Thoroughbred racing is big business. And Philadelphia’s Parx Casino owns a racing track that, in a recent article in the Philadelphia Inquirer, has seen a number of horse deaths. The article includes a single graphic worth noting, a bar chart showing the thoroughbred death rate. The graphic contrasts rising deaths at Parx with a national trend of declining deaths.
Traditionally rate statistics are shown using dots or line. The idea is that a bar represents counting stats, i.e. how many total horses died. I understand the coloured bars present a more visually compelling graphic on the page, and so I can buy that reason if you are selling it.
Labelling each datapoint, however, with a grey text label above the bar remains unnecessary. They create sparkling, distracting grey baubles above the important blue bars. If you need the specificity to the hundredths degree, use a table. This graphic is also interactive. The mouseover state is where a specific number can be provided, adding an additional layer or level of depth in a progressive disclosure of information.
Last week wrapped up the Coast Guard’s two-week inquiry into the sinking of the submersible Titan, which imploded on a dive to the wreck of Titanic. The BBC summarised the findings in an article at the weekend. It included a number of fascinating annotated photographs identifying parts of the wreckage. But it also included the following graphic, which captures the text messages sent by the Titan and the depths at which the messages were sent.
This is significantly better than a number of pieces I have seen lately, to be fair, most of those focus on the dive depths of various objects and creatures. Mostly that is because the graphics—this one included—do not scale the objects to the depths. I understand the why; many would be too small to see. But I think that difference in scale really hits home just how deep Titanic rests on the seabed.
Because this graphic does not focus on the dive depths of objects, but rather the texts Titan sent at what depth, the scale issue is less relevant. Though, the weird bit is how Titanic sits below 3800 m. She rests at 3840 and that little dip on the sea floor looks closer to 400 m.
Overall, though, a solid piece.
Credit for the piece goes to the BBC’s graphics department.
Because who does not recall the great Sharpie forecast track by the National Hurricane Center (NHC)?
Earlier this summer, in the middle of the hurricane season, the National Oceanic and Atmospheric Administration’s (NOAA’s) NHC released a new, experimental warning cone map. For those unfamiliar, these are the maps that have a white and white-shaded forecast for where the centre of the storm will track. Importantly, it is not a forecast of where the storm will impact. If you have ever been through a hurricane—would not recommend—you know you need not be near the centre to feel the storm’s impact.
I have been waiting for a significant storm to threaten the United States before taking a look at these. (It is also important to note, these new maps apply only to the United States.) But this is the current map for Hurricane Helene as of Wednesday morning.
For those of you who, like me, are familiar with these, you will see the red lines along the coast that indicate hurricane warnings. Blue lines indicate current tropical storm warnings. Not on this map are pink lines for hurricane watches and yellow lines for tropical storm watches. But all these lines only represent watches and warnings along the coast. Little dots indicate the storm’s forecast position at certain times and through letter indicators its strength. The full white areas are the forecast track for the centre of the storm through the first three days. The shaded area is for days 4–5.
Contrast that with the new, experimental version.
The background of the map remains the same. In my perfect world, I would probably drop the grey and blue back a little bit, but that is not the end of the world. Instead, the biggest change is that the tropical storm and hurricane watches and warnings, which have always been declared for full counties inland, are now shown on the map.
You can see the red hurricane warnings are now forecast to move through the eastern Florida panhandle and southern Georgia with tropical storm watches forecast for the inland counties north and east of those. And then the three- and five-day forecasts have blended into a single white cone track. Subtly, the stroke or outline for that has changed from black to solid white. That helps reduce the distracting visuals on the map and emphasise the forecast track and watches and warnings.
Overall, I think is a really strong and important and potentially life-saving improvement to the graphics. Could things be improved more? Absolutely. But sometimes the only way to make improvements is through slow and steady incremental changes. This update does that in spades.
Credit for the piece goes to the NHC graphics team.
The Teamsters Union decided to officially endorse neither candidate in the 2024 US presidential election. Prior to their non-announcement announcement, however, the union surveyed its members and then released the polling data ahead of the announcement.
Of course, the teamsters represent but a single union in a large and diverse country. More importantly, the survey results reported only the share of responses for either candidate—and “Other”—so we have no idea how many of what number opted for whom. But hey, it’s another talking point in the final six weeks of the campaign.
Naturally, I decided to visualise the data.
The trend is pretty, pretty clear. The union’s rank-and-file clearly support Trump for president, with the exception of the teamsters in the District of Columbia. (Note, no survey was taken in Wyoming.) In fact, in only eight states plus DC did Harris’ support top 40%.
This past weekend saw some flooding along the East Coast due to the Moon pulling on Earth’s water. In Boston that meant downtown flooding, including Long Wharf. The Boston Globe’sarticle about the flooding dwelt with more impact, causes, and long-term forecasts—none of which really warranted data visualisation or information graphics. Nonetheless, the article included a long time series examining the change in Boston’s sea level relative to the mean.
For me, the graphic works really well. The data strips out the seasonal fluctuations and presents the reader with a clear view of rising sea levels in Boston. If the noisiness of the red line distracts the reader—one wonders if an annual average could have been used—the blue trend line makes it clear.
And that blue trend line has a nice graphic trick to help itself. Note the designer added a thin white stroke on the outside of the line, providing visual separation from the red line below.
My only real critique with the graphic is the baseline and the axis lines. The chart uses solid black lines for the axes, with grey lines running horizontally depicting the deviation from the mean sea level. But the black lines draw the attention of the eye and thus diminish the importance of the 0 inch line, which actually serves as the baseline of the chart.
If I quickly edit the screenshot in Photoshop, you can see how shifting the emphasis subtly changes the chart’s message.
Today I have a little post about something I noticed over the weekend: labelling line charts.
It begins with a BBC article I read about the ongoing return to office mandates some companies have been rolling out over the last few years. When I look for work these days, one important factor is the office work situation and so seeing an article about the tension in that issue, I had to read it.
The article includes this graphic of Office of National Statistics (ONS) data and BBC analysis.
Overall, the chart does a few things I like, most notably including the demarcation for the methodology change. The red–green here also works. Additionally the thesis expressed by the title, “Hybrid has overtaken WFH”, clearly evidences itself by the green line crossing the blue. (I would quibble and perhaps change the hybrid line to red as it is visually more impactful.)
I also like on the y-axis how we do not have a line connecting all the intervals. Such lines are often unnecessary and can often add visual clutter, see yesterday’s post for something similar. I quibble here with dropping the % symbol for the zero-line. Since the rest of the graphic uses it, I would have put the baseline as 0%. And that baseline is indeed represented by a darker, black line instead of the grey used for the other intervals.
Then we get to the labels on the right of the graphic. Firstly, I do not subscribe to the view charts and graphs need to label individual datapoints. If the designer created the chart correctly, the graph should be legible. Furthermore, charts show relationships, if one needs a specific value, I would opt for a table or a factette instead. These are not the most egregious labels, mind you, but here they label the datapoint, but not the line. Instead, for the line the reader needs to go back to the chart’s data definition and retrieve the information associated with the colour.
Now compare that to a chart representing Major League Baseball’s playoff odds from Fangraphs.
Here too we have mostly good things going on, but I want to highlight the labelling at the right. This chart also includes the precise value, which is fine, but here we also have the actual label for the lines. The user does not need to leave the experience of the chart to find the relevant information, although a secondary/redundant display or legend can be found at the bottom of the chart.
If you can take the time to label the end value, you may as well label the series.
Credit for the BBC graphic goes to the BBC’s graphics department.
Credit for the Fangraphs piece goes to Fangraphs’ design team.
As a wee lad I grew up south of Downingtown, Pennsylvania, an old mill town situated along the banks of the East Branch of the Brandywine Creek. Drop a little stick in the Brandywine and it would float downstream until it joins the Christina River in Wilmington, Delaware and thereafter shortly into the Delaware River.
Delaware has tax-free shopping and movie theatres I frequented in my youth. First laptop purchase for university? Delaware. Furniture for moving out to Chicago? Delaware. In other words, when I posted my most recent map of where I have been, the three counties of Delaware were some of the earliest counties filled in.
Delaware—for better or worse—is seared into my mind. If you look at the state border, you will see the northern border is circular. Look at all other state borders and that circle is kind of weird. Most other borders are straight(ish) lines, mountain ridges, rivers, or bays. The reason is the border between Pennsylvania and Delaware was, essentially, taking out a protractor and drawing a circle twelve miles distant from New Castle, Delaware, the original capital.
Anyway, I have not thought about that in quite some time. But thankfully, xkcd did.
As many of you know, I love geography and so I am aware of many of these places. Lake Manicouagan is one of those places that has an island in it, which has a lake on that island, in which there is another lake. There might even be another island/lake combination, but I could be mistaken.
I grew up less than 15 miles away from the Limerick Nuclear Generating Station, located on the banks of the Schuylkill River northwest of the city of Philadelphia. Our house sat on the north-facing slope of the Great Valley and the cooling towers of Limerick were a ridge line and river valley away from view. But on a clear day, you can see the puffy, billowy clouds of steam rising over the distant horizon—Limerick is splitting the atom.
We all know—or should by now—burning coal, oil, and gas are not terribly great for the planet. They emit carbon dioxide and other gasses that warm the Earth. But the white columns rising over the Schuylkill are water. Fissile uranium is more dense than coal, oil, or gas. And not just by a wee bit. But by orders of magnitude. Splitting the atom provides mankind with enormous amount of energy.
And we need energy. This summer was hot. And I don’t like it hot. Consequently, my air con ran almost nonstop. And I am not the only one. But whence comes all the electricity to power those units? Yes, we can get electricity from the sun, the wind, and the water. But what about when the clouds block the sun? Or the hot, sticky summer air refuses to stir? Or the parched earth has sucked the water from the reservoir?
The uranium atom can still be split, and at a reliable rate. That makes it great to provide a high amount of electricity that can be augmented by the sun, the wind, and the water when conditions permit.
However, in recent years, the cost of oil and gas declined thanks to fracking, and the business cost to run coal plants lowered as environmental standards disappeared. The economics of running nuclear power plants made them less viable than carbon-spewing options. Electricity providers started shutting nuclear plants down.
Things have changed, though. As we run more air con, we need more electricity. As we run more electric busses and trains, we need more electricity. As we charge more electric cars, we need more electricity. As we run more servers for bitcoin mining or AI farms, we need more electricity.
We need more electricity. A lot more.
And so the economics of electricity is changing. The Wall Street Journal had a great article about the re-opening of nuclear plant in Michigan. It included some really nice photographs of the control room and the turbine room. But, the reason we are talking about it here today because the article includes a few diagrams and illustrations. This one caught my attention.
First, I really enjoy how the United States is reduced to a grey outline. Perhaps a very faint grey could have been used to infill the states, but here I think white works best because of the use of the light and medium greys for active plants.
The active plants—not the focus of the article—are in those greys, whilst the decommissioned and -ing plants are in tints of red. What I struggled with a long time ago when I made an infographic about southeastern Pennsylvania’s electricity generation was how to show the different plants at a single facility.
Ultimately, I listed each plant by name then an icon representing the type of fuel, because not every plant uses all the same type of fuel. Eddystone Generating Station just south of Philadelphia used both natural gas/oil plants and two coal plants, though those were retired in the 2010s.
Here the designer, not needing to label each plant and aided by the fact each plant is nuclear, simply encloses the dots within a container. Palisades, the plant in question, receives a thicker, black stroke to call it out against the rest of the plants.
Credit for the piece goes to, I think Adrienne Tong. She is credited for a different graphic in the article, but not the one I highlighted, so I’ll give her the credit unless and until someone else gets the credit.
I thought three-dimensional charts died back in the 2010s. Alas, here we are in 2024 and I have to discuss one once again. have been following the Titan Inquiry this week and the opening presentation included this gem of data visualisation.
To be fair, I do not know how many designers, let alone specialist information designers, the US Coast Guard had or made available to create a clear and compelling chart and presentation, but…this is not it. First I will go through a number of points and then, when I had written about half of the post this morning, I decided it would simply be easier to put a white box over the main chart area and just recreate the graphic myself.
Unfortunately, after digging around, I could not find the actual dive depth data the Coast Guard used and so I essentially traced out the chart by hand. Not ideal, but for proof of concept as to how this chart could have been improved…I think my reinterpretation ssuffices.
To start, the chart sits on the slide with a drop shadow. Drop shadows are not all bad. They create perceived depth between an object and it’s background. The interwebs love them. I have used them. But I do not understand why here the chart needs a drop shadow to sit on the slide. Especially since the shadow pushes the chart “above” the deck, only for the three-dimensional bar chart to push the data “below” the chart’s surface, which means the chart data is being represented on the slide surface.
Deep breath.
The chart background features some kind of coloured gradient that became pixellated upon export and import into the PowerPoint deck.
The type was too small that it too became pixellated and grainy to the point that the dive labels are illegible. I would argue labelling each dive beyond its number is unnecessary in the context of Titan’s final dive, but without having listened to the presentation I cannot say for certain.
Next we have the third-dimension. It adds nothing and creates more coloured areas—because the dimension is fake, this a two-dimensional representation of three dimensions—that distract the eye from the important dimension, the length of the bar.
After that, we can look at the axis labels. First, there are far too many. Second, the maximum depth labelling makes no sense. Sometimes, if a line or a bar exceeds the chart maximum once or twice by a small amount, you can let it poke above the top—in this case bottom—line. If you know the rules, you know when you can break the rules. Here, however, the maximum label is 3800 metres.
But Titanic rests at 3840. Ergo 13 different measurements will need to sit below the chart’s maximum—minimum, technically—axis line.
Deep breath.
If she rests at 3840 metres, just add 60 to the chart minimum and you will ahve a final axis label of 3900 metres. Look carefully, however, and you will see in the bottom left how after the final white line, the chart keeps going. Clearly, the designers knew the chart needed more space. This unlabelled minimum is probably 3900 metres given the 100-metre increments used throughout.
But, however, if you add 160 metres to the chart you have a nice, round, divisible number of 4000, which means you do not need to mark the depths in 100-metre increments. It means all the bars sit within the chart. It means fewer pixels on the slide to distract the eyes. (Especially if you drop the background colour.)
Furthermore, if you look carefully at the green boxes, which represent successful dives to Titanic, you can see how the bars break the dimensional rules and are actually flat two-dimensional bars. Perhaps this was only noticeable to me as I worked off the downloaded file at a high-level of zoom to try and figure out the depths as precisely as possible. Or perhaps it is an artifact of the pixellated export of the graphic. If the latter, more of a reason not to make the thing a three-dimensional bar chart.
Then we can get to the colours.
Deep breath.
To start, red-green colour blindness is a thing. I harp on this often and so I will not rehash everything here. No, it does not mean all green and red combinations will not work, one just needs to be careful with them. This one comes pretty close to not working so I would have avoided it.
Secondly, just look at the red. I mean, how can you not. It is very bright and draws your eye almost immediately to all those red bars, particularly the one nearly a fifth of the way in from the right edge. That one is next to one of the successful Titanic dives. My first thought? Oh, that was the final dive. Wrong.
Red means non-Titanic dives. Again, I have not listened to the presentation, but these would presumably be dives of relatively less importance than the Titanic dives. I would not have made the less important dives the one colour that stands out the most.
If you want to go green represents successful Titanic dives and red represents unsuccessful Titanic dives, that makes sense. I can understand the design decision. (Though you would still need to ensure the shades work with each other.) In that case maybe the blue bars represent non-Titanic dives.
Instead, here blue represents unsuccessful dives to Titanic, which of course means the final dive, which of course includes the inquiry’s raison d’être. Not only that, the chart’s background is also blue, which makes visually separating the bars from the background more difficult. This is particularly true at the sides of the chart where the gradient leaves the darker blue.
Finally we have a little orange box with some tiny type pointing out the final dive’s depth. That bit, more visible than the green and orange bars, was still lost to me behind the red bars.
And breathe.
All in all, a mess.
As I noted at the top, halfway through I decided this was such a mess I would prefer just to show how the chart could have been designed. It took a little over an hour to make the chart. Clearly I do not have the chart style guidelines for the Coast Guard, so I just chose a typeface I think worked and then picked some reasonable colours from the deck.
Call me biased, but my design substantially improves the chart. First, you can read the text. Second, the colours fit the brand, do not distract from and in fact highlight the final dive. If I started from scratch, I would prefer to use what looks like the full content area of the PowerPoint slide, but I simply traced over the existing chart. I.e., ideally the chart would have been a little bit taller. I did have to cut out the labels for each dive, but as I stated earlier, they were illegible.
Credit for the original piece goes to the US Coast Guard.
As it happens, the Latino culture largely remains x’ed out on using the term Latinx, according to a new survey from Pew Research.
The issue of supplanting Latino/Latina with Latinx as a gender neutral replacement—or as a complementary alternative—emerged in the general discourse in that oh-so-fun year of 2020 when everything went well.
One common argument I have heard is the inherent gender within the Spanish language. Broadly you use -o for singular masculine endings and -a for singular feminine forms and -os for plural masculine and mixed gender forms and -as for plural feminine forms.
Perhaps my biggest issue is that -x does not linguistically make sense. X is typically pronounced like a j or sometimes an s. Consider how Mexicans pronounce Mexico, May-hie-co. Latinx becomes La-teen-h, an almost silent ending that does not fit, at least to my ears. Pero, hablo solo un pocito Español. Aprendí a hablar por cuatro años en la escuela, dos años de niño, y trabaja en una cocina del restaurante. Thus Latinx, pronounced Lat-in-ecks, always seemed, daresay, a gringo solution to a problem that earlier polling of Latino communities did not indicate was a problem. With the potential exception of the young, but even then not terribly so.
Four years later, however, and not much has changed according to Pew. Their graphic shows as much.
Significantly more people are aware of Latinx as a term. Fewer people use the term, though not significantly. Although a shift from four to three percent can be seen as significant given its low adoption. Moreover, as a second graphic shows, more people who are aware of the term think it should not be used.
The article continues with a discussion of a new new alternative, Latine, which to my ears makes more sense. But is largely yet unheard of in the community—20%—and of those who have heard it, almost nobody uses it.
As far as the graphics go, I am not a huge fan.
For the first, we have two lines showing the movement between two datapoints. At the most basic level, the use of a line chart makes sense to depict two series moving between two points in time. But without any axis labelling one can only trust the lines begin and end at the correct position. Furthermore people need to read the specific labels to get a sense of the line charts’ magnitude. More of a tell, don’t show approach. If the chart had even a simple 0% line and 50% line, one need not label all four datapoints to convey the scale of the graphic.
Ultimately, though, does a chart with four datapoints even need to be graphed? Some would argue in most instances a dataset with fewer than five or six numbers need not be visualised; a table should suffice. Broadly I agree. This chart does show a particularly striking trend of increasing awareness of the term, but largely static to declining usage.
The second graphic, however, falls more squarely into that argument’s camp of “Why bother?” It shows simply two numbers. Numbers placed atop purple rectangles. Without any axis labelling, we presume these bars represent columns encoding the percent—at least the lines in the first chart were clearer to their meaning. Then we still have the issue of telling and not showing. Perhaps labelling to the left from 0% to 75% or 80% would help. Then you need not even add additional “ink” with the four digits sitting atop the bars and sparkling for unnecessary reader attention.
This falls into a broader trends I have witnessed over the last few years in the information design and data visualisation field of labelling individual datapoints within a chart. It is a trend with which I strongly disagree, but perhaps is best left for another post another day. Suffice it to say, if knowing the precise measurement is important, a chart is not the best form. For that use case I would opt for a table, best used to organise and find specific datapoints.
Overall, Pew shows that within the Latino community, very few use the term Latinx. Consequently, perhaps this entire post is, to use a Spanish-language expression, a tempest in a teapot.