Vídeos de K-Pop no Youtube

Disclaimer: This research was done using all our database and fine-grain monitoring for 3 months. It might loose precision with time, so from time to time the formula might be updated. Read the explanations below to better understand how it works. Read more about how every data is retrieved here

View Predictor

1. Relative viewership per age
AZUL são visualizações/dia HISTÓRICA, VERDE são visualizações/dia RECENTES
Números são relacionados à média mensal. Primeiro mês excluso.
Estes dados mostram a visualização atual dos vídeos de acordo com quanto tempo foram lançados, não quantas visitas receberam no passado. Todos os dados são dados do presente

2. Predictor analysis
Based on X where X is the number of views on the First Month, the following formula calculates how many views for each subsequent month are gathered:

V = X * 1.399 ((m+1) -0.743
This formula generates the red line at the above graphic and was derived from the points generated by the recent view/day (green)
Formula updated as of 2019-01-16

3. Sample video analysis
Video Data:
Views
CURRENT view/day
Date published

(Y-m-d)
Predict views for age

months

Understanding the mathematics

Graphic 1 gathers the top 15 videos released between 1 month and 60 months per month and analyses how the viewership falls the older a video is.

It does not display past data, it displays current data. For instance, a video released 12 months ago, displays the recent view/day and its historic views/day as of TODAY in the graphic, NOT the views that video had 12 month ago.

This means that a video that averaged, for example, 10m views/day on its first month, and the 60th month above is displaying 2%, should have 200k views/day when it reaches 60 month (5 years)

Therefore, this is a real-time display of how age disrupts viewership. For this reason, videos that have recently received a boost in views (like a comeback from the group, or some important link to it) might distort the graphic with a peak. That peak means a video, released on that time frame, is been viewed more than the average for videos that old

Since this graphic uses real-time data, every day it changes based on the viewership of the videos, which is interesting because it also take into account the current interest on past videos.

- The first data is about videos released starting the last day of the last month. Videos for the current month are not used because the view trending of the first 2 weeks works differently and would affect negativelly this graphic.
- Although this graphic have 60 data points, note that it does display one more month at the last line, that's because the first data is the 100% first month comparison.
- Numbers are in percentage from the top viewership (obviously, the first month)..
- A 45 day normalization is applied to reduce noise, otherwise famous high-profile videos or increased viewership for a comeback group would cause havok on the data (as you can see, there are still peaks in random places because of that).


The predictor analysis (2) uses the data from the first table to calculate the exponential approach for the graphic.
The formula was generated using Elsenaju.eu's Online Curve Fit Calculator using the plotted values from table 1

Sample video analysis (3) uses the predictor data to estimate views from a given video, where you insert the current view number and when the video was published, and at which age (in months) you wish to see the estimate viewership, and the system will predict. It works both ways, so it can predict past values as well. Since predicting data that already exists is prone to a definite error, needless to say if you request to see the views at "0 months" (the release day), it will probably not be 0 because this is an approximation.

Starting August 2018 we use this formula on the main Videos List showing predicted views for the 60th month (5 years), using the formula to estimate how many views a video had at its first month:
- Entries where the number of views are past 60 month are in green, but we only started recording it on August 2018, so videos released before August 2013 have an estimate past prediction for when they reached 60 months, while videos released after August 2013, which reached 60 month after we had this system in place, have the actual precise counter for 60 month
- Entries newer than 31 days cannot be calculated and are listed as "Too Soon" on the MV lists
- There are known fluctiations on the view distribution, as well known peaks (famous videos). Due to the nature of probabilistic mathematic and normalization in the formula, these do not affect the formula, however changes in user behaviour should be considered and therefore this formula will need to be kept updated at least every year.

Ads by google (click here to hide, consider tipping me to maintain the site)
PREPARING ...
Loading page...