A new detection method for ultra-low frequency gravitational waves (GWs) with a frequency much lower than the observational range of pulsar timing arrays (PTAs) was suggested in Yonemaru et al. (2016). In the PTA analysis, ultra-low frequency GWs ( 10 −10 Hz) which evolve just linearly during the observation time span are absorbed by the pulsar spin-down rates since both have the same effect on the pulse arrival time. Therefore, such GWs cannot be detected by the conventional method of PTAs. However, the bias on the observed spin-down rates depends on relative direction of a pulsar and GW source and shows a quadrupole pattern in the sky. Thus, if we divide the pulsars according to the position in the sky and see the difference in the statistics of the spin-down rates, ultra-low frequency GWs from a single source can be detected. In this paper, we evaluate the potential of this method by Monte-Carlo simulations and estimate the sensitivity, considering only the "Earth term" while the "pulsar term" acts like random noise for GW frequencies 10 −13 − 10 −10 Hz. We find that with 3,000 milli-second pulsars, which are expected to be discovered by a future survey with the Square Kilometre Array, GWs with the derivative of amplitude of about 3 × 10 −19 s −1 can in principle be detected. Implications for possible supermassive binary black holes in Sgr * and M87 are also given.