Statistics Levels Statistics - From June to October 2016

@vince The problem I see in the quartile method is not about weights. It's about considering rates and ratings separately. As shown in the diagrams, levels with scores of 6.00 or 8.00 can get into top 30. However, if a level get only one rate of 5 stars, it's already in the first quartile for the rating, and if rating's weight is 4, then it will already get 4*4=16.00 points. Worse yet, if a level has 20 rates of 1 star, it would be in the first quartile for rates and thus earning a score it doesn't deserve. The diagrams here don't confirm my theory though, and I'm still confused how the scores are calculated.

 

@richardfu_ We are looking at the numbers now of the examples you pointed out, you have a good eye, it does not look right to me either. Levels with almost exact same rates and ratings should have similar scores, not a difference of 20 points.

 

By the way, although I personally don't think "views" are a relevant factor, it wouldn't be hard to add it into the log formula if needed:

score = rating x log2(rates+2) + 0.2 x likes + log10(views)

 

Don't forget to vote, everyone!

 

Thanks @Gepeto for all the hard work put into this. I think this will be a great feature going forward. And thanks to @richardfu_ for the alternate suggestion, we had two viable methods but I think in the end the simpler one will be better.