Currently, this bot just subtracts yesterday’s subscriber count from today’s, and give a percentage difference.
Current Algo: percentage difference in subscribers, averaged over the past 7 days. Where it’s showing less than 7 days, it’s because I don’t yet have the data (mostly because this bot hasn’t run for 7 days yet, but also because either it’s a new community, or because I had to fudge things a bit in the first few days by excluding communities with less 100 subscribers)
Edit to add: since the posts from previous days will always be available, the first filter for each new day is now how much a Community has grown in that one day. That is then added to earlier day’s, and averaged out.
Previously, it was just showing the rolling-average for every Community, but that meant that sites that had had crazy growth in one day but had now plateaued were still showing up and the list wouldn’t change until that aberration had aged out.
Please use this post if you’d like share better ideas of what constitutes “trending”, and if you have any preferences for how the posts are displayed. Thanks.
I think splitting the post into sections dedicated to specific metrics might help. You could keep WIP metrics live but as different sections. For example…
Top N By…
Growth Percentage: <this would be your current metric> … … …
Absolute Number of New Subscribers: <eg. trendingcommunites +5000 new subscribers> … … …
Trending Formula (WIP) (Link to post detailing formula): <this is the formula talked about in a different comment> … … …
Etc.
I think that grabbing the easy metrics and offering them up initially will be more beneficial to growing your own subscriber count than trying to work out a clever way to do just the one thing.
Love the idea though, I’ve posted about this community in a couple places and I’m thrilled to see that it’s taking off a bit. Great job!
Thank you - for your support as well for for your suggestions.
I’m realising that, given the raw data, there’s more I could be doing with it, both in terms of how accurate the results are, as well as how they are displayed.
Now that I’ve got a few days worth, I’ll work on smoothing the daily bumps, to help distinguish between a genuine trend and incidental hiccup.
After that, I’ll look into breaking things up - using different formulas to show different things (e.g. at the moment it’s probably impossible for a 30k subscriber count Community to ever show up here). There’s such a wide deviation in the data that it likely calls for it.