Visualizing with SVG before D3: Timeseries

News headlines about the GameStop price swings this week reminded me of some old SVG visualizations of stock bubbles and crashes that I’ve done. SVG was around before D3. I generated SVG visualizations in the mid-2000’s well before D3.js. It was painful in comparison, but at the time it was a lot of fun.

The objective was experimentation: what could be done with a scalable vector graphics library? As such, it wasn’t constrained to screen resolution and screen dimensions (which at that time was typically 1600 x 1200) – rather you could do much more detail, import into Illustrator and print things at very high resolution with lots of detail, lots of transparency, and tiny text.

Analyzing timeseries

Here’s an example: Microsoft’s daily stock price from late 1992 to early 2004.

MSFT daily stock price 1992-2004 with different markers and shadings.

That’s 12 years of daily data with about 250 trading days per year resulting in a 3000 px wide visualization — which now draws just fine on my 4k display. With SVG, it is easy to layer in many different analyses creating different marks, lines and areas. Here’s a closeup of Lucent’s daily stock price during the Internet bubble:

Lucent stock price closeup of 1997-2001.

Many different graphical marks indicate data and derived indicators:

  • The blue line indicates the daily price.
  • The small green/yellow/orange/red bars behind the blue price line indicate the monthly price move from the beginning to the end of the month, colored by whether the price increases (green) or decreases (red). It aligns neatly with the monthly grid, filling in stripes within the grid.
  • The fat green/yellow/orange line behind that indicates yearly change in price. (There’s also light grey boxes behind that align with the thicker yearly grid.)
  • The green/purple circles indicating successive high/low points. Values for the highs and lows are indicated in text as well as the date. All successive low points are connected by a straight purple lines, all successive high points are connected by green straight lines. The zone between the green/purple lines form an envelope around the price range.
  • The many orange lines are moving averages each with a different time period, forming a guillochĂ©. When moving averages start to cross it is indicator of a change in trend, e.g. from up-trend to down-trend (as many stock traders know). In 1997, the averages start to converge but then the trend continues. However, in 2000, the moving averages successively start to cross, and by June most of the moving averages have crossed, before the much steeper crash.
  • The large arcs and corresponding fills indicate major trends. Up-trends start at trend low to the ending trend high (e.g. 3/19/1997 at $10.75 to 12/8/1999 at $71.31: up almost 7x!) Down-trends start at trend high and end at trend low (e.g. June 5, 2000 at $56.23 to a point outside the closeup: 9/27/2002 at $0.77: down to almost 1% from the start, ouch!) This is what a bubble looks like and how it plays out.
  • The filled shadows in pink and blue indicate the 52-week low price and 52-week high price. On the first two-thirds of this chart, the pink shadow is predominant as the stock keeps going up with a few “lakes” of blue filled in the occasional dips. In the last third of the chart, it’s almost entirely under the blue shadow as the stock tanks.

Why is it important to have all these different markers, bars, lines, arcs, text, and so on?

There are many techniques to analyze timeseries data – moving averages, standard deviations, envelopes, and so on. In finance, one can explicitly study timeseries analysis. And more broadly, these analyses apply beyond stock prices to electrical grid loads, network utilization, software performance, automobile diagnostics, and so on.

Long timeseries

SVG is scalable, so I applied it to longer and longer timeseries. Here’s the Dow Jones Industrial Average(R), when I re-ran the code with in 2010 to compare the 2008 financial crisis to the 1929 crash:

Dow Jones Industrial Average 1896-2010 daily: 28500 days, every point plotted.

So, 114 years of daily data results in a plot more than 28000 pixels wide. That’s more than my 4K screen can display in it’s entirety at full resolution. But paper can. The visualization uses a log scale: you can see the magnitude of the 2008 crash at the top right is far smaller compared to the crash of 1929. I.e. from the index peak in 2007 around 14000 to a low of 6500 in 2009, the index lost a bit more than half of it’s value; whereas in 1929 the index peaked at 381 in September 1929 then dropped to 44 in 1932, down 88% of its original value! There was a lot of pain in 2007-8, but the massive intervention in the markets by the Federal Reserve and central banks helped stave off a much bigger crash and avoid a much bigger, longer recession.

There’s a few other things going on with this experimental visualization. Here’s a closeup of the 1950s:

Closeup of Dow Jones Industrial Average during the 1950’s

In addition to the the many different shadings and markers, the grid lines also participate in indicating data range: rather than extend across the display, the grid lines are localized to the line +/- a range. Axis labels also follow data-driven rules. The price labels in this snapshot follow the grid (note the stock market high of 381 in 1929 wasn’t surpassed until 1954, some 25 years later). The date labels are completely data driven – indicating dates on the top side of the line when the price hits new highs, and on the bottom side of the line when the price hits new lows.

Why is it important to have such long timeseries in so much detail?

Our firm has clients with financial timeseries that are more than 200 years long. Seeing all the detail is important. The current GameStop bubble is not unique, there have been many, many more, going back to railroad mania in the 1840’s or the South Sea bubble in 1720’s. Different bubbles will play out in different ways: having detail allows for comparison to prior bubbles for insight into the current bubble. Recoveries from recessions will be different for different sectors. Some market experts will use this information to inform their portfolio strategies in response to GameStop, or in response to Covid, or in response to an election cycle.

And why the variations in grids, arcs and areas?

D3 is great and much can be done out-of-box. But, when you just use the standard examples applied to different data, you might not be indicating the things that matter to the end user (i.e. the purpose is insight, not pictures). As such, it’s important to use the toolset to experiment with the underlying graphics to highlight the core insights.

I’ve done much more experimentation with SVG before D3, but those will be for some future blog posts.

About richardbrath

Richard is a long time visualization designer and researcher. Professionally, I am one of the partners of Uncharted Software Inc. I have recently completed a PhD in data visualization at LSBU. The opinions on this blog are related to my personal interests in data visualization, particularly around research interests related to my PhD work- this blog is about exploratory aspects of data visualization not proven principles.
This entry was posted in Data Visualization, Line Chart, Timeseries. Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s