PitchKit
API Reference@pitchkit/coreFunctions

computeKdeGrid()

function computeKdeGrid<T>(layer, dimensions): KdeGrid;

Defined in: packages/core/src/kde/density.ts:65

Estimates a smooth density surface from point data via a 2D Gaussian kernel density estimate, sampled on a resolution x resolution grid over the pitch — mplsoccer's kdeplot. Unlike computeHeatmapBins, which is a hard histogram, each point spreads influence over its neighbourhood, so the result reads as a continuous surface rather than a grid of cells.

Pure and framework-agnostic (no pixels, no colour) like every other binning function here; render/canvas/paint-kde.ts turns the grid into pixels.

The kernel is axis-aligned with a per-axis bandwidth, defaulting to Silverman's rule (see silvermanBandwidth) so a caller gets a sensible surface without tuning; bandwidth overrides both axes when the data's own spread isn't the right smoothing scale (e.g. a handful of points). Values are densities, not probabilities — they're only ever consumed relative to maxValue, so the normalising constant is omitted.

Type Parameters

Type Parameter
T

Parameters

ParameterType
layerKdeLayer<T>
dimensionsPitchDimensions

Returns

KdeGrid

On this page