In Figure 1b, we plot the average u

In Figure 1b, we plot the average user rating score against the number of user ratings. The logarithm of the number of ratings has a long-tail distribution, which shows that most applications have very few ratings whereas only a few applications have more than 105 ratings. Except for two significant peaks at scores of 1 and 5, the distribution of average ratings follows a Gaussian distribution centered at a score of 4. The peaks are due to the large number of applications that only have a few ratings, all of which are either 1 or 5 stars.
By examining the relationship between average score and the number of ratings, we can
see that applications with higher scores tend to have more ratings. This suggests that high quality applications are downloaded and rated more often than low-quality applications. Additionally, extreme average scores (1 or 5) are only obtained for applications with a small number of ratings, typically less than ten. These two observations indicate that the score alone is an insufficient measure of quality; one also must take the number of ratings into account.