This paper describes a probabilistic distribution database of fall dynamics to develop preventive measures for childhood serious injuries due to fall. For this purpose, at first, we collected data of childhood activities at the sensor home which is a mock daily living space. This sensor home consists of a multivideo-surveillance embedded into a home environment and an acceleration-gyro-sensor attached to a child body. The acceleration-gyro-sensor is used for fall detection. Next, we calculate daily fall dynamics value by extracting fall videos from the recorded behavior data by a fall detection algorithm developed by authors and analyzing extracted fall videos, and developed database by accumulating fall data. In the developed database, a user can perform conditional search of fall data by giving search condition such as child attribution and fall situation. This database enables the user to conduct not only worst case analysis, but also analysis based on percentile in consideration of probabilistic distribution.