Outlier Detection

Talk Discussion in 'BeerAdvocate Talk' started by kscaldef, Dec 17, 2014.

Thread Status:
Not open for further replies.
  1. kscaldef

    kscaldef Initiate (0) Jun 11, 2010 Oregon

    I understand you probably don't want to reveal all the details of your algorithms, but I really hope this takes into account some examination of the individual user's history (personal average rating, and perhaps number of ratings). With the average rating in the database sitting at 3.74, this seems like it has a strong possibility of penalizing those of us who try to rate closer to a 3 average. I feel like I shouldn't need to justify giving a beer a 2 any more or less than I have to justify giving a 4, but I'm pretty sure with this change I'm going to have to.
     
  2. kevanb

    kevanb Pooh-Bah (2,705) Apr 4, 2011 Illinois
    Pooh-Bah Trader


    Hopefully the bros can correct me if I am wrong, but it would seem that, going forward, a review is required for any statistical outliers. So if you are, and have been, writing out full reviews, nothing will change. What I take away from that post is that old quick ratings and precise ratings that are statistical outliers, will be discarded from the overall score of the beer.
     
  3. Todd

    Todd Founder (13,518) Aug 23, 1996 Finland
    STAFF Mod Team Society Pooh-Bah

    Correct. We will not be revealing our algorithm or future updates to it. And your assumption that it penalizes those who rate close to a 3 average is not accurate.
     
    kscaldef likes this.
  4. Todd

    Todd Founder (13,518) Aug 23, 1996 Finland
    STAFF Mod Team Society Pooh-Bah

    Yes, and no. If you're a reviewer, you have nothing to worry about. Carry on. But we have no plans of excluding past ratings from the overall score. That wasn't mentioned anywhere.
     
    kevanb likes this.
Thread Status:
Not open for further replies.