Kari: “So, for SwordSTEM it’s ok so long as a topic is about swords and studied with science?”
Sean: “Yes…”
Up until about a year ago, I had been using Instagram for one purpose: documenting fashion challenges. I was part of fashion groups on Facebook that would host monthly challenges where you would tailor your outfits to the calendar’s daily theme. Themes like “Under the Sea” would get me to wear a skirt with a seashell print or “Favorite Color” would have me wearing my brightest orange dress. I’d post a photo of my outfit to Instagram with a hashtag for the others participating in the challenge to search on, and I’d be lucky if the likes on my photos reached the double digits. One day I posted a photo with my Blackfencer synthetic longsword and got nearly 100 likes on the photo:
I wondered if this was a coincidence, so of course I decided to test my theory of my photo getting more likes because of weaponry by posting my outfit photos posing with swords sometimes. After about a month of this experiment, I found that photos I posted that had swords in them got about 3 times as many likes on average as photos that did not contain swords.
The big question is: does this trend hold as a general rule? To answer this question, I became a stalker of various HEMA folks’ Instagram accounts.
Before We Get Into It…
Another phenomenon I noticed was that I started receiving some special messages from members of the HEMA community.
<24 Hours Pass>
<2 More Days Go By>
<Blocked>
I really wanted to write an article about harassment, but there was no systematic way to gather data on it. The most I can give is anecdotal evidence that it happens to me at least once a month, if not more, from members of the HEMA community. And in the interest of making this article fit for public consumption, the above screenshot contains one of the milder conversations; they get much worse. Don’t be like these people. This type of conversation is unwelcome at best and is, unfortunately, not uncommon for female HEMAists.
Data Collection
In order to make any sort of conclusions about whether swords make people look cool or not, I had to collect a sizable amount of data. I picked 100 different public Instagram accounts of various folks involved in HEMA– men, women, club accounts, accounts with tens of followers, accounts with tens of thousands of followers. These accounts needed to have both photos with and without swords present in the photos. A “sword” was defined as any weapon with an edge on it– so axes, daggers, and spears counted, but not boffers, polearms with hammer heads, etc. A photo’s likes were only analyzed if there was a person (or part of a person, like a hand) in the photo; so no photos of just swords or photos of a person’s dinner were included for the analysis. Their most recent 50 photos that met this criteria were used for the analysis.
Examples of sword photos:
Analysis
The methodology for this analysis is pretty straightforward:
- Document the amount of likes a photo has, and whether that photo contained a sword
- Average the amount of likes for each IG account for “Sword Photo” and “Non-Sword Photo”
- Normalize the averages
- Normalization is necessary because different accounts have different amounts of followers. Accounts with many followers are treated the same as accounts with few followers for this analysis.
- Run a paired t-test to see if there is a statistically significant difference between “Sword Photos” and “Non-Sword Photos”
- A paired t-test is used when you have one individual with two sets of values. Usually this kind of test is used when you’re dealing with something like a pre/post test, but it can also be used for something like this which contains an intervention (aka, having a sword in your photo).
If you remember from my prior article Do Fighters Game Rule Sets?, statistical tests are evaluated with a p-value. We want to see p-values less than .05 in order to say that any differences we see aren’t simply due to random noise in the data.
For this analysis, the p-value is less than 0.0001, which means that there is practically no chance that the differences seen in the likes on Sword vs Non-Sword photos are random.
Sword Photos received more likes than Non-Sword Photos.
Here are some additional interesting numbers:
- Of the 100 accounts looked at, 84 of them get more likes on their sword photos versus their non-sword photos.
- A photo with a sword in it will receive, on average, 41% more likes than a non-sword photo.
- The account with the highest increase when showing a sword in their photos received over 5x as many likes on their sword photos than non-sword photos.
The Control Group
Some of you might be thinking, “Accounts that have HEMA people in them are going to be followed by other HEMA people. And HEMA people like swords, so of course photos with swords in them are going to get more attention than photos without swords.” As such, we need a control group. But what kinds of people are going to be posing with swords if they aren’t necessarily training with swords?
Cosplayers.
For the cosplayer cohort, I used the same methodology as I did for the HEMA folks: 100 different accounts with varying levels of followers, using their most recent 50 photos containing a mix of swords and not swords. I did loosen the criteria on what a “sword” is by allowing foam or prop weapons rather than requiring them to be true weapons used for sparring.
After going through the same analysis steps as performed on the HEMA group, the p-value is less than 0.0001.
Once again, Sword Photos received more likes than Non-Sword Photos.
Here’s some additional facts about this group:
- Across the 100 accounts, there were 227 unique sword-wielding characters cosplayed from 128 different games/anime/movies/etc, with an additional 25 original designs.
- The most frequent sword-wielding characters cosplayed by these creators were Sora from Kingdom Hearts, Cloud from Final Fantasy VII, 2B from Nier Automata, Ezra Scarlet from Fairy Tale, Ciri from The Witcher, and Tanjiro from Demon Slayer.
- Of the 100 accounts looked at, 69 of them had a higher amount of likes for characters with swords than their characters without swords.
- A cosplay with a sword in it receives, on average, 17% more likes than a cosplay without a sword in it.
