Oblivion

Corrections to file stats

  • Comment
Hi folks,

It came to our attention last week, in a random forum post unrelated to the topic, that the unique file download stats have been broken for quite some time now. I honestly had no idea this was the case.

The unique download stat you see on file pages is supposed to tell you how many individual members have downloaded a file. Irrespective of how many times that user comes back to download that file (or multiple versions of a file on the same page) the unique download counter should only go up once per user who downloads from a file page. As an example, imagine a new user goes to the SkyUI file page for the first time. There are currently 14 files available for download for SkyUI. The first time the user downloads SkyUI the unique download counter will go up by 1. The second time the user downloads that file, or any other file on that page, the unique download counter should not increase by 1 again.

On top of that, each individual file uploaded to the file page (in SkyUI’s case, 14) has its own unique download counter as well. These are file specific. So if I download SkyUI 1.0 a total of 5 times, and SkyUI 2.0 a total of 5 times, then the unique download counter for SkyUI 1.0 will go up by only 1, the unique download counter for SkyUI 2.0 will go up by 1, and the unique download counter for the SkyUI file page as a whole will go up only by 1. In contrast, the total download counter for SkyUI 1.0 will go up by 5, the total download counter for SkyUI 2.0 will go up by 5, and the total download counter for the SkyUI file page as a whole will go up by 10.

As a result, you end up with two markedly different figures; unique downloads tells you how many individual members have downloaded the mod. Total downloads tells you how many times the file has been downloaded overall, unique or not. The disparity can go some way to showing you how many users like the mod enough to update it through multiple versions, among other things, though you take that with a pinch of salt, of course.

Users who download the file without being logged in (any files under 2MB can be downloaded without an account) do not count towards the unique download counter at any time, but will count towards the total downloads. This is because we cannot accurately track unique download statistics for non-logged in users due to the prevalence of dynamic IP addresses. As such, unique downloads are only based on registered members downloading files.

That is how it was supposed to work. That is how I actually thought it did work. However, that’s not how it was working up until today.

To explain how it was wrongly counting the figure before I’ll go back to the SkyUI example. Before, if I download SkyUI 1.0 a total of 5 times, and SkyUI 2.0 a total of 5 times, then the unique download counter for SkyUI 1.0 will go up by only 1, the unique download counter for SkyUI 2.0 will go up by 1, and the unique download counter for the SkyUI file page as a whole will go up by 2. As a result, if I were to download all 14 files on the SkyUI page then the unique download counter for the SkyUI file page as a whole would have gone up by 14.

This meant that up until now, the unique download counters for file pages were incorrect and not accurate. The unique download counters for the individual downloadable files themselves (e.g. SkyUI’s 14 individual files available from the “files” tab on the file page) were correct, just the overall total unique downloads for the file pages were wrong.

Over the past four days we’ve been running a script in the background to go through all 1.4 billion downloads we’ve logged to date in our file database and recalculate the correct unique download counters. We’ve also patched out the error in our calculations to ensure the correct counting method is used. The figures you now see on the site are the correct, fixed figures. Initial comparisons show a 30%-50% change downwards for most file page’s unique download counters.

I understand that it can be disheartening to log on to your file pages today to find your unique download counts revised down, a lot. I’m sorry about that. However, I’m sure everyone would rather the correct, accurate figures were shown rather than sticking with the incorrect figures.

65 comments

Comments locked

A moderator has closed this comment topic for the time being
  1. maxine24610
    maxine24610
    • premium
    • 0 kudos
    I dont know if this is the right place to post this but my downloads are missing .The last time i updated to nexus mod manager my mods disappeared .I was using 20 odd mods but have been reduced to 8 mods installed. The others have disappeared from my computer.
  2. NexBeth
    NexBeth
    • premium
    • 27 kudos


    Could you fix views now as well?

    1. Don't count the author
    2. Count only unique
    3. Ignore bots (preferably only logged in)


    i just started working on 1st mod page and noticed right off that I'm racking up the Views (I'm the only one viewing since the page is not published yet).  It probably doesn't matter that much but just wondered if this was supposed to work this way?
  3. unforbidable
    unforbidable
    • premium
    • 196 kudos
    Now I can see how many people downloaded a specific version compared to the number of all users who ever downloaded any of the versions if I understand correctly. The previous counter wasn't very useful and I can still reach it if I sum unique counters for all versions. It's still a shame some amount of downloads slip by if your file is below 2 MB but that cannot be helped.
  4. MrMason
    MrMason
    • Development Director
    • 12 kudos


    Not to be unpolite, but I think the counter (-s) are still broken somehow?

    One of my pages have the exact same frequency of Unique and Total Downloads as before the correction. At least that's my feeling.

    I notice this because I find it interesting how many different persons download my files and how many times in total, to see if some turn back a second time.

    Maybe I'm wrong, but I still think the Unique proportionally follows the same rate of the Total Downloads.


     
    Can you link to which Mod you're referring to please so we can investigate ? 
  5. Mushashi7
    Mushashi7
    • premium
    • 10 kudos
    Not to be unpolite, but I think the counter (-s) are still broken somehow?

    One of my pages have the exact same frequency of Unique and Total Downloads as before the correction. At least that's my feeling.

    I notice this because I find it interesting how many different persons download my files and how many times in total, to see if some turn back a second time.

    Maybe I'm wrong, but I still think the Unique proportionally follows the same rate of the Total Downloads.
  6. PatrickTheDM
    PatrickTheDM
    • premium
    • 174 kudos
    Just a thought, should the sorting for most downloaded now use this corrected stat as opposed to total downloads? IE. A mod that has many issues and is updated many times will jump queue over a mod that has no updates and appear to be more popular in the list until clicked on.
    I understand a mod that people are willing to come back and keep updating equals popularity. Maybe a sort option for total unique downloads or by ratio of endorsements to unique downloads would be good as that is a better gauge of a quality mod (in theory).
  7. willynexus
    willynexus
    • member
    • 0 kudos
    I cant be mad at you guys you bring me so much joy at so little a cost.
  8. Scrubster
    Scrubster
    • supporter
    • 0 kudos
    i'm more concerned with downloads being slow for me

    everything's fine when i download from other site, but mine's completely slow, tried the manual download, still the same
  9. geekminxen
    geekminxen
    • premium
    • 51 kudos
    Thank you for fixing it. As for the download count being lower, that just means my endorsement-to-download ratio is higher!
  10. hagyjalbeken
    hagyjalbeken
    • supporter
    • 154 kudos
    Wow, that makes me actually feel more content, but i encountered a small annoyance.

    I only just started all this modding thing, actually have only two puny more or less unknown mods,
    One of my mods has now less uniques than it should, because i accidentaly deleted the old version of the file when i uploaded a newer version some weeks ago, so the old one was not counted again for the uniques, only shows up for the old unchanged overall download numbers. Not that it bothers me much, it is just a small clumsy armour retexture with around 400 downloads, but maybe some people have the same problem with more popular mods. (I know that this probably can't be fixed, just saying.)

    But i am glad that it is fixed, it was bothering me. I couldn't figure out how on earth did my other mod have around 1800 unique downloads, when the the main file has only 1200 and all the other stuff i uploaded for it are body type patches. Human foolishness can be endless, but i was sure there are no 600 fools who only downloaded a single patch. Also i am notoriously bad at anything that is related to math, so it is good to know that the mistake was not on my end and i can maybe count more than my fingers.

    So thank you for fixing this
    1. Zaldiir
      Zaldiir
      • Moderator
      • 477 kudos
      Deleted files are counted as well.
    2. hagyjalbeken
      hagyjalbeken
      • supporter
      • 154 kudos
      Ok you are right, big facepalm, i think i counted something that i shouldn't have. But my first thought when i looked at the new numbers was that the endorsement-download ratio is now much better than it should and it's cheating (i mean more than 10% for a lousy retex with minimum downloads, what?) Nevermind, i'll think twice before i talk next time
    3. Zaldiir
      Zaldiir
      • Moderator
      • 477 kudos
      It's awesome to see a better endorsements/download ratio for all the mods out there.
    4. jeepshaker
      jeepshaker
      • supporter
      • 0 kudos
      if others are like me, having a better DL/Endorsed ratio will increase interest in a mod. always thought it was strange when even the best mods had only 10:1 endorsement ratios.