Problem definition: We study a service system where true service times are unknown and customers exhibit the cognitive bias of being overconfident (in particular, overprecise) in their beliefs about service times— customers underestimate the variability of service times. Academic/practical relevance: Our models and results generalize those in several seminal queueing-game papers. We also provide overconfidence theory as a plausible explanation for the common long-queue phenomenon. Methodology: We formulate the problem as a stylized queueing model to examine the implications of overconfidence in service systems. Results: First, in an unobservable queue, the revenue-maximizing price is strictly higher than the welfare-maximizing price and consumer surplus is always negative. However, in an observable queue, consumer surplus is always negative for sufficiently small and sufficiently large system loads, but is positive (negative) for intermediate system loads when Ve (Ve := Rμ/c) is sufficiently large (small). Second, the revenue-maximizing price is strictly increasing in overconfidence and thus always higher than in the classical model in an unobservable queue, while it can be non-monotonic in overconfidence, being higher or lower than in the classical model in an observable queue. Investigating the impact of queue length information, we show that revealing queue length information always improves revenue for both sufficiently small and sufficiently large system loads, but increases (decreases) revenue for intermediate system loads when Ve is sufficiently small (large). Further, revealing queue length information always decreases consumer surplus for sufficiently small system loads, but decreases consumer surplus for sufficiently large system loads if and only if Ve is sufficiently small. Managerial implications: This paper unravels the role of overconfidence in service systems and its important implications on the manager’s pricing decision and queue-length-information provision policy, as well as consumer surplus.