Use line charts instead of horizon graphs when similarity should ignore amplitude and vertical offset changes

Use a line chart instead of a horizon graph when viewers must quickly judge which candidate time series is most similar to a query and differences in amplitude or y-offset should not break the match. For example, show the query and candidates as stacked line traces rather than banded horizon graphs when similarity is defined with z-normalization or another amplitude- and offset-invariant comparison.

Horizon-graph band boundaries turn small vertical differences into visible band and color changes, which slows matching when viewers are supposed to treat those differences as unimportant.

Mechanism: A line chart preserves the continuous shape without adding band thresholds, so viewers can compare overall form more quickly when amplitude or vertical shift is not the deciding factor.

Evidence: In the normalization experiment, line charts were faster than horizon graphs for choosing the most similar time series to a query, with mean completion times of 21.1 seconds for line charts versus 28.8 seconds for horizon graphs, and the collated result ranks the line-chart condition ahead of the horizon-graph condition with a significant pairwise difference (Gogolou et al., 2019; Zeng & Battle, 2023).

Notes: The source also found horizon graphs slower than the other tested designs in this condition.

• User Goal: Choose the most similar candidate time series to a shown query.
• Task: Visual similarity comparison where amplitude and vertical-offset changes should be discounted.
• Data: Multiple time series shown over the same time window and common scale.
• Chart Setting: Stacked small-multiple views with equal or comparable vertical space per series.
• Audience: Viewers making subjective similarity choices rather than exact numerical readouts.
• Success Criterion: Faster selection of the closest matching pattern.

• Break it when: Local temporal stretching or shifting is the main deformation you need viewers to tolerate. Why: The source did not show a significant speed advantage of line charts over horizon graphs in that warping condition.
• Break it when: Vertical compactness is the dominant design constraint. Why: Horizon graphs were introduced as a more space-efficient encoding, so replacing them with lines gives up that compression.

Sacrifice: You give up the vertical space savings of the horizon graph. Risk: A line chart uses more display height for the same time series. Mitigation: Use the line chart when faster similarity judgment matters more than dense vertical packing.

Mistake: Keeping a horizon graph when viewers should ignore amplitude and y-offset differences. Why it fails: Band thresholds can exaggerate exactly those differences and slow the match.

Failure Sign: Viewers hesitate longer with horizon graphs when matching amplitude- or offset-varying candidates to a query. Quick Check: Show the same query-and-candidates set once as line charts and once as horizon graphs, then compare which version yields a faster confident choice. Stronger Test: Time a small set of representative similarity judgments under your amplitude-invariant criterion and compare median completion times across the two chart families.

• Replace each banded horizon graph with a line trace for the same query and candidate series.
• Keep the same time window and common scale when you compare the two designs.
• Re-test the revised view on representative similarity prompts where amplitude or y-offset should not decide the answer.
• If compactness matters more than speed in your layout, keep the horizon graph and accept the slower matching.