Thread by @RobertGReeve

OK nerds, it’s time for some data literacy. You should be deeply skeptical of this map of D&D demographics.

Let's dip our toes in politics in a Monday night TIPSY THREAD 🥂

https://twitter.com/cawoodpublish/status/1442149297294741511

First of all, no shade to Cawood or the marketing manager at Zippia who wrote up this content, because as a marketing piece, it’s successful! zippia.com/advice/states-…

Cawood's viral post cited the source, and in the marketing world, perfect content is the enemy of good content.

The consequences for being wrong are low; it’s not covid data. But the flaws in methodology show why scientific research needs peer review (and why ANY news about a study that hasn’t been peer reviewed should give you pause).

🚩 #1: Undefined terms

This map is pretty. But look closely and there aren’t any numbers on our key. What does “most” mean? Players? Games? Hours? Is it just Dungeons and Dragons, or is it all tabletop RPGs?

The first line of the piece says “there are an estimated 13.7 million D&D players” without a citation

If I fire up Microsoft Edge and Bing “number of D&D players,” the top result is a 2yr-old article of napkin math based on 2018 data from Roll20

(yes Google's #1 result matches)

Meanwhile in 2018, @NathanBStewart estimated 12 to 15 million Dungeons & Dragons players in North America alone, and a 2019 infographic cited "40 million fans around the world." I trust his estimate a little more, even if "fan" is an unclear term. seattletimes.com/life/lifestyle…

🚩 #2: Methodology

Okay, lots here. “very comfortable” using 1yr of “search volume as a proxy for users.” Proxies are OK with appropriate controls, but this one isn't. And wait, what’s this?

“The 15 most common DND classes." Hmm. We’ll get back to that. First, Google Trends.

a-HA. Google Trends finally defines that ambiguous "Most" for us: it's a PERCENTAGE, not a count. So we’re seeing the states where the greatest percentage of searches are about D&D.

Not the states with the most searches about D&D.

That's why this piece says Utah, Idaho, Alaska, and Montana are D&D hotspots, and Texas and California are tabletop wastelands.

New York City has over 8 million people with competing searches. That’s why NY is dead last. The percentage of searches about D&D is smallest there.

🚩 #3: Contradictory data

The election haruspices at 538 crunched D&D Beyond data back in 2017. Top ten combos:
1. Human fighter (bo-RING)
2. Elf ranger (hi Legolas)
3 – 8. yet more humans and elves
9. Dwarf cleric
10. Tiefling warlock

https://twitter.com/FiveThirtyEight/status/918506562972540928?s=20

Yet Zippia's piece also uses Google Trends data to guess each state's favorite class:
· Monk: favorite in 20 states
· Druid: 9 states
· Ranger: 9 states
· Paladin: 3 states
· Artificer, barbarian, bard, cleric, fighter, rogue, wizard: 1 state each
· No state likes sorcerer 😢

This brings me to 🚩 #4: expert opinion

Have an expert gut-check your data. The piece mentioned "the top 15 classes." Even accounting D&D Beyond's 12 base classes, artificer, and @MatthewMercer’s Blood Hunter, what's the 15th class in our data?

Monk and ranger aren’t exactly beloved by the community. So we know we should double-check the data and our assumptions about it.

Could it be that more people search monk, druid, and ranger because those classes’ mechanics are more complex and less understood?

NOW LET’S GET POLITICAL.

You might be going “wow, is percentage a flawed measurement? We’re putting our finger on the scales to favor sparsely-populated states like Utah, Idaho, and Montana.”

Yeah. This same problem exists with the STRENGTH OF YOUR VOTE.

When Vox’s @imilhiser looked at 2020 census data, he found California has 68× more people than Wyoming. Yet both contribute 2 senators to the US Congress.

He also found that in the Senate, Democrats represent 41.5 million more people than Republicans. vox.com/2020/11/6/2155…

It's a percentage problem.

California's senators represent 18.6 million people each, while Wyoming's senators represent only 284,000 people each.

(@TheGreenPapers's website is charmingly dated but they're longstanding, respected election data wonks)
thegreenpapers.com/Census10/FedRe…

This 2012 Slate article explains that "One Wyoming voter has roughly the same vote power as four New York voters."

This is why people say the Electoral College is undemocratic. By Electoral College logic, D&D is a bunch of Utahns playing monks.

slate.com/articles/news_…

Compare Zippia's map to Slate's map measuring the power of your electoral vote.

The data's dated so not a perfect match, but look at West Virginia, Vermont, Wyoming, the Dakotas.

Look at New York, California, Texas, Florida. slate.com/articles/news_…

And I lived in DC for 10 years, their 700k residents have 0 senators representing them. Zippia left DC off completely!

DC elects a "shadow delegation" (yes it's really called that) of senators and representatives whose hypothetical votes count the moment #DCStatehood happens

In conclusion, 1) learn how to read data to become a better, more informed citizen

2) the Electoral College allots power disproportionately

2) if you're a group of Mormons playing a kung fu campaign, start live streaming.

This thread took off but I made some small oopsies. It’s @imillhiser with a double-L, and my conclusion goes 1, 2, 2. “Write drunk; edit sober.”