In radio-requency identification (RFID) systems, when multiple tags transmit data to a reader simultaneously, these data may collide and create unsuccessful identifications; hence, anticollision algorithms are needed to reduce collisions (collision cycles) to improve the tag identification speed. We propose a one-time collision arbitration algorithm to reduce both the number of collisions and the time consumption for tags' identification in RFID. The proposed algorithm uses Manchester coding to detect the locations of collided bits, uses the divide-and-conquer strategy to find the structure of colliding bits to generate 96-bit query strings as the 96-bit candidate query strings (96BCQSs), and uses query-tree anticollision schemes with 96BCQSs to identify tags. The performance analysis and experimental results show that the proposed algorithm has three advantages: (i) reducing the number of collisions to only one, so that the time complexity of tag identification is the simplest O(1), (ii) storing identified identification numbers (IDs) and the 96BCQSs in a register to save the used memory, and (iii) resulting in the number of bits transmitted by both the reader and tags being evidently less than the other algorithms in one-tag identification or in all tags identification.