Nakamoto consensus is prevailing in the world largest blockchain-based cryptocurrency systems, such as Bitcoin and Ethereum. Since then, various attempts have been studied to attack Nakamoto consensus worldwide. In recent years, network delay has won more attention for making inconsistent ledgers in blockchain-based applications by virtue of attacking Nakamoto consensus. However, so far as we know, most of the existing works mainly focus on constructing inconsistent ledgers for blockchain systems, but not offering fine-grained theoretical analysis for how to optimize the success probability by flexibly dividing computational power and network delay from the viewpoint of adversary. The paper first utilizes network delay and the partition of controlled computation power of honest miners for making forks as long as possible. Then, formally analysis is provided to show the success probability of the proposed attack, and compute the optimal network delay and splitting for adversarial computation power in theory. Finally, simulation experiments validate the correctness of the formal analysis.