Just group by version. Time-based decay might work too but linking a rating to a specific version (maybe only a major version - there are pro & cons) has more meaning.
The date should not only be taken into account for app RATING, but for app RANKING. I see many apps that haven't been updated in over two years that are ranked higher (due to more downloads) than current releases similarly or higher rated with fewer downloads. I would favor an time-based exponential decay type of factoring of ranking. (<a href="http://en.wikipedia.org/wiki/Exponential_decay" rel="nofollow">http://en.wikipedia.org/wiki/Exponential_decay</a>)
Amazon does a similar thing to seller ratings: <a href="https://images-na.ssl-images-amazon.com/images/G/01/sellerfeedback/time-weighted-graph-2._V389181558_.png" rel="nofollow">https://images-na.ssl-images-amazon.com/images/G/01/sellerfe...</a> .<p>Ratings have an initial slow impact (this is more to do with getting the goods to the user and allowing the returns process to run it's course) then the feedback has a 100% affect that tails off over a year or two.<p>A similar approach could be used by google.
You can see a graph of the ratings over time here:<p><a href="http://www.appbrain.com/app/touristeye-travel-guide/com.touristeye" rel="nofollow">http://www.appbrain.com/app/touristeye-travel-guide/com.tour...</a><p>(in the "Changelog" tab)
Seems like common sense to me? If I want to know if an application is any good, I look at the ratings and if it has a lot of good ratings, I look at the latest comments to make sure that the lower ratings are for older versions.<p>Surely it'd be a bit more convenient for Google to take versions of apps into account, and additionally devices, too, if they can (a lot of times an app will be okay for most devices, but users of a certain device, different from my own, might have troubles), but it's -- in my experience -- not too much of a hassle.
A version/ratings graph might really help here. That will also dissuade devs from frequently releasing updates to hide the fact that their product is sub-par. If you have a whole bunch of recent releases with a handful of early 5-star ratings that are glowing with praise, something is fishy.
Ratings trends over time can go in the other direction too. I got bit by an update the other day on a mature and very popular app which just released a tragically flawed update. I over confidently accepted the update and then started reading the latest reviews which we're shouting in all caps not to get this update.<p>And what you said about excluding ppl who aren't on G+.
A well known tor hidden service weights customer feedback by age, spend history volume, and variety of orders.<p>I think this would work well for android apps. you might want to change spend history for number of apps downloaded but with higher rating for paid app downloads.
The problem is not the average rating. The problem is the faked ratings. I think it's not normal that a lot of ios applications has five starts. It cannot be that a new app that anybody knows about it has a lot of positive comments.
FTA: <i>We don’t have a bad average rating (3.98 with almost 2.900 ratings), but I’m pretty sure it doesn’t reflect the quality of the last version we launched.</i><p>Really? The whole thesis of this article is that app ratings are broken because you have a 4 out of 5?