Purpose: To evaluate the test-retest reliability of a contour-based stereoacuity test using a head-mounted display (HMD) and compare it with other stereotests. Methods: Thirty-two healthy adults aged 23-47 years were recruited from a tertiary hospital between August 2017 and July 2018. Two separate contour-based circles (crossed disparity: 135-1,350 arcsecs) were generated on a high-resolution phone display (Galaxy S7; Samsung, Seoul, Korea) using an HMD (Galaxy Gear VR). Two images were independently projected to each eye as graded circles with a random dot background. The results of the new HMD stereotest were compared to those of the standard Randot and TNO stereotests. The test-retest reliability was assessed using the Bland-Altman plot and Cohen’s kappa statistics. Results: Among the 32 study participants, 17 (53%) were males and the mean age was 30.1 ± 4.8 years (range: 23-47). The mean stereoacuity was 160.3 ± 53.5 arcsecs in the first HMD stereotest (HMD1), 28.4 ± 12.5 arcsecs in the Randot stereotest, 96.1 ± 83.5 arcsecs in the TNO stereotest, and 143.3 ± 47.7 arcsecs in the second HMD stereotest (HMD2). The Bland-Altman plot showed a mean difference of 0.042 (-0.189 to +0.272, 95% limits of agreement) between HMD1 and HMD2. The reliability analysis showed an intraclass correlation coefficient of 0.499 (p = 0.022) and agreement of 81.25% in Cohen’s kappa statistics (Cohen’s kappa index = 0.119, p = 0.017). Conclusions The HMD stereotest without monocular cues showed fair test-retest reliability and reproducibility. Further studies using a high resolution display are needed to confirm the validity of the HMD stereotest.