1. Home
2. How accurate is the scoring?

Measurement

How accurate is the scoring?

Updated 3 August 2026

Where the timestamp actually comes from

Every MIDI message from your kit arrives carrying a hardware timestamp, applied at the point it reaches your iPhone, iPad or Mac. iQuantize reads that timestamp rather than the moment its own callback happens to run, which can lag behind the strike for reasons that have nothing to do with your playing.

That stamp is trusted only within 50 ms either side of the app’s own receive time. A timestamp of zero — which some virtual MIDI sources send — or one that drifts further than that is discarded, and the app falls back to the moment it received the message locally.

One more filter sits underneath all of this: if the same drum fires again within 30 ms, it is treated as the pad double-firing, not a second note, and is absorbed without a dot. That is a known trigger quirk, not a rhythm mistake, and it should not cost you one.

Windows that get tighter as the chart gets busier

A hit within ±80 ms of the grid scores green; within ±180 ms it scores orange; anything wider is red. Those are the defaults, and they can be narrowed or widened between 5 ms and 300 ms for a specific chart or player.

On a dense passage the green window shrinks further, capped to a quarter of the gap between consecutive notes on the same drum. A sixteenth-note hi-hat pattern is not judged on the same 80 ms window as a half-note crash.

All of this — window sizes, velocity zones, everything dynamic scoring depends on — is fixed the instant a take starts. Change a setting mid-take and it applies to your next take; the one you are playing right now cannot be recoloured or rescored underneath you.

Doesn’t a ±80 ms window make anyone look good?

On its own, yes. We make that exact complaint about other apps one page over, so here is the answer to it. The green band is ±80 ms by default, which makes the target 160 ms wide, and inside it timing with the click cannot tell 10 ms from 70 ms. It returns full credit for both. That band decides the colour of the dot over each note on the chart, and it supplies 70% of the click score. The other 30% comes from how tightly you cluster — and so does a whole separate skill.

Two players, both inside the window. The first sits 15 ms behind the click on every note and never varies by more than about 5 ms: 99 for timing with the click, 97 for timing stability. The second averages dead on the beat but scatters 60 ms either side of it: 90 for timing with the click, 67 for stability. The player who never once landed on the grid outscores the player whose average is perfect, because one of them is in control and the other is not.

Timing stability is the number that resolves all the way down. It is your own scatter measured against 180 ms, so every millisecond of wobble you take out of your playing shows up, and spread moves it more than three times faster than it moves the click score. The offset is never hidden from you either: the per-bar map under your result labels the bar in plain figures — 8 notes late (+15 ms) — whatever the click axis decided.

And the window is a setting. If you think ±80 ms is generous, take it down to 30 ms and play the same groove again before you decide how good your time is.

What the score cannot see

A MIDI stream tells you when a drum was hit and how hard. It does not tell you how you got there. Grip, posture, stick height, the angle of a fill — all invisible. Two players can land an identical timestamp with completely different technique, and the score cannot tell them apart.

It cannot judge taste either. Play a fill that was not the one written but arguably better, and the app still scores against the notes it was told to expect. Whether that substitution was a mistake or an improvement is a judgment only a teacher’s ear can make, which is why this is built as a second opinion between lessons rather than a replacement for one.

A skill it could not measure shows no number, never a zero

Every diagnosis needs enough evidence before it is shown at all. Internal pulse — whether your tempo holds once the click drops out — needs two complete bars played with no click after a valid count-in, plus the section that follows and a bar to recover in. Anything thinner comes back as not measured, and the drill that earns it is a separate piece of work.

• Dynamic range is not scored unless the chart itself asks for real contrast between soft and loud.
• Limb independence is not scored unless the chart demands hands and feet working against each other — a hands-only groove does not earn a fake 100.
• When note accuracy on a take falls below 55%, the timing, dynamics and limb diagnoses are all withheld. You cannot read someone’s pocket on a take where the notes did not land in the first place.

A few smaller cases work the same way: a disqualifying mode, too few matched hits, evidence too thin to interpret.

The measurement chain in numbers

Green window

±80 ms

Orange window

±180 ms

Adjustable range

5–300 ms, per chart or player

Dense-chart cap

¼ of the nearest same-drum note spacing

Timestamp trust window

50 ms either side of receive time

Echo debounce

30 ms same-drum retrigger absorbed

Extra-hit weight

0.5 — about half a missed note

Common questions