A plotly.js parcoords trace auto-detected by MAIDR. The
columns here are not one scale — miles per gallon beside horsepower
beside weight — so the pitch is scaled per axis rather than for the layer.
Arrowing along one car hears it high here and low there; arrowing down a
column hears the cars ranked on that variable; and two cars whose pitches
swap between adjacent axes are the crossing lines that mean negative
correlation. Plotly draws these lines to a canvas rather than to SVG, so
the layer sonifies and reads out but has no visual highlight.