covers one unit at a time. For the shape of the market the most useful public work is a 12-unit test spanning $400 to $5,000, run indoors and outdoors against a single reference system. It reports tier trends rather than per-model error tables, so no figure from it belongs to a named device.

• Ball speed held within roughly 1 to 2 percent of the reference indoors and outdoors, including in the $400 to $700 group.
• Carry deviation reached four to five percent on several models, with outdoor numbers spreading more than indoor.
• Some lower-priced units deviated by 20 to 30 percent on spin rate indoors.
• Some models showed double-digit percentage variation in launch angle, while higher-priced systems clustered tighter.

The reading nobody else offers: price buys spin fidelity and club data, not ball speed. A $499.99 Garmin Approach R10 and a $6,999 Foresight GC3 are far closer on ball speed than the gap between those two prices suggests, and far apart on spin and impact location. The question worth asking is not which unit is accurate, but which units among the monitors we rank on data quality measure spin instead of inferring it. Both prices were read on 20 August 2026 and move often.

How to read an accuracy claim without being fooled

Almost every accuracy number here is published by someone with an interest in it, ours included. Three questions separate a finding from a slogan.

1. What was the reference device? A number measured against nothing is not a claim. The lab study above names a TrackMan 4. A page promising accuracy within two yards, without saying two yards of what, has told you nothing.
2. What was actually being tested? Some vendor comparisons feed the reference device's shot parameters into the challenger's ball flight algorithm and report the carry difference. That validates the flight model, not what the sensor saw.
3. Who ran it, and who paid for it? Testing commissioned or performed by a party that sells the unit is not disqualifying, but it is context, and it belongs in the sentence rather than a footnote.

A percentage with no metric, no reference and no method attached is decoration. That describes most of what circulates here, including the 90 to 95 percent accuracy figure that appears on page after page with no traceable origin.

The inaccuracy you cause yourself

Most reports of a wrong simulator are reports of a wrong install. Five causes account for the bulk of it, and four of them cost nothing to fix.

• Aim. A unit set one degree off the target line reports a push on a shot that was straight, every time, all night. Square it with a string line or a laser, not by eye.
• Space. A radar placement figure is an accuracy specification wearing a tape measure, and the flight window a sensor needs is one of the numbers that decides how much room a bay needs.
• Light. FlightScope specifies a minimum of 300 lux, held consistently within a 3-foot radius of the hitting area, for its face impact location data. It is the only published lighting threshold we found in this category, and it explains a class of complaint that gets blamed on the hardware.
• Ball. Range balls and surlyn-covered balls spin differently from urethane. Change ball model and the numbers change with no swing change at all.
• Drift. Gear gets nudged. Levelling and alignment move gradually and quietly, which is why almost nobody diagnoses them.

A ten-minute accuracy audit

Calibration gets named as a factor on every page about this subject and explained on almost none. This is the procedure we run on our own bay, and it takes about ten minutes.

1. Square the unit to the target line with a string line or a laser, then confirm it sits level. Alignment set by eye is the most common source of a push that never happened.
2. Clean the lens or sensor face, then check mounting height and distance against the manual rather than memory.
3. Hit twenty balls with one mid-iron and read the spread, not the average. A tight spread with an offset mean is a calibration problem you can correct. A wide spread is not.
4. Cross-check that same club against a carry number you trust from outdoor play, comparing carry to carry with the same ball.
5. Repeat the whole thing after any knock to the unit, any move, and any change of ball.

The principle underneath it: repeatability is testable at home and absolute accuracy is not. A unit that reads the same swing the same way every night is delivering what practice needs, whatever its absolute error turns out to be.

How much error actually matters

Percentages hide their own consequences, so do the arithmetic in the open. Five percent on a 150-yard 7-iron is 7.5 yards, most of a club and enough to break a gapping session. Two percent is 3 yards, which is nothing. The same error is fatal for one job and irrelevant for another. A 300 rpm swing in backspin changes how a wedge stops and almost nothing else in the bag, which is why what a spin number means in isolation misleads more often than it helps.

The biggest use case sits in the top row. Most owners are grooving a swing, and for that job absolute accuracy matters less than the worry about it suggests, while consistency matters more than the marketing admits.

Where accuracy genuinely runs out

Four places where the format has real limits, stated without hedging.

• Putting. Ball speeds of a few miles an hour are a hard read for hardware tuned to a driver, and green break is generated by software rather than measured. Auto-putt and gimme ranges are the standard workaround, and a separate putting mat does the real work.
• Chipping and partial wedges. The shortest shots give radar the least arc to work with, which is why limited flight modes exist and carry minimums of their own.
• Mishits. Heel and toe strikes are where measured impact location earns its price, and where inferred club data quietly flatters a swing.
• Scores. A simulator round reads lower than a course round for reasons unrelated to tracking: no wind that was not modeled, perfect lies, no rough, and a generous gimme radius. A carry number the hardware got exactly right still lands on a surface the software invented, and rollout, turf firmness and green speed are modeling choices that the two main course engines make differently.

A simulator score is not a course score, and it should not be posted as one.