Twitch Data Reveals Secrets About Gaming

Publicly available Twitch data tells a fascinating story about the gaming industry today. First, let’s reflect on how big the Twitch audience is. In 2015, Twitch averaged 1.7 million broadcasters (people who stream game footage on the site) each month and 100 million viewers. Throughout the year, those video game streams were seen by more than 500 million people who watched an average of 422 minutes of games during the year. The peak concurrent viewers on Twitch in 2015 was 550,000, surpassing the average viewership of many cable television networks, like CNN and E!. Incredibly, some Twitch users are broadcasting for more than 10 hours per day in an effort to turn a gaming hobby into a business. This ecosystem is creating millionaire gamers, like UNIVeRse and Pewdiepie.

In an effort to quantify more details about the Twitch ecosystem, we analyzed data based on concurrent streams for a set of games (listed at the bottom of the post) on June 14, 2016.

Despite League of Legends being the most popular game in the world (by a long shot) with 27 million players per day and reigning as the most popular game on Twitch in 2015, on June 14th, the new release, Overwatch had 1.4x more concurrent streams — Overwatch is clearly a hit with the Twitch community!

As might be expected with a growing user base and a snapshot in time, the accounts that were streaming on June 14, 2016 were overwhelmingly created in the last 2 years. Each game has a slightly different distribution of account creation. For instance, Call of Duty: Black Ops 3 had a high percentage of channel creation in 2015 (the game was released in November 2015). But, this 2015 peak in channel creation holds true even for games that were released many years ago, like Counter-Strike: Global Offensive. This snapshot may demonstrate a tendency for the majority of streamers to stick around for a short time before moving on once they realize their audiences are not increasing in size fast enough.

Twitch channels with significantly more followers (2 standard deviations above the mean…i.e. the 95th percentile) than their counterparts broadcasting the same game tend to have been created several years before the vast majority of players.

Clearly, it takes years of consistently playing on the site to build an audience that would qualify you to build a following significantly above the mean for a particular game. In fact, based on this snapshot, 75% of the players with the largest followings for each game created their channels in 2013 or earlier.

The same holds true for Twitch Partners. These channels have significantly more followers than non-followers and were created several years before channels that are not considered Twitch Partners. There’s definitely a high correlation between the Outlier channels and the Partner channels because having a lot of followers seems to be a significant criteria in becoming a Twitch Partner.

This is the first post in our series examining Twitch data. We will investigate several storylines related to this data and analyze interesting trends further.