The superior soft-tissue contrast offered by the integrated magnetic resonance imaging (MRI) of the Unity MR-linac compared to the x-ray-based image guidance on conventional linacs potentially allows for liver stereotactic body radiation therapy (SBRT) without the need for implanted markers or other surrogates. On conventional linacs, liver SBRT patients are typically positioned with their arms above their heads (arms-up) to minimize exposure to healthy tissue. However, the spatial confinement of the MRI-bore and increased treatment delivery times can make the arms-up position straining. Therefore, we assessed the plan quality for MR-linac treatment plans with the patient in the arms-up and in the arms-down position. Additionally, we compared the MR-linac plans with clinically used arms-up treatment plans made for a conventional linac. Fifteen consecutively treated patients with oligometastatic liver disease were included in this retrospective study. For each patient, a planning computed tomography (CT) with delineations, a diagnostic MRI, and a 3×20Gy dual-arc volumetric modulated arc therapy (VMAT) plan, which was used to treat the patient in an arms-up position on the conventional linac, were available. For the MR-linac, 15-beam step-and-shoot intensity-modulated radiation therapy (IMRT) plans were created for four patient positioning scenarios: arms-up, mimicking current clinical practice; arms-down, with treatment beams avoiding the arms on the entrance side; arms-through, arms are down but not avoided, and right-arm-up; only the right arm is up and the left arm is avoided on the entrance side. Resulting treatment plans were compared. Bonferroni-corrected two-sided Wilcoxon signed-ranks tests were used to assess statistical significance (P<0.05). No significant differences were found in gross tumour volume (GTV) coverage (D , D , and D ) or liver sparing (liver-GTV V ) between the clinical plans and any of the MR-linac plans. The median target conformity [exterior V /planning target volume (PTV)] was significantly better in the clinical plans (5.8) than in the MR-linac scenarios (arms-down: 6.6, arms-up/right-arm-up: 6.2, arms-through: 6.3). No MR-linac plan violated any additional organ-at-risk (OAR) constraint that was not already violated in the clinical plans. In the arms-down scenario a significantly increased median spinal cord D (14.5Gy) was detected compared to the clinical setup (7.2Gy). For the arms-down (arms-through) scenario, the median left arm D was 1.5 (2.7)Gy, the median right arm D was 5.8 (22.7)Gy, and the median right arm V was 0.0 (14.7) cc. These differences were statistically significant. For the right-arm-up scenario, the median left arm D (2.3Gy) and V (0.0) were not significantly different compared to the arms-down scenario. Mimicking the current clinical practice by treating patients in the arms-up/right-arm-up position on the MR-linac leads to plans which are dosimetrically very similar to the conventional linac plans. Treating in the arms-down position is expected to increase patient comfort at the cost of a small reduction in OAR sparing for individual patients. Treating through the arms is not encouraged due to substantial dose deposition in the arms.