Lens search performance based on a single epoch. The lens search algorithm is tested using difference images from only one single epoch for all the 13 epochs in the HSC transient survey; the ROC curves of the epochs with the best performance (left), the average performance (middle), and the worst performance (right) are shown. For lens search algorithm based on a single epoch, only p_thrs is used as constraint. In each panel, the ROC curves are plotted by varying p_thrs%; the diamonds, triangles, and circles indicate the points on the ROC curves at p_thrs% = 90%,95%, and 97.5%. The seeing of each epoch is also indicated in each panel. The best lens search performance happened in the single epoch with a seeing of 0.72″, capturing the bright lensed quasars with wide separation at (TPR, FPR) = (97.6%,2.4%) when p_thrs% = 97.5%. In a single epoch with average seeing (∼0.98″), the lens search algorithm can detect the bright lensed quasars with wide separation at (TPR, FPR) = (94.2%,5.0%) when p_thrs% = 95.0%. The worst lens search performance happens in the epoch with an exceptional seeing value of 0.48″ due to artifacts appearing in the difference image pipeline when the seeing is “too” good.