Who is Winning Between Upshot and FiveThirtyEight (Answer: We Are)

In the battle between FiveThirtyEight.com and NYT’s Upshot, we are the winners.



NYTimes.com Upstart

There is a great battle taking place. I’m not talking about the race for President. I’m talking about the battle over which website is best at using election-oriented explanatory graphics (or “data journalism”)–NYTimes.com’s Upshot or ESPN.com’s FiveThirtyEight.com?*

Currently, users of both sites are winning because both sites are using Edward Tufte-inspired methods of displaying real-time trends and variations. (Tufte is a Yale professor emeritus who thinks PowerPoint is evil and passionately dislikes the cartoon illustrations with numbers that are popularly called infographics.)

2016_Election_Forecast___FiveThirtyEight 3Here’s an example of a Tufteian influenced display of data found on 538. When one typically sees a Red and Blue map of the U.S. that shows which candidate is leading in each state, the map adheres to a geographic display.

2016_Election_Forecast___FiveThirtyEight 2However, because each state, thanks to the Electoral College, has a say in the election based primarily on its population, that key data point is ignored when the map is displayed geographically. A more insightful way to display a Red and Blue map is this way, as demonstrated on FiveThirtyEight.com.

Another impressive use of data journalism is the way in which Upshot displays the ever-changing “paths to victory” that each candidate has to victory. (below)

nytimes-pathsThis is perhaps the most revealing display of the daunting challenge Trump faces that I’ve seen. Using the current assumptions from Upshot’s forecasting model applied to current data, Clinton has 998 paths (or 97% of the ways) while Trump has only 24 paths (or 2.3%).

*There’s a lot of history between Nate Silver (founder of fivethirtyeight.com) and the New York Times. I’ve never had much interest in such back-stories. However, competition is a good thing.

Sometimes, a ‘wisdom of crowds algorithm’ needs the wisdom of a lone editor

A five-day old story is popular with readers, yet the "news" has lapped that story.

The website of the Wall Street Journal, WSJ.com, has a feature most news sites have, a leader board that ranks the popularity of articles on the site. The WSJ.com ranks “Most Popular” by articles Read, Emailed, and Commented on. It also has a ranking of videos that are being watched.

Sometimes, such algorithmic determinants of popularity go nuts, however. Like now, for example. A few minutes ago (Tue., Feb. 21 and 3:52 ET), I snapped the accompanying screen grab showing that the #2 most popular “read” article now on WSJ.com is one dated Feb. 16 at 12:58 ET with the headline, “Colbert Report Suspended.

Anyone not living under a rock (which apparently a lot of WSJ.com readers do) knows that, according to the WSJ and every other news outlet on the web, Colbert’s show is back on the air.

Leader boards of trending or popular news stories can be helpful in giving you an idea of what people are interested in.

But as my favorite online “popularity tracking” algorithmist, Gabe Rivera of Techmeme, MediaGazer, Memeorandum.com, et al, discovered a long time ago, even algorithms that track popularity of content need a little editing now and then.

Lies, damn lies and the other kind of statistics

I just saw two items as I was glancing down the headlines in my news reader.

“Google’s Android becomes the world’s leading smart phone platform,” says the headline on the press release from a company conducting research that claims Google’s open source mobile operating system was on 33.3% of the smart phones shipped in the 4th quarter of 2010.

“Apple’s 4% mobile market share rakes in over half the industry’s profit,says the headline on a blog post about another mobile phone industry research that focuses on “profitability.” Wow! Only 4% of mobile phones have Apple’s iOS operating system on them, but they account for over 50% of the industry’s profit. Gee, that also sounds like something that would be leading.

I have no reason to believe if either  of these items are accurate. Frankly, I doubt either  is.

Nonetheless, when paired, they demonstrate a point I’ve tried to make here time and again: Never believe any statistic, unless it confirms what you already believe.*

*For readers unfamiliar with this blog, that was said in jest.

How a chart can suggest the opposite of what is says

chartdirection-20101223-213736I’ve read conflicting claims about which day of the year is the busiest travel day. (The day before Thanksgiving and the day after Christmas are favorites of bored TV news crews.) No matter what precise day you choose, if you combine one of the “busiest travel days of the year” with an eastern seaboard blizzard, I’m guessing this could be one of the worst air travel days ever. So, I’m glad my family has no one “enplaning” today.

And since they’re already so overwhelmed, I guess I should feel a little bad that I’m posting this item I wrote a few days ago (while delayed in an airport for six hours) but didn’t realize until a few moments ago that I had mis-dated it, and it had not shown up on my blog.

This post has nothing to do with travel or weather. It’s about how to inadvertently communicate the opposite of what you intend to — with a simple bar chart.

Let’s go back and pick up the post where I had originally started it:

For some reason, the Metropolitan Nashville Airport Authority sent me a printed copy of its annual report (Here’s a PDF version. As I was flipping through the report hoping I’d find an item that says they’ll be offering free wifi soon (to no avail), I noticed two charts on page 18 that I’ve included with this post (if you are reading it on my blog). The charts compare the number of passengers and the weight of planes originating that the airport for the past three years.

Like most Americans and Nashvillians, I read bar charts in the traditional western way, based on the left-to-right display of ascending numbers. That should come as no surprise, as most of us were taught to expect the left-to-right ascending number concept by everyone from Bert & Ernie to college math professors. I, therefore, assume that bar charts that compare year-to-year data have the oldest data represented in the left-most bar, and then, as Bert & Ernie taught us, the next-oldest data to its right, and so-on. So, when I saw the bar charts moving higher, from left-to-right, I assumed this bar chart was a visualization of a metric that had gained in volume each of the past three years.

Yet upon second glance, I noticed something I thought to be rather odd — or, perhaps mis-leading. The years were descending, left-to-right.

Was this an intentional switcharoo? Or was it just a mistake? Apparently, neither. A quick googling of past annual reports by the commission convinced me they’ve always used this convention: descending years in bar charts that compare year-over-year. What’s with that? I can only guess that some accounting requirement is at work, rather than the desire to use a common-sense, “user-centric” visualization.

[I still think Nashville should join the dozens of airports providing free wifi so that travelers can better endure travel every day, and especially days like today.]

Magical Mystery Math: The Beatles First Day on iTunes

While it ended up being a lot less eventful than a typical Apple announcement, the much delayed (by 8 or so years) appearance of 13 Beatles studio albums and the “Beatles Box Set” on the iTunes store finally occurred today (Tuesday). While I didn’t join in the buying binge, I was curious what was selling. So, 12 hours after the big announcement, I looked at the iTunes store best selling albums page.

On first glance of the screen-grab of the top 20 albums (from about 10:30 p.m., CST), it may appear odd that the top-selling Beatles album is only #6 (Abbey Road).

But a second look reveals some rather amazing stats:

40% of the top 20 albums are by the Beatles.

The 11 non-Beatles albums have been available to fans for an average of three days. The 8 Beatles albums have been available to fans for an average of 40 years.

The 8 Beatles albums (including compilations) have previously been purchased (certified) 246 million times. The 11 non-Beatles albums (including compilations) have been purchased previously zero times.

The average price of a non-Beatles album is $11.77. The average price of a Beatles album is $32.63 (the $149 box set ups the average).

$141 – The cost of purchasing all 12 non-Beatles albums.

$261 – The cost of purchasing all 8 Beatles albums.

Or, if you purchased all 20 albums, the Beatles would account for 54% of the price. The next highest: Taylor Swift, 3.5%

Bottomline: Despite me making no purchases, the Beatles and Apple sold lots of music today.