Short fatigue crack density evolution on specimen surface of railway LZ50 axle steel is studied experimentally by local and overall viewpoints, respectively. Results indicate that the density of effective short fatigue cracks (ESFCs) around dominant crack, which results in the specimen failure, evolutes sensitively to the changes of the dominant crack size and tip locations. In accordance with the previous criterion of ESFCs, this density evolution is more suitable for describing the intrinsic and dynamic localization and randomization of fatigue damage. At the same time, statistical deriving is applied to address the randomness of the evolution behavior. It is revealed that the density data increases with the dominant crack growth in the micro-structural short crack (MSC) regime, and keeps declining in the physical short crack (PSC) regime. Coefficient-of-variations exhibit a contrary tendency. Appropriate distribution is further determined using a previous comprehensive statistical comparison method. Random characters and quantitative measurements of the density data have been well depicted.