Steepster ratings
I’m really perplexed about the genesis of these “steepster” scores which can create unfortunate biases against certain teas. I just added a new tea from Taiwan Sourcing, reviewed it and gave it a 94. When I looked at the page minutes later, a mysterious 81 appeared next to my score of 94, even though I’m the only person who rated it. ?
This is the explanation behind the Steepster rating system. It doesn’t simple average posted numberical ratings, which is why a tea with few ratings (even at 100%) can have a much lower score than a popular tea with many varied ratings. It is to prevent “gaming” the system.
http://blog.steepster.com/post/247562587/brewing-a-better-rating-system-the-2nd-steep
I guess I buy the explanation for a tea that has been rated multiple times, but it’s still weird that a tea rated once at 94 can have a steepster rating of 81.
Thanks for the link!
Oh, the system has plenty of critics! I just wanted to explain so it made marginally more sense. :)
It works the other way too. A single rating of 50 would result in a rating of about 79. It appears that 80 is the “golden mean”.
The reason ratings are weighted even when there’s only one, is to prevent floods of single-rating 100’s to the “Best Teas” section. It prevents bots and less-than-moral tea companies from rigging the system with a few 100’s. Just like how the Best Teas page itself’s highest rating right now is only 91 (lower than your 94!), which a few teas have, and those teas are further sorted by number of tasting notes.
It aint perfect, but.
Besides, all of Sanne tea’s teas are not available now on Steepster now…..it is weird. I can’t find a place to edit it…:(
You can’t edit them. You need to email someone at Steepster with a link to your buy page next to a link to the Steepster listing. They will change the availability.
Thanks for the link. The problem though is that all tea rating variables are qualitative and subjective, not quantitative. Bayesian inference requires large populations parameters as a basic assumption which we never have in tea tasting for any tea, and Bayesian inference requires 0,1 binomial variables, which are success/failure, yes/no and not a range of possible outcomes.
I think most people here know to ignore the ratings and use the reviews as a guide for shopping.
Login or sign up to leave a comment.