The current study evaluated the degree to which novice visual analysts could discern trends in simulated time-series data across differing levels of variability and extreme values. Forty-five novice visual analysts were trained in general principles of visual analysis. One group received brief training on how to identify and omit extreme values. Participants rated 72 continuous time-series graphs. Inferential analyses were used to estimate the probability of correct responses. Participants who received the additional training were more likely to correctly identify intervention effects across all conditions. Nevertheless, extreme values had a substantial impact on decision accuracy for all participants. The impact of extreme values was exacerbated by increases in overall variability. Results support the notion that automated trend lines are useful but not infallible when interpreting continuous time-series data. Implications for practice and avenues for future research are discussed.