Changes to the algorithm and what's considered an active beer

The minimum (min) number of ratings required for an active beer to be scored and ranked has been 10 ratings for ages and the reality is that's too high given our current userbase, its activity, and consumer and industry changes over the last 5-10 years. On top of this, the threshold for what's considered active (rated within the last 2 years and not retired) is clearly too low. Combined, this hasn't been giving us an enough data to play with.

So, what's changing?

After proposing several ideas to the Mod Team, Pooh-Bahs, and Society members, I've decided to apply the following changes based on the feedback so far.

• Expand what's considered active:
  • Beers rated within the last 5 years (up from 2 years) that aren't retired (discontinued by the brewery).
  • This will enhance top rated and brewery lists (active and inactive tabs).
• Remove the minimum (min) as a requirement to be scored/ranked:
  • All active beers that have been rated will be scored/ranked.
  • The min will still be used within the algorithm that's used to rank beers for top rated lists and to generate the BeerAdvocate Score.
• Increase the algorithmic min:
  • After a ton of testing, increasing the algorithmic min from 10 to 20 should cover our needs and we can reevaluate as needed.
  • Overall, this will strengthen active top beer lists, while improving discovery on less active lists by allowing beers with a lower number of ratings to break through until this list becomes more active. Those beers will also be subjected to a considerably lower weighted rank compared to their averages.
  • Example: a beer with a 5.0 average and only one rating might have a weighted rank of 3.94+, as it's being pulled much closer toward the average across all beers.
• Add a scaling factor when under the min:
  • For beers under the min, we slightly reduce the beer's weighted rank based on the difference between the min and the number of ratings the beer has. This is experimental and may adjust over time based on feedback and further testing, but it should help soften the impact of beers with only a handful of ratings.
  • Example: that beer with 5.0 and only one rating now has an even lower weighted rank, because there's a 19 rating difference.
• Include new beers with 0 ratings on active lists:
  • Specifically, the "active" tab location on brewery pages for better discovery and to help reduce the chance of duplicate entries.

Other ideas are still being discussed, but this is a solid step forward. And the end result will increase the number of ranked beers from over 26k to over 253k, while also increasing the overall discovery of beers.

Updates to support these changes are rolling out now…

Questions? Feedback? Let me know.

Thanks!

 

No problem. I’ll pull this together later this week.

 

Update: There was a bug with calcluating the number of inactive beers. Fixes have been applied and numbers will update in a couple of hours.

 

Update: For beers under the min, we slightly reduce the beer's weighted rank based on the difference between the min and the number of ratings the beer has. This is experimental and may adjust over time based on feedback and further testing, but it should help soften the impact of beers with only a handful of ratings.

Rolling out now and should be complete in less than an hour.

 

I took care of this, but submitting a report for entries like this allows the mod team to quickly retire beers that are discontinued/no longer brewed.