Among the various types of concurrency bugs, the data race is one of the primary causes of other concurrency bugs. Thus, it is important to detect as many data races as possible during the development step of multithreaded programs. A hybrid data race detection technique that uses the Lockset algorithm and happens-before relation, can detect actually occurred and hidden data races in one execution trace. However, high runtime slowdown obstructs the frequent use of hybrid detectors. In this paper, we empirically demonstrate that most data race bugs are caused by the absence of a lock, and that multiple locks are rarely involved in a data race bug in the real world. Thus, we propose a fast hybrid detection algorithm that does not introduce additional false positives and false negatives to the current hybrid detectors. The suggested algorithm replaces the lock-set intersection by a simple comparison operation that focuses on exploring data-race-prone locking patterns. The experimental results indicate that the proposed algorithm detects the same data races as Multilock-HB, which is the most accurate hybrid detector, with a 1.18x slowdown of FastTrack for eight large-scale benchmark programs.