Compare Tap Cards Without Hiding Their Methods

Reviewed concrete example

Compare complete cards, not detached BPM numbers

The YouTube BPM session comparator accepts two, three, or four completed tap cards. Each card carries a selected BPM plus the recipe, interval count, wall-time coverage, time basis, and playback speed that produced it. Those fields matter because two identical numbers can describe different evidence, while two different observed numbers can represent the same source pulse at different speeds.

The page does not rank sessions or compress them into an average. Instead, it creates every unordered pair: one row for two cards, three rows for three cards, or six rows for four cards. A pairwise view preserves which methods actually differ.

Establish a comparison rate for each pair

Direct comparison is allowed when both cards already declare Corrected source time. It is also allowed when both declare Observed wall time at the same external speed. In either case, the selected BPM values share a clock basis.

If cards use different observed speeds or mix observed and corrected time, supply a source-normalized BPM copied from each card’s correction record. The comparator does not perform that transformation. It requires the completed result and labels the pair Supplied source basis. Without normalized values for every affected card, the pair is blocked rather than silently comparing unlike clocks.

This rule prevents a familiar false disagreement. An observed card at 90 BPM and 0.75× and another at 150 BPM and 1.25× differ by 60 observed BPM. If both carry a supplied source-normalized value of 120, their source-basis signed delta is zero and their ratio is 1. The speed column still reads Mismatch, preserving why the observed values differed.

Read the pairwise arithmetic

For cards A and B, signed delta is B minus A. Absolute delta removes direction. Ratio is B divided by A. A result of 1.01 means B is one percent higher than A on the declared comparison basis; it does not say why.

Suppose two source-time cards report 120.0 and 121.2 BPM. The signed delta is +1.2 BPM, the absolute delta is 1.2, and the B-to-A ratio is 1.01. Reversing card order would reverse the signed delta, so the ledger preserves entry order and labels both sides.

There is no mean, median, minimum, maximum, range, winner, or “best session.” With only two to four manual observations, an aggregate can look more authoritative than the evidence supports. This tool is designed to reveal comparability decisions, not summarize a dataset.

The method matrix explains what the BPM row cannot

Each pair receives exact-field checks for summary recipe, time basis, and playback speed. It also shows the absolute interval-count difference and the ratio of larger wall coverage to smaller wall coverage. These are prompts for review, not pass/fail quality scores.

For example, Card A may report Overall elapsed at 120.0 source BPM from 15 intervals over 7.50 wall seconds. Card B may report Median interval at 121.2 source BPM from 24 intervals over 12.00 seconds. Their delta is 1.2 and ratio 1.01, but the matrix also shows Recipe mismatch, a count difference of 9, and coverage ratio 12 ÷ 7.5 = 1.6. The pair cannot establish a tempo change because method and evidence depth changed together.

If both cards use Overall elapsed, source time, 24 intervals, and twelve seconds of coverage, the matrix is more aligned. Even then, agreement only describes the entered cards. It does not prove the intended pulse, eliminate shared tapping bias, or verify the external media.

Treat recipe differences as evidence

Overall elapsed uses all included intervals and total elapsed time. Median interval converts the middle interval duration. Trimmed mean excludes a declared portion of interval extremes. Recent window uses a trailing subset. Comparing these recipes may be useful when investigating sensitivity, but it is not a clean repeatability comparison.

The matrix therefore says Match only when recipe labels are exactly the same. It does not pretend that close BPM values make methods equivalent. Reopen the original cards to check trim count, recent-window length, warnings, and pulse labels; those details remain outside this compact comparator.

Coverage and count add context without grading

A longer card may average over more human variation or cross a changing passage. A shorter card may isolate a stable phrase but react more strongly to one delayed tap. Interval-count difference and coverage ratio expose that asymmetry without deciding which card is better.

Coverage ratio is always larger coverage divided by smaller coverage, so it is at least one. Two twelve-second cards produce 1.00. Twelve seconds versus eight produces 1.50. The tool does not normalize BPM by coverage; it simply reports the evidence scale beside the pair.

Preserve chronology outside the matrix

Cards remain in entered order, but pairwise comparisons do not detect acceleration, drift, or a mix transition. If chronology matters, use card labels such as Early section and Late section and preserve the source observation notes. A positive B-minus-A delta is not proof that tempo increased; different pulses, sections, devices, recipes, or tapping behavior can produce it.

Raw-interval questions belong elsewhere. The stability checker describes one interval distribution, and the outlier reviewer applies a declared median-deviation rule. This comparator never receives or removes individual intervals.

Local, manual, and independent scope

No YouTube URL, player, audio, video, microphone, API, upload, account, database, catalog, or saved card library is used. The page cannot identify whether cards came from the same recording. Entered card fields and generated pair rows remain in current-page memory until Reset, reload, or closure.

Copying the matrix creates a plain-text clipboard record controlled by the browser and device. The copy includes card order, comparison basis, deltas, ratios, and method fields. It does not invent dates, media names, or session conditions.

Frequently asked questions

Enter complete cards, inspect each pair’s arithmetic and method matrix, then return to the original evidence wherever recipe, time basis, speed, count, or coverage differs.