The ensuing impact of Elon Musk’s now famous ‘Gamestonk!!’ tweet over time | Skanda Vivek

On Jan 26th at 4:08 PM EST, Elon Musk tweeted “Gamestonk!!” in reference to the “wallstreetbets” Reddit page, which now has over 9 million subscribers. Interestingly, this was just a few minutes after the stock market closed, one has to wonder if Musk timed this.

The following day, Gamestop stock (NYSE: GME) jumped 2.5x. We know how the stock market reacted to Musk, the data is out there. Just google NYSE: GME and you will see the day to day changes in Gamestop stocks.

But do we know how Musk’s tweet actually impacted the stock market, I mean can we follow this impact beyond simple correlative measures, or worse — presumptions? One step in this direction, is to evaluate how Musk’s tweet impacted Twitter itself, in particular, tweets referencing Gamestop.

The good news is, twitter’s new API v2 has more features than v1, and gives some excellent data access, especially for academic research.

Similar to how this paper investigated social tweeting networks around hashtags, let’s look at the #gamestop tweet networks during that time.

The top figure shows a small spike when Musk tweeted, just after the stock markets closed.

Interestingly, there is a small bump from nearly no one tweeting #gamestop to a maximum of 1000 tweets referencing #gamestop around 6–7 PM on Jan26th, and then a slow die off.

After the night, during the day on Jan 27th, there is a slow upraise, with a sudden jump around 9 AM, the opening of the stock exchange.

Possibly people tweeting in response to the large jump in Gamestop stock opening price. Later in the day there is a plateau with a consistent number of ~3k hourly tweets mentioning #gamestop till the end of 27th.

On the 28th there is an even higher jump than on the 27th, even though we see a contrast in the stock market. Gamestop stock prices on the 28th opened lower than on 27th.

Maybe twitter users who made a risky bet and bought Gamestop stock at the peak, that thought they could increase the price back up by franticly tweeting :)

Apart from my half-baked hypotheses, there is clearly something really interesting going on.

On 26th, Musk’s tweet raised stock prices the following day (at least that’s what most people say), but on 27th and 28th, it’s possible the reverse happened — crazy stock markets prompting a frenzy of tweets around #gamestop.

Most Frequent Usernames

From each tweet, it is possible to get the username (starting from@), of the user, as well as anyone mentioned in the tweet. From the ~170k tweets, there were ~130k unique users either tweeting or mentioned. Let’s look at the top 10 frequent usernames:

Surprisingly, the most common is DonaldJTrumpJr! Musk barely makes it to the list, at #10! Another way of visualizing this is through a word cloud. You might be able to see elonmusk if you squint!

Complex Network Analysis

Apart from simple frequencies, we can build the network of tweets referencing usernames, to build the #gamestop graph network where nodes are users, and directed edges denote users referencing others with through @. I used the python NetworkX package for this.

After building the graph, you can look at various centrality metrics, which basically give information of the importance of nodes by various criteria.

The most intuitive is degree centrality, where a node importance is based on how many connections it has relative to other nodes.

For directed graphs we use in degree, which is how many edges reference a username, as the usernames importance, Below you see the 10 higher in degree centrality nodes:

Again, surprisingly DonaldJTrumpJr is right at the top! It is possible that a lot of users are referencing DonaldJTrumpJr are themselves not too popular. Instead, let’s filter out the graph to include only popular usernames, above a certain in degree (I chose 15):

Basically the above 10 usernames are the 10 top sub-influencers amongst the top influencers. Interestingly, DonaldJTrumpJr is not the top anymore.

Conclusions and Unanswered Questions

In starting analyzing #gamestop tweets during the crazy stock markets, I thought it might be interesting to observe tweet numbers following stock market trends.

In this naïve presumption, I didn’t realize how fascinatingly complex the whole scenario could be. Tweets influencing stocks, stocks influencing tweets, and complex interdependent, dynamic networks. Here’s some food for thought:

1. Why are there peaks in #gamestop tweets on 27th and 28th, days after Musk’s tweet? Is it in response to the stock market fluctuations? And why is 28th so much higher than 27th, considering the stock market itself didn’t soar on the 28th?
2. Why the sudden plunge in number of #gamestop tweets at 12 PM EST on the 28th?
3. What’s up with Donald Trump Jr. being so important during #gamestop?

If you are interested, the Google Colab notebook with the code is detailed below.

Channeling my inner

Aleksa Gordić: the more I explore, the more I am aware how poor my understanding and hypotheses are. But that makes it all the more interesting to delve deeper into interdependencies between societal networks.