YouTubeBPMFinder.online
A flag is a question, not a deletion
The YouTube tap outlier review compares each inter-tap duration with the list median. The user chooses a percentage threshold. Values beyond that distance are flagged for inspection but remain visible and included in the copied ledger.
This tool does not decide that a tap is wrong. A long gap may represent a missed activation, pause, expressive timing, or pulse change. Context belongs to the observer.
The median-deviation rule
First sort a copy of the list to find the median. For each original value X and median M, deviation percentage equals |X − M| ÷ M × 100. A row is flagged only when that percentage is greater than the selected threshold.
Using “greater than” instead of “greater than or equal” makes the boundary deterministic. A value exactly 10% from the median is retained at a 10% threshold. The result shows original row order, signed millisecond difference, percentage, and status.
Worked missed-tap example
Consider intervals 500, 498, 502, 501, 999, 499, and 500 ms. The median is 500 ms. At a 20% threshold, the 999 ms interval differs by 499 ms, or 99.8%, and is flagged. The other values differ by at most 0.4%.
The 999 ms duration is close to two normal 500 ms intervals, which may suggest a missed tap. That is an inference to investigate, not a fact the page asserts. Listen to the session context or repeat the passage.
Worked near-boundary example
With median 500 ms and threshold 10%, a 550 ms interval has exactly 10% deviation and remains unflagged. A 551 ms interval differs by 10.2% and is flagged. The ledger exposes the calculation so users can see why adjacent values differ in status.
Changing the threshold changes flags but never raw values. A stricter 5% setting catches smaller differences; a wide 50% setting highlights only large gaps. There is no universally correct threshold across music and devices.
Review threshold sensitivity explicitly
For a median of 500 ms, a 5% threshold marks values below 475 or above 525, subject to the strict-greater rule. A 20% threshold marks values below 400 or above 600. A 50% threshold marks values below 250 or above 750. These boundaries make the control’s effect visible.
Suppose one interval is 620 ms. It is flagged at 20% because its deviation is 24%, but retained at 25% because 24 is not greater than 25. The interval itself never changes. Copy output preserves the threshold and row percentage so a reviewer can reproduce status.
Do not move the threshold until only a preferred row is flagged. Choose a tolerance before inspecting desired outcomes, or record every threshold tried. Threshold shopping can make arbitrary cleanup appear evidence-based.
The raw order remains part of the evidence.
Decide what happens outside this tool
If a flagged interval is clearly caused by a double click or missed tap, return to the live card and document any manual exclusion in a new derived note. Do not rewrite the original history invisibly. If the interval reflects real tempo movement, keep it.
The outlier page intentionally does not recompute a cleaned BPM. Automatic recomputation would make a threshold choice look like a verified correction. Users can create a separately labeled analysis if they choose to exclude a value.
Difference from stability statistics
The stability checker reports quartiles, IQR, standard deviation, coefficient of variation, and BPM range across every valid value. This outlier route applies one simple median-distance rule row by row. It does not grade spread or summarize a whole session beyond counts.
Both tools begin with interval milliseconds, not media. They answer different questions: “How dispersed is this list?” versus “Which rows cross this declared threshold?”
Limits and privacy
Median distance can miss clusters where the median itself represents the wrong pulse. It can flag intentional syncopation or tempo change. It is not a robust musical detector, performance score, or substitute for the original observation note.
No URL, player, audio, video, microphone, API, account, or database enters the page. Intervals and flags remain in current memory and vanish after Reset, reload, or closure.
Frequently asked questions
Are flagged rows removed?
No. The ledger preserves every valid interval. The tool makes no automatic cleaned result.
Which threshold should I choose?
Choose a documented review tolerance appropriate to the session. There is no universal musical threshold.
Why compare with median?
Median is less moved by one extreme value than mean, providing a transparent central reference.
Can a valid musical interval be flagged?
Yes. Tempo changes, swing, and pulse choices can cross a numeric threshold. Review context before acting.
Paste the interval list, select and document a threshold, and inspect each flag in original order. Preserve the raw ledger even if you later create a separately labeled analysis.