Average review scores are useless for predicting whether you'll like a game. Here's what to look at instead.
Average session length
HowLongToBeat data on median playtime tells you a lot about pacing. A game with a 4-hour median and 200-hour outliers is very different from one where everyone lands between 20 and 25 hours. Match this to your own session patterns.
Review sentiment vs. score
An 8/10 with negative Steam user tags around 'repetitive' is a different product than an 8/10 with 'great story' and 'relaxing'. The score is meaningless without the texture of the sentiment underneath it.
Refund rate window
Steam's 2-hour refund window is a natural filter. Games with high refund rates at launch signal something that hooks previews but doesn't hold. If a game gets refund-rate discourse in launch week, that's a real signal.
