Restart most-downloaded counter every 1 month on repository


commander-beef

Very Active Member
Joined
May 1, 2012
Messages
971
Location
Polandowo
It seems that most downloaded counter is not very accurate to which new application was most downloaded, so i think there should be a reset counter function every month to see what is on top in currently fresh apps..


Why current idea of most downloaded - apps is bad: the more time app is in repo, the more download gets: and if You uploaded a game [even real garbage] in 2010 you can be assured that it gets more "most downloaded" than any other fresh game in 2012 since average download of fresh app/game is about 20 downloads in first week, and 2-3 in next weeks..
 
I do not agree that even garbage will make it up in the most downloaded list if you just wait long enough, but most websites offer to view the most downloaded of all times, last month, or this week. Same for "top rated", where only votes from that time period are counted. Also having this in each category could make sense.


These kind of statistics can still be very misleading if you try to learn something about the quality of the app, though.
 
Maybe have a 'download frequency' counter instead - displaying average downloads per month
 
What is flawed in the download counter is that it is counting redownload.


So If I upload a small PND, and upload a small update every month, I can easily reach the largest number of download as many users will update (using PNDstore or else)


The download counter dont represent the number of users...
 
Last edited by a moderator:
Yes, but if download counter will be restarted every month , this "trick" will not work..
Sure, but then, most my apps would have a 0 download count this month even if it have many dayly users (like audacious and such) which wouldnt be fair.


I would rather propose to only show download of the lastest version ;) which will work around the trick yet show a number closer to the number of real users ;)
 
@commander-beef - ridiculous idea IMHO


I know you have a record of throwing the dummy out of the pram when things don't go your way, but the best way to show the popularity of any app is how many times it's been downloaded.


@sebt3 - your idea of an option to show the download count of the latest version is probably the fairest way but it should be as a filter. Remember that not everyone will download the latest version. Especially beta versions.


IMHO the system as it stands works well.
 
.. but the best way to show the popularity of any app is how many times it's been downloaded.

Dont be mad, but this is most ridiculous thing i heard from long time, because of

So If I upload a small PND, and upload a small update every month, I can easily reach the largest number of download as many users will update (using PNDstore or else)

And downloading counter can show popularity only in small percent.


If you want to have really good info about quality of apps you need to consider: app rates + downloads + [especially] comments + [maybe] searching behavior of users on repo + external info from google..
 
Last edited by a moderator:
Don't want to get into an argument here but the metrics your talking about include subjectives -comments, search behaviors and external sources.


Resetting the download counter would not effect these to any great degree.
 
And also, rather than having a fixed reset date every month, which would hurt the new submissions a bit before the reset date, simply have a measure of the download count in the past 30 days or so.
 
It seems to me that except in a few rare circumstances (i.e. extremely infamously bad things, like this), number of downloads tends to be a lot more useful than most rating systems, since not everyone bothers to rate.


I'd have to agree with sebt3 in a sense. Record the number of downloads for each version separately, and display the biggest number out of those. The point of allowing people to order apps by view count is to be able to see what people are frequently downloading (and, therefore, approximately what is popular), and resetting the download counter regularly wouldn't help people know what's popular at all. The result would be current "exciting" or novel new programs being listed as popular rather than programs that have been consistently downloaded by users in the long-term. On the other hand, if each version had its own counter, the number of downloads would still be a long-term representation while eliminating any advantage a program gains by being updated more frequently.
 
Some kind of "trending" list could be useful. It could be implemented as simply as "most downloads in the last 30 days" or so, like sverm suggests. The "most downloaded" list still works nicely as a list of essential downloads for new users. Could the repo spec be augmented with download counts too? :p
 
at the scale of users we have, I think the current system works well. There's not an insane amount of apps to filter through and not a whole lot of garbage apps.


for things like the iOS and Android, there is a timeframe for displaying most popular. A problem was with android at least is that the original kings of downloads remained at the top of the list regardless of it being outdated and basic yet being the first to arrive. Lots of people re-download the several year old app thinking it's good when in reality it is just old.


they fixed this recently and you are displayed "top apps" based off some revolving timeline likely 3-6 months. But like is mentioned here... with the maximum number of users here being <10,000 it's hard to accurately guage who's downloading what and quality of app based off downloads.


average app rating and # of ratings should determine the order of what is displayed the highest IMO. With the size of our community, I would suggest you use that.
 
Last edited by a moderator:
The best way of tracking usefulness is by tracking the amount of users still using the app, and how much/long they do. That'd require the OS the track and (anonymously ofc...) upload that.


You could also show the downloads/usages per day/week/month and/or use a moving average, by lowering the amount by 1% every day. That gets rid of old and stale stuff.
 
Last edited by a moderator:
The best way of tracking usefulness is by tracking the amount of users still using the app, and how much/long they do. That'd require the OS the track and (anonymously ofc...) upload that.

That would be spyware and it would also be bloat. We don't need that (and I personally don't want that; Facebook and Google are tracking me enough). If it could be easily disabled (and the OS made it clear that this was happening and how to disable it), it wouldn't be spyware anymore, but then the system would be useless.


I think having a counter for the current month/day/year/whatever would be a perfectly good idea as long as it supplements, rather than replacing, the current "all-time" download counter.
 
I think having a counter for the current month/day/year/whatever would be a perfectly good idea as long as it supplements, rather than replacing, the current "all-time" download counter.

I think that makes most sense. 3 or so counters for each: all time/last 30 days/last 7 days
 
It would be best to focus on having ratings matter, not downloads.


- for each item


- get the number of ratings newer than x months


- calculate the average number of ratings


- for all items of number of ratings x or greater


- list the best x
 
I propose that we use number of ratings - not the actual ratings themselves, but the number that people actually bothered to log in and rate. You could then have a "breakdown" treeview of actual ratings. Number of downloads is, as Streak says, not a good metric. One uploader (who isn't really doing any actual porting, just building) might upload twenty really crappy terminal-style apps in a short space of time, but refrain from giving a reasonable description and gather a large amount of downloads simply by people grabbing it to see what it's about. If we could deter people from doing that, then the pandora's repository would grow with much more high-quality software - decent interfaces, emulators that run at reasonable speeds, that sort of thing.


D.
 
Back
Top