ShotIQ Learn/📡 Launch Monitor Metrics
Is Your Improvement Real?
Your average moved. The question nobody can answer from the number alone is whether it moved because you changed, or because ten golf shots never produce the same average twice.
Put your own numbers in and find out.
Sample size calculator
Put in your own numbers. It uses the same test ShotIQ runs on your sessions.
Your 6.0-yard change sits inside your 6.1-yard noise band. On these shot counts it cannot be told apart from ordinary shot-to-shot variation.
What it's doing
Two averages built from a handful of shots will differ even when nothing about the swing changed — that's what a spread of outcomes does. So before a difference can be called real, it has to be bigger than the amount your average was going to move on its own.
That threshold is the noise band, and it depends on three things: how wide your shot-to-shot spread is, how many shots you hit this session, and how much history you're comparing against. The calculator runs Welch's t-test at 95% confidence — the same test ShotIQ runs on every club, unequal sample sizes and unequal spreads allowed for.
A result inside the band isn't "probably nothing". It's indistinguishable from nothing on the evidence you have — which is a different and more useful statement. It doesn't mean you didn't improve. It means this session can't tell you either way.
The three levers, and which one to pull
Your spread is the one that matters most. Halving your spread does more for your ability to detect change than doubling your shot count. It's also the thing worth improving for its own sake — a tighter spread is a better scoring club regardless of what the average says. Try it: set the spread to 5 instead of 8 and watch the band collapse.
Shot count has fast diminishing returns. Going from 5 shots to 15 buys a lot. Going from 30 to 50 buys very little. There's a practical ceiling on what one session can prove, and it arrives sooner than most golfers expect.
History is the cheap lever. A session measured against 40 shots of history has a wider band than the same session against 150. You don't have to hit more balls today — you have to have hit them at some point, and kept them. Set the history to 150 and watch the band tighten without touching the session at all.
What to do with the answer
If your change is inside the band, nothing is wrong and nothing is proven. Keep the change if it feels better, log another session, and let the evidence accumulate. Small real gains show up over several sessions; they are almost never visible in the one where you made the change.
If your change is outside the band, it's real — but "real" only means it happened, not that you know why. A carry drop with unchanged club speed is a strike problem; the same drop with lower club speed is effort or fatigue. See smash factor and club head speed for how to tell those apart.
Either way, the number to watch alongside the average is the spread itself. It usually improves first, and if you only track carry you'll miss the earliest evidence that something is working.
Frequently asked questions
What is a noise band in golf launch monitor data? It's the smallest change in your average that can be distinguished from ordinary shot-to-shot variation, given how many shots you hit and how much they vary. Any difference smaller than the band is indistinguishable from what an identical swing would have produced anyway.
How do I find my carry spread? Most launch monitor apps report a standard deviation or a consistency figure per club. If yours doesn't, take a session's carries, and roughly two thirds of your shots will land within one spread either side of the average. As a starting point, a wedge is often ±3 to ±5 yards, a mid-iron ±6 to ±9 and a driver ±10 to ±14.
Why does a bigger history make the band tighter? Because a comparison is limited by both sides of it. If the thing you're measuring against is itself built from only a few shots, it's wobbling too, and the two errors add. Comparing ten shots to a long history is meaningfully more sensitive than comparing ten shots to another ten.
Does a result inside the noise band mean I didn't improve? No. It means this session can't tell you either way. Genuine small improvements are common and usually take several sessions to become visible — the honest response is to keep going and gather more data, not to abandon the change.
What confidence level does this use? 95%, the standard convention, via Welch's t-test — which allows the two samples to have different sizes and different spreads. A stricter threshold would call fewer changes real; a looser one would start calling noise a result.