How are the review ratings calculated?

Comments

• Cam_n_Stu UKMember Posts: 368

  November 2012

  Looking at http://www.brickset.com/detail/?Set=79003-1 the review summary says 3.8 (1 Review), but when you click on the review it's Huw 5* review!

  0
• leemcg Member Posts: 607

  November 2012

  I've seen Huw discuss this before (but I'd never be able to find the post).
  
  He uses a clever method to calculate the average, which gives a set with 20 5* reviews a better average than a set with only 1 5* review. He adds in another virtual review (I think with 3 stars), which will gradually get less and less weight as there are more reviews.

  0
• LostInTranslation UKMember Posts: 5,572

  November 2012

  I think this has been queried before - I believe it's because there's only one review and some kind of weighting is applied. I'll try and find the previous discussion.

  0
• Joseph Member Posts: 629

  November 2012 edited November 2012

  ^ & ^^ Here you go: http://www.bricksetforum.com/discussion/comment/27119/#Comment_27119

  Huw said:

  There's a weighting score, of 2.5, added which causes the number of ratings as well as the rating values themselves to influence the overall rating.
  
  So, 1 score of 4
  
  4 + 2.5 / 2 = 3.2
  
  1 score of 5
  
  5 + 2.5 / 2 = 3.75
  
  5 scores of 5
  
  5*5 + 2.5 / 6 = 4.5
  
  100 scores of 5
  
  100*5 + 2.5 / 101 = 4.97
  
  It stops a single score of 5 giving the same rating as a set that has 100 scores of 5.
  
  Clever eh? :)

  0
• Cam_n_Stu UKMember Posts: 368

  November 2012

  OK. I guess it is one way to do it, but it does result in misleading summary scores in these sort of instances. Also, as I wasn't aware of this weighting I've been taking summary scores at face value for a few sets. Thinking it was disappointing and not bothering to read the actual reviews!
  
  I think I prefer IMDB.com's approach which doesn't show a summary score at all until it has an specified number of reviews, 10 in their case I think.
  
  Surely it would be better to only apply the weighting when sorting sets on score rather than for each individual on-page summary score where you can see it only has 1 or 2 reviews and hence should treat the score with caution?

  0
• Huw Brickset Towers, Hampshire, UKAdministrator Posts: 7,059

  November 2012

  Thanks everyone. It may have some disadvantages but I think they are outweighed by the advantages and its simplicity.

  0
• BooTheMightyHamster Northern edge of London, just before the dragons...Member Posts: 1,504

  January 2013

  @Huw
  @atkinsar
  
  Sorry chaps, this probably ought to be in 'The Database', rather than 'The Forum'.
  
  If somebody could shift it, that would be great.
  
  Thanks,
  
  Boo

  0
• CCC UKMember Posts: 20,525

  January 2013 edited January 2013

  See here:
  
  http://www.bricksetforum.com/discussion/8005/review-summary-showing-3-8-out-of-5-on-set-page-when-only-5-review#latest
  
  It is to stop sets with just one review dominating the ratings.

  0
• BooTheMightyHamster Northern edge of London, just before the dragons...Member Posts: 1,504

  January 2013

  @CCC
  
  Ah... Clever stuff!
  Thanks for the explanation.

  0
• Huw Brickset Towers, Hampshire, UKAdministrator Posts: 7,059

  January 2013 edited January 2013

  This method was copied from the way it was done back in the day at Lugnet.
  
  It's very simple, but works very well in ensuring that sets with one rating don't skew the top rated sets table

  0