I have previously recommended trusting others’ experiences over your own expectations because the preponderance of the data suggests that your expectations are about as high quality as you would expect from a social primate born into an information age. I have recently been going through the many games that have come with Humble Bundles and Steam packs, and I am finding Metacritic really valuable.
Example: Guardians of Middle-earth. “I like Tolkien, that sounds interesting. I missed the LotR games that came out when the movies did.” Click from Library to Store page to see what it’s all about. “Hmm, a MOBA. I don’t know if I need another of those, but LotR DotA could be…” Metacritic: 56%. “Holy crap.”
A useful application of Bayes’ theorem is that sometimes an indicator in one direction means a lot more than an indicator in the other direction. There are classic examples, but gaming journalism provides another: negative reviews mean a lot more than positive reviews, particularly for major studio releases. High scores are the default and low scores on advertisers tend to be rare and dangerous. Games with bugs that would format your hard drive have gotten scores on the standard 7-9 scale. So a 56%? Holy crap. Okay, next game…
Update: commenters mention going beyond scores. Agreed.