Evolving Pokemon Network

September 5, 2016

Do you remember my Pokemon network? Well, like a real Pokemon, it evolved significantly in the last few weeks to become even more massive! To complement what I wrote earlier, let’s do a bit of data-science and bring some facts to the table.

Quick recap for the latecomers: I created a network of all Youtube videos from the playlists of Pokemon fans. In this network, each node is a video and videos are linked if they appear consecutively in the same playlist. You can read all about it in my Pokemon Data Viz article and even enjoy a 3D visualization of the previous network.

Disclaimer: All the data presented here are timestamped to the 3rd of September 2016.

Some facts about the Pokemon network

48k VIDEOS

79k LINKS

153M LIKES

9.1M DISLIKES

7.6B VIEWS

Yes, you read that right! The total number of views for the 47,927 videos is exactly 7,635,026,806. It’s huge but let’s not forget that the Gangnam style video has more than 2,637,228,375 views itself…

Cumulative distribution stats

An interesting thing to do is to look at the cumulative distribution of views. Basically, we sort all the videos for the most viewed to the least viewed and we show how many percents of the total number of views the k most popular videos represent.

Cumulative stats for the most viewed Pokemon videos in log scale

Wait, how do I read this chart?

To make it clearer, let’s take the black curve representing the cumulative distributions of the number of views. If we look at the first 10,000 most-viewed videos, we can see that it represents 90% of the total number of views for over more 47k videos.

It is not that impressive but if you look closely you will notice that the horizontal scale is in a logarithmic scale. It means that for each grouped bars we change the order of magnitude and that the top-100 videos account for 15% of the total number of views (1,245,584,007 views). Now this is huge!

So what are the most appreciated videos?

The cool thing about Youtube is that anyone can make playlists of videos without necessarily uploading anything. Here I’ll share some of the top videos in form of Youtube playlists so you can enjoy them comfortably from your couch.

Most popular Pokemon videos
Most liked Pokemon videos
Most liked appreciated videos

To define the appreciation I simply divided the number of likes by the number of dislikes. This ratio is interesting because it reveals the most loved videos from the community without the haters.

Evolving Pokemon network

Now that we learned a bit more about the network of Pokemon videos it could be interesting to visualize it and see how the Pokeball shape of the network changed in the last 15 days. The last image shows in flashy green the difference between the previous network and this new one.

Pokemon video network
The new Pokemon network at the 3rd, September 2016
The new Pokemon network at the 3rd, September 2016
In green, videos added in the Pokemon network between the 17th, August 2016 and 3rd, Septembre 2016

The hype is on!

I heard that the popularity of Pokemon was declining in the last couple of weeks. Let me tell you that it is plain wrong and I have the data to prove it.

Since the last post, the network gained 2.3 B of views! In other words:

30% of all views on Pokemon videos since the beginning of all Youtube videos happened in only 15 days.

This fact really demonstrates that the hype around Pokemon Go is still raging and is continuously growing. Take that, naysayers!

Of course, in a few months, the acceleration of views will probably decrease like any other viral phenomenon but until then: go catch ’em all!

Concluding thoughts (let’s get serious)

From an academic standpoint, it is really fascinating (at least to me) to study the evolution of dynamic networks. This example of ever-changing graph reveals a lot but also leaves some interrogations that are yet to be answered.

For example, how will the structure behave once the hype is over? Is the topology (structure) of the network similar to the one around another popular game, or to the one another completely different topic? How can we compare these dynamic networks?

I’ll leave some of these answers for the next episode. Until then, don’t hesitate to share this post and follow me on the social media. @KirellBenzi /kirell.benzi

Enroll now to learn how to create art with data.

Thank you! Please check your email to validate your subscription :)
Oops! Something went wrong while submitting the form.