Human mood varies according to both endogenous and situational factors. Analyzing patterns of online media engagement can provide additional insights into a third dimension of mood potentially involving time-of-day factors. Using data collected through the online brain-training app, Lumosity, we report real-world trends in mood expression among 500,000 respondents. Endorsements of positive mood by Lumosity gamers peaked midday, while endorsements of negative mood peaked overnight. These data suggest that there may be inherent bias towards negative mood when individuals are awake at night. More broadly, they suggest that internet data-trails have the potential to uncover fundamental aspects of human behavior.