Research about viewbots on Twitch
What is a viewbot?
Let's define a viewbot, beyond absolute simplicity and also talk about what some streamers use it for
Viewbots are automated computer processes that pretend to be real users. We all know that on instagram / twitter there are fake followers, let's say the equivalent of fake followers on twitch are viewbots.These viewbots serve to increase the number of views that our stream reflects that it has. This can attract new users, simple logic, if in a stream there are 5 people watching it and in another there are 100, our most basic mind dictates that the streamer with 100 views is better.
What do streamers use viewbots for?
In the first place they are used to raise the status of the stream: "If many people watch me, I am better, more interesting, a better player". We could say that it is not only for "status", it can also have traces of a big ego
It is also used to position the stream. If we have 5 viewers, people who potentially want to find us through the "explore" function will have to spend a period of time scrolling down. In a category like Valorant, perhaps to find 5-view streamers we have to scroll for 4 minutes.
If we have 500 or 800 views, probably people can find us without scrolling. This means that people who would never have found us, can find us without effort, because we have paid for the positioning with viewbots.
Summary: Viewbots help us to better position our stream, making new people who should never find us, find us.
How do viewbots work?
We do not go into technical details of viewbots from a developer's point of view, to avoid this article being used as a guide on how to create these bots
Viewbots are usually hired through web portals. These give us a dashboard in which we can decide different modes: chatbots, viewbots or followers
The most used are viewbots. We can simply enter a number n of viewbots that we want to be on any twitch channel, it doesn't have to be ours, an important detail.
These viewbots usually simulate a user that is not logged in, therefore when we click on our chat to see the list of users, it may be the case that we have 50 total views and in our list of users, we only find 40 registered users.
Simple mathematics 40 - 50 = 10, we could say that we have 10 users that we do not know what they are, they could be viewbots, views that come from a player embedded in a web or from users not logged in.
How can we determine which of these three types of "unidentified views" are those 10 users? With data, analyzing how they behave.
Let's look at data from a professional esports player in Spain, who has never made a stream before. It is logical that a profile like this is seen by many people who are not logged in to twitch, since their audience did not know the platform.
If we look at the behavior of unidentified viewers, it is exactly the same as that of real views, but on a smaller scale.
This is an example of a stream that is being visited by legitimate people and a large mass not logged into twitch.
With this in mind, let's look at another stream and see how our real views behave compared to unidentified views.
What conclusions do we draw from this data? The real views behave as they should, we see an upward spike given by a raid. But when we look at the unidentified views ... we see erratic behavior, behavior that does not represent the same as the real people we already have.
Let's zoom in on the data a bit to see it in much more detail
Aappreciate how real viewers describe a predictable trend, with no ups and downs. While our unindentified views, they jump from 100 views from in one minute
Let's look at more data from "suspicious user", data from a different day, to see if we determine any repeating patterns, or anything that gives us a clue.
What we see today from our suspicious user is absolute chaos in unidentified views, and the expected calm of our real views. Do we have to think that our unidentified views are real people who naturally have erratic behavior? all at the same time? Let's zoom in on our data again, to see chaos in all its glory
We see that the line of total views is absolutely manipulated by the oscillation of unidentified views, and our real views act as a base, super stable and barely oscillating over time, which if it effectively describes a human pattern, does not as our unidentifed views
Let's see more data about our suspicious user and then let's see legitimate users, with similar numbers, playing the same game in the Spanish community
Now lets see the legitimate data
About data collection
The suspicious user and the legitimate user are Spanish content creators of the Valorant game, all data is taken at 1 minute intervals. The data collection times are from 10:00 a.m. to 2:00 a.m.
One of the main people investigated has been banned by twitch indefinitely and I continue to collect data in bulk in order to draw better conclusions