Introduction
Twitch is a video live streaming platform owned by Amazon that mainly serves as a conduit for broadcasting video games. Streamers use this online service, with over 15 million users watching daily, to broadcast their gameplay and interact real-time with their viewers. As of 2020, the platform had over 3 million broadcasters on a monthly basis, many of which are company partners.
Some video games streamed on Twitch have reached immense popularity, such as Epic Games’ Fortnite, Riot Games’ League of Legends, Valve’s Dota 2, PUBG’s Battlegrounds, Blizzard’s Hearthstone, Overwatch and Valve’s Counter-Strike: Global Offensive which combined, have surpassed 356 million watch hours on the platform. This report aims to analyze relationships between streamers, watchers, and watch behavior among the most popular games played on the platform.
Tools
Excel & Tableau Public.
Data
Top Games on Twitch 2016-2021 – Kaggle
Methodology
Using the data collected from Twitch, this study will analyze relationships between attributes and activity of the plethora of games played on the platform. In order to understand the relationship between popularity among broadcasters and popularity among the audience, I will be evaluating how hours streamed and hours watched correlate. While aggregating data across several years will provide a substantial starting point for this analysis, it is important to note that these trends tend to vary greatly over time. Analyzing each year individually could produce very different results.
Results
In order to set a proper precedent for this study, it is important to understand which games are most popular among the endless list of video games played on Twitch. When a streamer goes live on the streaming platform, they tag their stream with the game that they are playing. This data can be very easily tracked, thus a running list of streams and their corresponding games exists. By ranking games played on the platform from 2016 to 2021 by hours watched, the following list of most popular video games on Twitch is created (Figure 1). Note that the “game” labeled as Just Chatting, is actually a category of content that involves the streamer having some sort of casual conversation with their viewers rather than playing a game. The decision to leave this category of streams in with this analysis is to see if there are any patterns with this sort of content that contrast with those of video game streams.
![](https://i0.wp.com/studentwork.prattsi.org/infovis/wp-content/uploads/sites/3/2022/10/Sheet-1-1024x196.png?resize=840%2C161&ssl=1)
Although these are the most popular games by watch hours, it is also important to note that there is a difference between hours watched and hours streamed. The scatter plot visualization shows the general relationship between these two attributes (Figure 2). The trend line indicates that there is an obvious positive correlation between hours watched and hours streamed, but there are some significant outliers to this pattern. On average, hours watched is approximately 25 times higher than hours streamed; the Just Chatting category, for example, breaks from this trend having almost 74 times more hours watched than streamed. This indicates that either fewer streamers are broadcasting this category, and/or that more people are watching it.
![](https://i0.wp.com/studentwork.prattsi.org/infovis/wp-content/uploads/sites/3/2022/10/Sheet-3-1-1010x1024.png?resize=840%2C852&ssl=1)
To further analyze this relationship for the outliers, each of the most popular games can be evaluated on the basis of watch and stream hours. The size of each nested area in the tree map below is determined by hours watched while the gradient of color deepens as hours streamed increases. It is evident that the correlation between the two attributes may be more complicated. For example, while Riot Games’ League of Legends is most popular by number of hours watched, it is far behind Epic Games’ Fortnite in hours streamed, topping the charts at 300,000,000 hours. While some games may be streamed very frequently, that does not necessarily mean they will be as frequently watched.
![](https://i0.wp.com/studentwork.prattsi.org/infovis/wp-content/uploads/sites/3/2022/10/Sheet-2-1-1024x601.png?resize=840%2C493&ssl=1)
While hours streamed is an important metric to consider, it is also valuable to note the concentration of streamers in each of these video game categories. One would expect a direct correlation between the number of streamers broadcasting a game and the number of hours of a game streamed, and looking at the first few entries in the table of values below this is true (Figure 4). In most cases, it seems to be that if there are more streamers broadcasting the game, it will evidently be getting more views. There are exceptions to this in that games which have very high view hours such as Valve’s Dota 2 or Riot Games’ Hearthstone, do not have as many people broadcasting it, in fact, they are not even in the top 10. Although this level of granularity is not possible with the dataset in use, a breakdown of viewership by streamer would be helpful in understanding why some of these streamer counts are inflated.
![](https://i0.wp.com/studentwork.prattsi.org/infovis/wp-content/uploads/sites/3/2022/10/Sheet-5.png?resize=531%2C319&ssl=1)
Although these general findings are quite interesting, it is important to note that they are based on averages over the course of several years. Some of these statistics vary greatly if periods of popularization and release date are taken into account. The area chart in the figure below shows how over time, some of the popularity in games on Twitch has shifted dramatically. The contrast between the landscape in 2018 and currently is noteworthy; during this year, Epic Games’ Fortnite’s viewership was approximately 10 times bigger than that of the Just Chatting category. As of 2021, Just Chatting is more than 3 times larger than Fortnite. While overall the hours watched of tried and true video games is high, the landscape of the platform is quickly changing to popularize a more casual form of content in Just Chatting.
![](https://i0.wp.com/studentwork.prattsi.org/infovis/wp-content/uploads/sites/3/2022/10/Sheet-6-1024x661.png?resize=840%2C542&ssl=1)
Reflection
This study undoubtedly produced some noteworthy results, but they only scratch the surface of correlations in viewer and streamer activity. Given that the initial interest was quite open, the resulting statistical analysis was also quite broad. If this topic had been narrowed further then more telling results could have been produced to potentially explain why some correlations are happening.
If the study were to continue, an analysis game popularity aggregated by category would be very interesting. The trend of public interest on Twitch in a variety of video game categories would be very telling as to how the times have impacted viewer preference. A similar study in ESRB ratings linked to this data would be very fascinating.