Homework Help BlogTips that A+ students use to get ahead…
|
Tutor ratings are one of the best features of UniversityTutor – they allow parents and students to quickly go through a large number of tutors and find the ones with the most experience.
However, the rating system was always somewhat basic and just used a simple average of all ratings to sort tutors.
In practice this led to some less-than-optimal results. For example….
To solve this I started investigating smarter ranking algorithms and eventually settled on this Bayesian rating system.
You can check out the details on that page, but the essence of it is that it you can become more confident about the rating when you have more votes. So a tutor with few ratings will have their average rating discounted slightly (toward the site wide average), whereas you don’t need to alter the value as much for a tutor with more votes.
As you can see it leads to better results and fixes the two problems mentioned above.
Hopefully the new rating system gives some better results.
Until next time, happy tutoring!
Brian Armstrong
![]() |
A free copy of "10 Ways To Raise Your Grades By Studying Smarter, Not Harder" Instantly delivered to your inbox! |
|
Or Subscribe by RSS → |
|
3 Responses
Christian Stapfer
10|Jan|2011 1A blasphemous thought: How come you can rate a tutor on a *linear* scale in the first place?
For example, a tutor might be friendly, competent, willing to do the homework for the student (I don’t know whether that’s a good thing to do, but let that pass), et cetera, et cetera to *different* degrees.
Worse still, a tutor that might be really bad for some students might be excellent for others. This happens, in my experience. – I fear, therefore, that one cannot rate a tutor with sufficient precision *without* knowing something about the student he is expected to tutor…
[Reply]
Brian Armstrong Reply:
January 11th, 2011 at 7:12 am
Hi Christian,
I think you’re right a personalized rating system is even better. This is a bit like what NetFlix uses to recommend movies. To make something like that even somewhat accurate it requires more data points about what tutors (or movies) you liked in the past. And unfortunately we don’t have that yet. In the mean time, I think a rating system like the one in the article is still far superior to having none at all. But you’re right it is a matter of degree. Thanks!
Brian
[Reply]
ödev
13|Jun|2011 2I think you’re right a personalized rating system is even better. This is a bit like what NetFlix uses to recommend movies. To make something like that even somewhat accurate it requires more data points about what tutors (or movies) you liked in the past. And unfortunately we don’t have that yet. In the mean time, I think a rating system like the one in the article is still far superior to having none at all. But you’re right it is a matter of degree. Thanks!
[Reply]
Leave a reply