Many different techniques and algorithms exist for visualizing data of multiple time-series. The performance of a visualization technique is affected by multiple factors, predominantly data characteristics and the tasks that needs to be solved. The aim of this thesis was comparing the readability between two of them - stream graphs and pixel-based techniques, and the task was determined to be peak detection. We conducted an online study with 12 participants, during which we measured the accuracy and respond time of task completion for each stimulus. The study result illustrates that pixel-based visualization is more suitable for peak detection, as participants made less errors and responded more quickly when extracting related information from pixel graphs. There are other interesting results worthy of being interpreted as well. In the end, we proposed what the study lacks and can be improved in the future.