Update to popularity score algorithm
- Andrew Lee
- Posts: 3063
- Joined: Sat Feb 04, 2006 9:19 am
- Contact:
Update to popularity score algorithm
I have updated the algorithm for calculating the popularity score, which will now take the sum of activity score over the past 30 days for each app (instead of over its entire lifetime).
The impact will only be seen after 30 days, because the current score is used as the baseline for "day 1". This means that after 30 days, the baseline will be dropped and we will see the full impact of this change.
Expect to see some shakeup in the "Popular titles" listing, which will hopefully reflect the popularity of apps in a more dynamic manner.
The impact will only be seen after 30 days, because the current score is used as the baseline for "day 1". This means that after 30 days, the baseline will be dropped and we will see the full impact of this change.
Expect to see some shakeup in the "Popular titles" listing, which will hopefully reflect the popularity of apps in a more dynamic manner.
Re: Update to popularity score algorithm
Thank you! ![Smile :)](./images/smilies/icon_smile.gif)
![Smile :)](./images/smilies/icon_smile.gif)
My YouTube channel | Release date of my 13th playlist: August 24, 2020
Re: Update to popularity score algorithm
Sounds good -- thanks for working on that.
Re: Update to popularity score algorithm
Sounds good. Though I would use something that doesn't ignore old downloads but instead treats them with lower score. This would be even more dynamic and react when there's some boom yet wouldn't ignore booms that dropped just over month before.
Re: Update to popularity score algorithm
I don't know about this change. What if there is a app that is extremely popular, but has been around for a while, everyone would have voted for it, but all those votes would be older than 30 days.
You would need new people constantly voting for apps. Are there that many new members voting each month?
Universal Extractor for example, it's basically a one-of-a-kind app, how many people voted for it over the last 6 months, by month. Would it still be highly ranked with this new system?
You would need new people constantly voting for apps. Are there that many new members voting each month?
Universal Extractor for example, it's basically a one-of-a-kind app, how many people voted for it over the last 6 months, by month. Would it still be highly ranked with this new system?
Re: Update to popularity score algorithm
I think this is a great change and look forward to seeing the results...
Re: Update to popularity score algorithm
Should be interesting.
- Andrew Lee
- Posts: 3063
- Joined: Sat Feb 04, 2006 9:19 am
- Contact:
Re: Update to popularity score algorithm
That's an interesting idea. We will see how it goes, and make adjustments as we go along.Sounds good. Though I would use something that doesn't ignore old downloads but instead treats them with lower score. This would be even more dynamic and react when there's some boom yet wouldn't ignore booms that dropped just over month before.
One potential problem is some old apps have such huge scores, if we position a decreasing window over the data points, the old scores will still cast a huge shadow over the more current scores.
The popularity score is not based on members' votes, but activity from anonymous Internet browsers. Hence the whole idea of this change is to find out which apps are more popular on a more recent basis, rather than in its totality.I don't know about this change. What if there is a app that is extremely popular, but has been around for a while, everyone would have voted for it, but all those votes would be older than 30 days.
Re: Update to popularity score algorithm
hrm. Maybe im misunderstanding this, but im a little iffy on it also.
i think that the 'Popularity score' of each app, as voted by registered users should remain unchanged,
to me, this is like what the community likes and feels is worth having,
plus, itd be annoying if you have to like revote for your favorite apps to keep the score up.![Razz :P](./images/smilies/icon_razz.gif)
now as for the 'Popular titles' list on the site, im All for that changing based on like monthly activity,
(though, i like numbers, so i wouldnt mind a 'weekly' and 'yearly' list either,
)
i think where im getting confused, or have been confused is this:
i mean, i know it can add them to our fav apps list, but i though the whole voting bit was to publicly show how much we as a community liked each app? O_o
i can see changing the 'Popular titles' list to track like clicks on the 'Website - Screenshot - Download - Comments - Post comment - Permalink' links to track how popular something is, (though i must say whatever you do, apps might get a slight boost from just being on the front page,
)
But. Yeeeah. I think im confused here. ^-^;
...im also rather Ticked Off at HP right now, so that might be playing into it, But im posting anyways,![Razz :P](./images/smilies/icon_razz.gif)
Disclaimer: I honestly dont vote for stuff, as im sure Andy can tell,![Razz :P](./images/smilies/icon_razz.gif)
i mean theres stuff i like, alot, and use, but its just not my thing, unless i somehow feel its important, ^-^;;
So im just posting this in the interest of figuring out how this all works...
i think that the 'Popularity score' of each app, as voted by registered users should remain unchanged,
to me, this is like what the community likes and feels is worth having,
plus, itd be annoying if you have to like revote for your favorite apps to keep the score up.
![Razz :P](./images/smilies/icon_razz.gif)
now as for the 'Popular titles' list on the site, im All for that changing based on like monthly activity,
(though, i like numbers, so i wouldnt mind a 'weekly' and 'yearly' list either,
![Wink ;)](./images/smilies/icon_wink.gif)
i think where im getting confused, or have been confused is this:
Sooo. Why do we vote for apps again? O_oAndrew Lee wrote: The popularity score is not based on members' votes, but activity from anonymous Internet browsers.
i mean, i know it can add them to our fav apps list, but i though the whole voting bit was to publicly show how much we as a community liked each app? O_o
i can see changing the 'Popular titles' list to track like clicks on the 'Website - Screenshot - Download - Comments - Post comment - Permalink' links to track how popular something is, (though i must say whatever you do, apps might get a slight boost from just being on the front page,
![Razz :P](./images/smilies/icon_razz.gif)
But. Yeeeah. I think im confused here. ^-^;
...im also rather Ticked Off at HP right now, so that might be playing into it, But im posting anyways,
![Razz :P](./images/smilies/icon_razz.gif)
Disclaimer: I honestly dont vote for stuff, as im sure Andy can tell,
![Razz :P](./images/smilies/icon_razz.gif)
i mean theres stuff i like, alot, and use, but its just not my thing, unless i somehow feel its important, ^-^;;
So im just posting this in the interest of figuring out how this all works...
Re: Update to popularity score algorithm
There are two independent popularity scores.Firewrath wrote:hrm. Maybe im misunderstanding this, but im a little iffy on it also.
i think that the 'Popularity score' of each app, as voted by registered users should remain unchanged,
to me, this is like what the community likes and feels is worth having,
plus, itd be annoying if you have to like revote for your favorite apps to keep the score up.
now as for the 'Popular titles' list on the site, im All for that changing based on like monthly activity,
(though, i like numbers, so i wouldnt mind a 'weekly' and 'yearly' list either,)
i think where im getting confused, or have been confused is this:Sooo. Why do we vote for apps again? O_oAndrew Lee wrote: The popularity score is not based on members' votes, but activity from anonymous Internet browsers.
i mean, i know it can add them to our fav apps list, but i though the whole voting bit was to publicly show how much we as a community liked each app? O_o
i can see changing the 'Popular titles' list to track like clicks on the 'Website - Screenshot - Download - Comments - Post comment - Permalink' links to track how popular something is, (though i must say whatever you do, apps might get a slight boost from just being on the front page,)
But. Yeeeah. I think im confused here. ^-^;
...im also rather Ticked Off at HP right now, so that might be playing into it, But im posting anyways,
Disclaimer: I honestly dont vote for stuff, as im sure Andy can tell,
i mean theres stuff i like, alot, and use, but its just not my thing, unless i somehow feel its important, ^-^;;
So im just posting this in the interest of figuring out how this all works...
The internal popularity score is visible only if you are logged in and determines if the application is public or private.
The public popularity score is displayed to all visitors. When the visitor clicks the "Most popular first" link, the entries are ordered by the public score. Similarly, the "Popular titles" box displays the applications which have the highest public scores. (BTW, it might be interesting to see the list of entries with highest internal scores.
![Smile :)](./images/smilies/icon_smile.gif)
AFAIK, this change will affect only the public popularity score.
My YouTube channel | Release date of my 13th playlist: August 24, 2020
Re: Update to popularity score algorithm
It all depends on how do you set score reduction curve. Make it 1/exponential with halving period of one week and votes older than 2 months won't matter at all. The system is flexible.Andrew Lee wrote:That's an interesting idea. We will see how it goes, and make adjustments as we go along.Sounds good. Though I would use something that doesn't ignore old downloads but instead treats them with lower score. This would be even more dynamic and react when there's some boom yet wouldn't ignore booms that dropped just over month before.
One potential problem is some old apps have such huge scores, if we position a decreasing window over the data points, the old scores will still cast a huge shadow over the more current scores.
The popularity score is not based on members' votes, but activity from anonymous Internet browsers. Hence the whole idea of this change is to find out which apps are more popular on a more recent basis, rather than in its totality.I don't know about this change. What if there is a app that is extremely popular, but has been around for a while, everyone would have voted for it, but all those votes would be older than 30 days.
Re: Update to popularity score algorithm
What will happen to apps like "FastStone Capture"?Andrew Lee wrote:I have updated the algorithm for calculating the popularity score, which will now take the sum of activity score over the past 30 days for each app (instead of over its entire lifetime).
It's very popular, but who has it won't download it again ... and there will be no change/update (because it's shareware now).
That means: Very less activity
![Exclamation :!:](./images/smilies/icon_exclaim.gif)
Re: Update to popularity score algorithm
I suspect what's happened to this point is that FastStone Capture has gotten a great deal of attention due to the main site dropping it and PFW becoming its unofficial homepage. Its presence at #1 on the front page has only kept this going artificially, even though there are better capture tools here on the site. In essence, it really isn't the most popular program on the site anymore. It only stays in that position because it got a LOT of votes when we became its homepage.Checker wrote:What will happen to apps like "FastStone Capture"? It's very popular, but who has it won't download it again ... and there will be no change/update (because it's shareware now).
This situation isn't unique to FastStone Capture; a program's presence as a link on any popular site will drive a lot of traffic and a lot of votes that don't really represent actual popularity. Hopefully Andrew's new method will work a bit better or at least mix up the site content.
- Andrew Lee
- Posts: 3063
- Joined: Sat Feb 04, 2006 9:19 am
- Contact:
Re: Update to popularity score algorithm
Voting by logged in members is to help elevate private entries to public status. For logged in members, the popularity score for each entry is displayed as (ppp/nnn), where nnn is the voting score by members. Once the score is >= 10, the entry is public. Also, nnn is NOT DISPLAYED for non-members.Firewrath wrote:Sooo. Why do we vote for apps again?
ppp, on the other hand, aggregates public/member activity for each entry. It is added for each entry every time a browser is interested enough to click through to the app's website, or click through to the download link (with duplicate elimination by source IP address). It does this for both public and member clicks, but the public scores will tend to overwhelm member scores thanks to redirects from search engines.
I think webfork has essentially summarized the intent for this change. It is to make the top app list more dynamically reflect recent activity because it is very difficult to unseat older apps due to time advantage.
The scores are not deleted, so we can tinker with the ranking algo any time we want. Applying a moving window to the scores is definitely possible.
I will report back on 6 Sep (20 days after the mod) the list of top apps and their scores. Then we still have 10 days to make any changes if we want.
Re: Update to popularity score algorithm
maybe 2 lists: "recent popularity" and "all-time-favourites"?