Steam filters out key-based customer reviews from weighted score

Valve have made some changes to their customer review system on Steam, adding a set of filters and altering which reviews contribute to the overall community ‘score’ shown for a given game. From now on, reviews written by customers who purchased the game on Steam itself will be the only ones which affect the user score.

The justification for this is an internal study which revealed to Valve that “at least 160 titles have a substantially greater percentage of positive reviews by users that activated the product with a cd key, compared to customers that purchased the game directly on Steam”. There have been previous instances of ‘get a Steam key for my game in return for leaving a positive review’ and, worse, marketing services specifically set up to leave fake positive reviews in return for payment, so this is a legitimate problem.

Valve’s default approach to most problems of this type is to automate a wide-ranging solution (since they don’t have the individual man-power for a more nuanced system), and the unwitting victims this time around will be people who owned key-activated games and left legitimate reviews. You’ll still be able to read reviews left by those people, so it’s not all doom and gloom, but the reviews themselves won’t influence the overall score. It’s a necessary step, but taken in a crude manner.

It’s estimated that this step will change the review score categories for “about 14%” of games. These will mostly be titles on the cusp of one category or another being bumped up or down by a small percentage change.

The new filtering system added today allows you to quickly look at only positive (or only negative) reviews, view reviews from Steam purchasers vs key activations (or, as noted, all of them), and filter reviews for your primary language. You can change your language preferences in your account options.

At the end of the article, Valve say they’re continuing to look at how the customer review system is functioning (or malfunctioning, as the case may be). Their main points of interest are “titles where the most helpful reviews don’t seem to accurately match the general customer sentiment”, “titles where a small group of users are able to consistently mark specific reviews as helpful, and as a result can present a skewed perception of what customers are saying about the game”, and when “off-topic reviews get marked as ‘helpful’ simply because they are funny”.