A measurement of piano playing

High-loud slope

Whether higher notes are played louder.

What it sounds like

The same passage — Schubert Op. 90/3, bars 1–2 — from the recordings that sit lowest and highest on this measurement. Every clip is a note list replayed on one sampled grand at one gain, so the only difference you are hearing is the playing.

Least

Walter Gieseking, 1955 — bars 1–2, high-loud slope 6.3 vel/oct8 s
Peter Rösel, 2007 — bars 1–2, high-loud slope 6.6 vel/oct9 s

Most

Khatia Buniatishvili, 2019 — bars 1–2, high-loud slope 16.2 vel/oct10 s
Kun-Woo Paik, 2013 — bars 1–2, high-loud slope 15.4 vel/oct10 s

How it is calculated

Slope of a regression of melody velocity on pitch, in velocity units per octave.

How much pianists differ on it

Every piece in the library, with the range across its recordings. The units mean the same thing in every row, except for basic tempo, which is in each piece’s own beat.

PieceRecordingsLowestMedianHighest
Bach Prelude in C27-4.80.54.9
Chopin Étude 10/1310.01.43.7
Nocturne Op. 9/232-6.04.09.9
Ravel Pavane33-0.14.110.1
Chopin Ballade 4124.05.07.3
Chopin Étude 10/12192.95.17.1
Ravel Ondine611.85.78.7
Schubert Op. 142/393.96.07.7
Mozart K. 545372.77.212.4
Rachmaninoff Prelude294.67.610.3
Fantaisie-Impromptu206.39.111.8
Brahms Op. 118/2282.89.613.8
Clair de lune276.89.712.1
Schubert Op. 90/3432.810.116.2
Gymnopédie 1236.910.713.8
Träumerei314.010.716.5
Pathétique II478.412.216.0
Nocturne Op. 48/1278.012.315.1

How reliably it can be measured

A Disklavier capture — the pianist’s own key sensors, no microphone — was rendered to audio, transcribed as if it were a recording, and re-measured. The error the transcription step introduces, as a multiple of how far pianists actually differ on this dial, is 1.62. noisy The transcription moves this dial about as far as pianists differ on it, so no claim should rest on where a single recording sits.