Ground configurations are an essential part of rail vehicles during wind tunnel tests. Reasonable reproductions of the infrastructures are of significance for experimental accuracy. In this work, the effects of ballast length were numerically studied via improved delayed detached eddy simulation (IDDES). Three extension lengths of the ballast accompanied by a 1/8th three-unit train model were proposed, with a pseudo infinite scenario as the benchmark. The numerical algorithm was firstly validated through a wind tunnel test in terms of aerodynamic loads and pressure distribution. The results revealed that the extended length of ballast had a significant impact on the leading and tail cars while less affected the middle car, due to the relatively farther distance from ballast ends. Ignored the drag coefficient, the influence of the ballast length on the separate cars was that the aerodynamic coefficients predicted by much longer ballast got closer to the benchmark. The ballast-edge-induced unrealistic flow and the train-nose-tip-induced realistic flow could easily be merged and interacted with each other and reinforced, thus causing a noticeable low-velocity zone at the leeward side. With the increasing ballast length, those two separations gradually moved away from each other, decoupled, and receded. A minimum length of E/H = 2 is recommended considering the aerodynamic performance and practical constraints of wind tunnel width. This study provides potential guidance for the arrangement of ground configurations for rail vehicles during wind tunnel tests.