Tower height vs model P(unstable)

10 random towers; P(unstable) averaged over noise σ ∈ {0, 2.5, 5, 10} mm × push Δv ∈ {0, 0.1, 0.2, 0.3} m/s, 25 runs each, threshold 2.5 cm. Hover a point or column for its values.

Idea 1 · Towers as markers

Each drawing stands on its data point (blue dot at its base). Towers that tie are nudged sideways, with a short line back to the point. Most direct, but crowding grows with the number of ties.

0 0.25 0.5 0.75 1 P(unstable) 0 10 20 30 40 50 60 Tower height (cm)

Idea 2 · Callouts

Plain dots keep the scatter exact; each drawing takes the nearest free spot around its dot and a hairline joins them. Never hides a point, but leaders can get long in dense regions.

0 0.25 0.5 0.75 1 P(unstable) 0 10 20 30 40 50 60 Tower height (cm)

Idea 3 · Row of drawings

Plain dots above; the drawings stand in a row under the axis, ordered by height, each tied to its dot. Drawings never collide and can be larger; following a leader takes a moment.

0 0.25 0.5 0.75 1 P(unstable) 0 10 20 30 40 50 60 Tower height (cm)

Idea 4 · Ranked columns

No height axis: towers are ordered by height and drawn front-on at one true scale, so height is read off the drawings (cm axis below), with P(unstable) as a lollipop above each. Loses even spacing in height.

0 0.25 0.5 0.75 1 P(unstable) 0 10 20 30 40 50 Height (cm) tower_005 tower_004 tower_008 tower_006 tower_002 tower_009 tower_001 tower_003 tower_000 tower_007

Idea 5 · Silhouette glyphs

Each point is a tiny front-on silhouette of its tower, so the marker itself hints at the shape while staying small enough to keep the scatter readable. Detail is limited to what fits in ~30 px.

0 0.25 0.5 0.75 1 P(unstable) 0 10 20 30 40 50 60 Tower height (cm)

Idea 6 · Numbered key

Dots are numbered 1–n in height order; a key beside the plot shows each numbered drawing. Nothing covers the data and there are no lines to follow, but matching numbers takes a glance back and forth.

0 0.25 0.5 0.75 1 P(unstable) 0 10 20 30 40 50 60 Tower height (cm) 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Idea 7 · Lineup table

One row per tower, sorted by height: drawing, height bar and P(unstable) bar. The relation shows as the P bars growing down the table. Scales to many towers; gives up the 2-D scatter.

Height (cm) P(unstable) 0 60 0 1 tower_005 15 0.11 tower_004 20 0.08 tower_008 30 0.56 tower_006 30 0.95 tower_002 45 0.85 tower_009 45 0.93 tower_001 45 0.99 tower_003 55 0.63 tower_000 55 0.97 tower_007 55 1.00

Idea 9 · Annotate the exceptions

A logistic trend fitted through all points, with drawings only for the three towers furthest from it, the ones that tell you something height alone does not. Others are dots (hover for details).

0 0.25 0.5 0.75 1 P(unstable) 0 10 20 30 40 50 60 Tower height (cm)
Table view
TowerHeight (cm)P(unstable)
tower_00515.00.107
tower_00420.00.080
tower_00830.00.555
tower_00630.00.953
tower_00245.00.853
tower_00945.00.932
tower_00145.00.995
tower_00355.00.632
tower_00055.00.970
tower_00755.01.000