Chart.js Decimation Plugin: Choosing LTTB vs Min‑Max Algorithm

Algorithm Selection

The choice between lttb and min-max depends entirely on whether your priority is the visual "story" of the data or the absolute preservation of extremes.

• Use LTTB (Largest-Triangle-Three-Buckets) for trend analysis. It is designed to preserve the visual shape and slope of a line, making it ideal for datasets where the overall pattern is more important than individual outliers.
• Use Min-Max for spike detection. It ensures that the highest and lowest values in every bucket are retained, preventing narrow peaks or deep valleys from being smoothed away.

Implementation Strategy

To standardize the decimation plugin, follow these requirements and configuration patterns:

1. Mandatory Configuration

Decimation will not function correctly unless the following constraints are met:

• parsing: false must be set in the dataset options.
• The x values must be sorted in ascending order.
• The x axis must use a linear or time scale.

2. Dynamic Sampling

Avoid static constants for samples. A static value may result in over-sampling on small mobile screens or under-sampling on 4K monitors. Instead, derive the sample count dynamically based on the rendered chart area:

// Conceptual logic for dynamic sampling
const chartWidth = chartCanvas.clientWidth;
const devicePixelRatio = window.devicePixelRatio || 1;
const dynamicSamples = Math.min(chartWidth * devicePixelRatio, MAX_PERFORMANCE_CAP);

Interaction and Fidelity

It is important to note that the decimation plugin modifies the data passed to the renderer. When points are dropped:

• Tooltip Accuracy: Tooltips and hover events only trigger for the points that remain after decimation. Dropped points are not rendered and cannot be interacted with.
• Visual Fidelity: LTTB may smooth out narrow spikes, while Min-Max may create a "jagged" appearance by overemphasizing noise.

Verification Steps

1. Render the same dataset using both algorithms at multiple canvas widths.
2. Compare the decimated output against the raw source to ensure critical peaks (for Min-Max) or trends (for LTTB) are preserved.
3. Test hover interactions to confirm that the loss of tooltips for dropped points is acceptable for your UX.

Missing Diagnostic: To provide a final recommendation, please confirm if your primary requirement is trend shape preservation or extremum/spike detection.