Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

frequency of list link clicks #10

Closed
collinbarrett opened this issue Jan 23, 2017 · 6 comments
Closed

frequency of list link clicks #10

collinbarrett opened this issue Jan 23, 2017 · 6 comments
Labels
wontfix will not be worked on

Comments

@collinbarrett
Copy link
Owner

collinbarrett commented Jan 23, 2017

Use the frequency of clicks of the View/Add links (unique per visitor per day) from FilterLists to calculate the FilScore.

This data is collected from our anonymized, self-hosted Piwik installation.

@collinbarrett collinbarrett added this to the v3.0 milestone Jan 23, 2017
@collinbarrett collinbarrett changed the title Determine a list popularity rating Calculate a list popularity rating Jan 23, 2017
@collinbarrett collinbarrett changed the title Calculate a list popularity rating Calculate a list score Jan 31, 2017
@collinbarrett
Copy link
Owner Author

collinbarrett commented Feb 1, 2017

Today, I am rolling out the first iteration of this. FilScore is currently calculated as a normalized 0-100 value (100 being the most popular in comparison with other lists of its type) based on the number of unique daily Views/Adds of the list.

Glaring Bug:

Currently, popularity is only tracked by Piwik "downloads" which refer to the "View" and possibly the "Add" links which typically lead to a .txt file. Lists that are not a "download" type (having an extension like .txt) are currently ignored (but shouldn't be).

This should be fixed ASAP as follows. It is possibly as simple as replacing Actions.getDownload with Actions.getOutlink. Alternatively, clicks of any links for a given list should be tagged as the list name in Piwik so that "popularity/interest" in FilScore calculations is determined by a visitor clicking on any of the list's links rather than just view/add. Once this implementation is in place, it should be slowly weened in over a 30-day period.

Other Concerns:

This scoring method is prone to false manipulation as well as false skewing (lists defaulting at the top get clicked more by new visitors not necessarily interested in any given list). But, these are issues I hope we can improve upon over time by implementing the other potential factors outlined in the OP.

@collinbarrett collinbarrett changed the title Calculate a list score Calculate a list score (FilScore) Feb 1, 2017
collinbarrett added a commit that referenced this issue Feb 1, 2017
@collinbarrett
Copy link
Owner Author

collinbarrett commented Feb 1, 2017

Update to "Glaring Bug" mentioned in last comment:
Now, if Piwik returns null for the getDownload API call, try a getOutlink api call. This gets the unique click counts for lists that are not recognized as a download (i.e., don't have an extension of .txt). This code is ugly and will need to get refactored later, but is now deployed to production.

It would still be good to track clicks on any link related to a list rather than just the View/Add links.

@collinbarrett collinbarrett changed the title Calculate a list score (FilScore) FilScore: Add frequency View/Add link clicks Feb 2, 2017
@collinbarrett collinbarrett changed the title FilScore: Add frequency View/Add link clicks FilScore: Add frequency of View/Add link clicks Feb 2, 2017
@collinbarrett collinbarrett changed the title FilScore: Add frequency of View/Add link clicks Incorporate frequency of View/Add link clicks Feb 2, 2017
@collinbarrett collinbarrett changed the title Incorporate frequency of View/Add link clicks Frequency of View/Add link clicks Feb 2, 2017
@collinbarrett collinbarrett modified the milestone: v2.1 Feb 2, 2017
@collinbarrett
Copy link
Owner Author

Re-opening and re-naming.

Clicks of only View/Add links are heavily skewed because many software tools come with default subscriptions. A uBlock Origin user, for example, would likely never click EasyList's View or Add links even though it is one of the top lists because it is already built into UBO.

To smooth out this skew somewhat, it'd be good to track if a unique user clicks any of the links for a given list (signifying some level of interest) rather than just the View or Add links. Clicking multiple links for the same list (or the same link multiple times) by the same visitor in the same day should be ignored (capped at 1 click).

This still won't be perfect, but it would provide a better indication of visitor interest in a list.

@collinbarrett collinbarrett reopened this Feb 3, 2017
@collinbarrett collinbarrett changed the title Frequency of View/Add link clicks Frequency of list link clicks Feb 3, 2017
collinbarrett added a commit that referenced this issue Feb 4, 2017
related to #10, after 30ish days, it’ll no longer be necessary to make
the second set of api calls for outlinks when calculating the FilScore
@collinbarrett
Copy link
Owner Author

Won't expand to other links besides View/Add for now. Adding other factors to the FilScore calculation is more important and will be more effective.

@collinbarrett collinbarrett self-assigned this Feb 9, 2017
@collinbarrett collinbarrett removed their assignment Oct 23, 2017
@collinbarrett
Copy link
Owner Author

re-opening for v3

@collinbarrett collinbarrett reopened this Aug 13, 2018
@collinbarrett collinbarrett changed the title Frequency of list link clicks frequency of list link clicks Aug 13, 2018
@collinbarrett
Copy link
Owner Author

No click-tracking has been implemented in v3. The idea of "tracking" is certainly contrary to the point of this site, no matter how anonymous it is.

In addition, this could be a self-fulfilling data point as lists which have a higher filscore will naturally get clicked on more, thus increasing their filscore.

Closing as won't implement for now.

@collinbarrett collinbarrett added the wontfix will not be worked on label Aug 13, 2018
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
wontfix will not be worked on
Projects
None yet
Development

No branches or pull requests

1 participant