<p>Several existing visual object tracking benchmarks, including OTB100, NfS, UAV123, LaSOT, and GOT-10K, mostly feature day-time scenarios. However, the challenges posed by the night-time remain relatively underexplored. We attribute this primarily to the lack of a large-scale, well-annotated night-time benchmark for rigorously evaluating tracking algorithms. In this paper, we first introduce NT-VOT211, a novel benchmark specifically designed to thoroughly evaluate visual object tracking algorithms under a wide range of challenging night-time conditions. NT-VOT211 consists of 211 diverse videos, offering 211,000 well-annotated frames with 8 attributes including camera motion, deformation, fast motion, motion blur, tiny target, distractors, occlusion and out-of-view. After conducting a comprehensive analysis of the results from 42 distinct tracking algorithms on the NT-VOT211 dataset, we develop a simple yet effective zero-shot domain adaptation method to significantly enhance the tracking performance of state-of-the-art trackers under low-light conditions. In this method, a newly designed module aims to distinguish the background token from the target token before feeding into the MLP Head. Our module enables for more accurate position estimation. Remarkably, it accomplishes this enhancement with merely 11 epochs of fine-tuning on a standard daylight dataset. Our dataset, code and other assets can be found at: <a href="https://github.com/LiuYuML/NV-VOT211">https://github.com/LiuYuML/NV-VOT211</a></p>