The VIX is the price the options market puts on the next thirty days of turbulence. Plot every one of its 9,261 daily closes since 1990 as a line and you get spikes. Plot each year as a distribution and you get something a line cannot show: that most years look alike, and the memorable ones are shaped differently rather than merely taller.
A line chart of the VIX answers "when was it high?". It is a perfectly good question and a well-worn answer: 2008, 2020, a few others. What it cannot answer is "what did a normal day feel like that year?" — because a line spends its ink on the extremes and says nothing about where the index actually sat for the other 240 trading days.
A ridgeline answers the second question. Each year becomes a kernel density of its own daily closes on one shared axis: how often the index sat at each level. A calm year is a narrow spike hugging 12. A frightened one drags a long tail to the right. Thirty-seven of them stacked lets you compare the texture of one year against another, not just their worst afternoons.
2017 is the flattest, tightest ridge on the chart. Its median was 10.8, the lowest of any year here, and it never once closed above 30 — not a single day. 2009 is its opposite: a broad smear with a median of 28.6 and 44% of its days at or above 30, the largest share of any year. The single highest close in the whole record came on 16 March 2020 at 82.7.
The dashed line at 30 is the spine of the chart. Some years never reach it. One year lived above it. That contrast is the argument, and it is drawn rather than asserted.
Bandwidth per year, not one global number. A kernel density needs a smoothing width, and a single value for all 37 years was wrong in both directions at once: wide enough for 2020's fat tail flattened 2017's spike into a hump, and narrow enough for 2017 made 2020 lumpy and false. Each year gets Silverman's rule computed on its own distribution.
Shared height, stated out loud. Every ridge is scaled so its tallest point is drawn the same. That makes the shapes comparable, and it means a ridge's height says how concentrated a year was, not how many days it had. The notes on the chart say so, because a reader is entitled to assume the opposite.
The incomplete year is marked. 2026 is 169 trading days in. It is drawn dimmed and asterisked, rather than presented alongside 37 finished years as though it were one of them.
The VIX measures expected volatility, not direction. A high reading is not a forecast that prices will fall — it is the market pricing a wider range of outcomes, in either direction. Every label here is a date and a number; none of them names a cause. The years anyone can name are visible in the shapes without the chart telling you what to think about them.
One row per observation, a date and a value. The agent buckets by period and computes the densities.
| date | value |
|---|---|
| 2020-03-16 | 82.69 |
| 2020-03-17 | 75.91 |
| 2017-11-03 | 9.14 |
Novice tip: a ridgeline wants many observations per band — roughly 50 or more. Twelve monthly points per year will not make a density worth reading; 250 daily ones will.
Distributions carry information that summary statistics throw away. "Average VIX in 2013: 14.2" and "average VIX in 2015: 16.7" sound like neighbours. Their ridges do not look like neighbours at all — one is a tight spike, the other has a shoulder that reaches past 40. If your data has a shape, show the shape.
Have a dataset with a story in it? Drop it into the AK.VIZ agent for three chart proposals, see more of the work, or read about bespoke commissions.