Eye movements that are dependent on cognition hold promise in assessing sports-related concussions but research on reliability of eye tracking measurements in athletic cohorts is very limited. This observational test–retest study aimed to establish whether eye tracking technology is a reliable tool for assessing sports-related concussions in youth and adult athletes partaking in contact and collision team sports. Forty-three youth (15.4 ± 2.2 years) and 27 adult (22.2 ± 2.9 years) Rugby Union and soccer players completed the study. Eye movements were recorded using SMIRED250mobile while participants completed a test battery twice, with a 1-week interval that included self-paced saccade (SPS), fixation stability, memory-guided sequence (MGS), smooth pursuit (SP), and antisaccades (AS) tasks. Intra-class correlation coefficient (ICC), measurement error (SEM) and smallest real difference (SRD) were calculated for 47 variables. Seventeen variables achieved an ICC > 0.50. In the adults, saccade count in SPS had good reliability (ICC = 0.86, SRD = 146.6 saccades). In the youth, the average blink duration in MGS had excellent reliability (ICC = 0.99, SRD = 59.4 ms); directional errors in AS tasks and gain of diagonal SP had good reliability (ICC = 0.78 and 0.77, SRD = 25.3 and 395.1%, respectively). Four metrics were found in this study to be reliable candidates for further biomarker validity research in contact and collision sport cohorts. Many variables failed to present a sufficient level of robustness for a practical diagnostic tool; possibly, because athletic cohorts have higher homogeneity, along with latent adverse effects of undetected concussions and repetitive head impacts. Since reliability of a measure can influence type II error, effect sizes, and confidence intervals, it is strongly advocated to conduct dedicated reliability evaluations prior to any validity studies.