Data Science

A Comprehensive Twitch Streamer Exploratory Analysis

Not just pixels, not just kids!

7 min readSep 27, 2021

· Motivation Behind Choosing This Topic?
· What is Twitch?
· So you still play video games?
· Exploratory Data Analysis
∘ 1. Loading necessary libraries
∘ 2. Sourcing data from Kaggle (Top 1000 Twitch Streamers of 2020)
∘ 3. Importing the data into a data frame
∘ 4. Wrangling and pre-processing
· Defining features
· Analyzing the highest-ranked channel for each feature
∘ I) Watch Time
∘ II) Stream Time
∘ III) Peak Viewers
∘ IV) Average Viewers
∘ V) Followers
· So who wins the lucrative follower game?
∘ Describing the calculated metrics
∘ Which channels bagged a long viewership with lesser streaming time?
∘ Who had the smallest gap between peak and average viewership?
∘ Which channels scored the highest peak viewer-to-follower ratio?
∘ Which channels scored the highest average viewer-to-follower ratio?
· Languages used by channels
· Scope for further analysis
· Sources

Motivation Behind Choosing This Topic?

Nestled in the comfort of my home last year, I indulged myself in video games, as did my friends. Until then, I had never been so involved in gaming. Whenever a discussion of the latest A.A.A. titles struck up, I considered myself an outcast. After playing CounterStrike: Global Offensive for over a thousand hours, watching gameplay streams and tournaments, I am amazed by the Esports industry and its vastness. More as an admirer than a player.

What is Twitch?

Twitch is an American live streaming service that focuses on video game live streaming, including broadcasts of Esports competitions. In addition, it offers music broadcasts, creative content, and, more recently, “in real life” streams. It is operated by Twitch Interactive, a subsidiary of Amazon Inc.
It was founded as a spinoff feature on now-defunct Justin. Tv by Justin Kan, Emmett Shear, Michael Seibel, and Kyle Vogt in 2007 to allow anyone to broadcast gaming online.

So you still play video games?

Without passion, you are already dead.
(Max Payne)

Contrary to the belief that gamers were stigmatized for spending endless hours fighting over pixels, gaming needs risk-taking, leadership, and strategizing as its core qualities. It has presented enormous economic opportunities for players, strategists, game data analysts, and many more.

With companies battling for breathtakingly fast processors, better graphic cards, and churning gaming titles of the highest aesthetic appeal, the gaming industry is here to make a mark rather than looking back. Formally denoted as Esports, the industry is growing in millions of dollars.

Tech-giant Microsoft came up with its own streaming service called Mixer, paying Shroud and Ninja approximately $20–$30 million to leave Amazon-owned Twitch for its platform. However, Mixer shut down its services and merged with Facebook gaming, while Ninja and Shroud returned to Twitch.
Twitch has been thriving under Amazon’s able leadership since 2014 and clouded its nearest competitors, Facebook Gaming and YouTube.
Amidst all hullabaloo, Twitch emerged as the undisputed champion of video game streaming.

Time to dive and crunch some numbers.

Exploratory Data Analysis

1. Loading necessary libraries

2. Sourcing data from Kaggle (Top 1000 Twitch Streamers of 2020)

3. Importing the data into a data frame

Top 1000 Twitch Streamers Data for 2020

4. Wrangling and pre-processing

Dropping columns followers gained and viewers gained since we do not have comparative data from the previous year (2019)

Defining features

Watch Time — total watch time by all viewers simultaneously
Stream Time — No brainer, the total time the channel has streamed live video
Peak Viewers — The highest number of viewers at any moment in the live stream
Average Viewers — Calculated over a period of one year
Followers — Number of Twitch users following the channel
Partnered — Twitch’s second stage of monetization program, after Affiliate
Mature — If the channel streams explicit content
Language — The language primarily used on the channel’s streams

Note: Viewers may or may not be following the channel

Analyzing the highest-ranked channel for each feature

I) Watch Time

Félix Lengyel, better known by his online alias as xQc or xQcOW, is a French streamer, popular internet personality, and a former Overwatch professional player representing Dallas Fuel and Team Canada in various events.
Currently, he is a variety streamer on Twitch for Luminosity Gaming living in Canada. He has won and been nominated for multiple awards as a professional Overwatch player and Esports streamer.
However, his career was still marred by account suspensions, warnings, and other violations due to his abusive in-game chat, stream sniping, and streaming explicit content.

Rifftrax (not a gaming channel) is an American company that involves roasting particular television films and programmes by comedians Michael J. Nelson, Kevin Murphy, Bill Corbett, and others.
The three comedians’ commentating style originated from their earlier television series, Mystery Science Theater 3000 (MST3K), in which they would similarly mock (or “riff”) films aloud while watching them.

Initially, an American game developing firm, publisher, and esports tournament organizer, Riot Games, was the town’s talk when it released its first-person shooter game, Valorant, in February 2020.
Valorant is a spinoff from Riot’s first-ever and most successful League of Legends.
It improved game mechanics and cut down on in-game chat abuse by establishing a sophisticated report procedure. The hype of the new title release could be a significant reason for having the highest number of peak viewers.
In 2011, Chinese firm Tencent Games bought 93% shares in Riot and later completed the acquisition in 2015.

The International is an annual Esports World Championship Tournament consisting of 18 teams for the video game Dota 2, hosted and produced by the game’s developer, Valve Corporation. It has a prize pool of a whopping 1 million USD.
The number of monthly active users of Dota 2 regularly exceeds 7.6 million.
Dota 2 was the fourth most played game on Steam in 2020, Valve’s video game digital distribution service.

Turner Tenney, also known by his online alias Tfue is an American streamer and esports player best known for playing Fortnite.
Tenney previously streamed games such as Call of Duty, Destiny, and H1Z1, but he transitioned to Fortnite Battle Royale as it quickly gained popularity.
He later joined FaZe Clan, a professional Esports organization. He left Faze after a series of payment allegations and mudslinging.

So who wins the lucrative follower game?

The stakes are high up top, so is the competition.

Tfue has been an active streamer since 2014 and played for three esports teams to date. He regularly played Call of Duty, Destiny, and H1Z1 but later transitioned to Fortnite.
One of the most talented and famous former Esports athlete, shroud is often recognized as an aim god and has won international championships. He streams plays of Counter-Strike: Global Offensive, PlayerUnknown’s Battlegrounds, Apex Legends, Escape from Tarkov, Valorant, and more.
Riot Games is the only channel owned by a firm. The rest are all solo streamers amongst the top ten. This reinforces the hype buildup around its first-person shooter game Valorant.

Five numeric features gave me a lot of values to fiddle with

So I created a new data frame streamer_metrics with some of my features.

Describing the calculated metrics

time_score- This ratio of stream time to watch time is significant to gauge the popularity of a particular channel, then scaled to its max value. A higher ratio implies that simultaneous viewers watched content for longer than it was streamed.
avg_viewer_pc-Since average viewers give us an idea of central tendency, this statistic helps us compare the total followers with the viewers across the year.
peak_viewer_pc-A ratio of peak viewers at any moment to the total followers of the channel
peak_avg_diff_pc -Difference between avg_viewer_pc and peak_viewer_pc as a measure of consistency in terms of viewers

Which channels bagged a long viewership with lesser streaming time?

Minimum effort, maximum results.

dota2ti and dota2ti_ru channels stream Dota 2 gameplay, tournaments, and guest plays.
Developed by Valve Corporation, Dota2 tops the list in maximum active players internationally
auronplay is one of Spain’s most popular solo streamers of Minecraft
L.C.S. is the official channel of League of Legends tournaments, developed by Riot Games

Who had the smallest gap between peak and average viewership?

Consistency is the key.

None of the top-rated channels that we saw until now featured at the top of our consistency parameter.
A potential reason for an ASMR channel significantly outperforming others here is that gaming channels usually peak around significant tournaments, big news breaks, controversies, etc.
Japan_ASMR is way simpler. It plays audio against static image background with the sole purpose of being soothing.

However, there is a catch…

The above channels, albeit similar in peak and average viewers, score far less on avg_viewer_pc and peak_viewer_pc .
The reason for ASMR content is understandable; it is not the adrenaline-filled rush of shooter games. Loeya and DanucD are female streamers, the former ranking as the top-ranked female Fornite player globally.
There are two probable reasons: a different target audience and a skewed gender ratio in the gaming arena.

Which channels scored the highest peak viewer-to-follower ratio?

Numbers don’t lie, or do they?

Sinner666 has had the highest peak viewers by far. Interesting to note, Sinner666’s channel doesn’t exist on Twitch anymore.
With a meteoric rise, followed by disappearance, the channel raises suspicions of pushing viewership by deploying bots.
JokerdTV is a player of World of Warcraft from Malta, an MMORPG genre.
Sigma is an American streamer who plays a wide variety of F.P.S. games.

A high peak viewership percentage can imply that viewers do not find the content quality to be consistent and do not follow.

Which channels scored the highest average viewer-to-follower ratio?

The loyal arrive daily.

Again we come across the fateful statistic of Sinner666.
The channel has a little less than three times the average viewer percentage from its closest rival, KEEMSTAR, and the two Dota2 channels.
This raises suspicions of malpractice, reinforced by the tremendous rise and disappearance.

Languages used by channels

The diverse languages are a testament to Twitch’s reach and globalization.

Languages beyond German have been clubbed together

As we can see from the above pie-chart

English is used by almost half of the channels.
Korea, Russian, French, and Portuguese have an approximately similar share.
The popularity of Spanish has been significantly contributed to by auronplay, as seen before.

Scope for further analysis

The five generated variables can be analyzed further for interesting insights, or even better analysis parameters can be generated from original data.
Mature , Partnered coupled with existing features can possibly explain trends elaborately.

Here’s the Jupyter Notebook with the code!