Abstract: The paper attempts to address a worldwide difficult problem. This problem is how to predict the time of happening landslide. There is no solution yet because of the lack of clear recognition of mechanism about landside and no correct quantitative model to be set up. Accordingly, this paper is based on the geological condition, landslide examples, and external data at the Three Gorge project's bank slopes and southwest mountains, as well as experiences aboard. It demonstrates the necessary and full condition of landslide happening. So, this paper eliminates the terrain control theory and stratigraphic control theory for causing landslide. The paper elucidated that the necessary conditions of landslide is a slope that has easily slipping structure and the full conditions of landslide is the effect of inducing factors. So the dynamic variation of inducing factors is decisive for the time of happening landslide.  Whereas rock-landslides are the most widely distributed. They are induced by underground water. In order to set up a correct water-starting model, the paper sums up the recent theories of typical rock landslide characters as follows: (1) The sliding face is the through penetrative discontinuities with the most difference of water transmissibility; (2) The landslide body's length is bigger than its thickness. The ration of the length to the thickness is about 20; (3) The front of landslide is starting early; (4) There is the phenomena of water seepage or water sprinkling in the front of slope body before the occurrence of sliding; (5) The sum of effluent water width on the slope is much smaller than the total width of the front of landslide body. Further more, the paper applies the hydraulics and hydrogeology theory to pointing out the following problems associated with a rock stability model proposed by others in 1970. It is found that (1) The old model didn't point out conditions of the distribution of the weight of rock over through the penetrative discontinuities; (2) The distribution characters of crack water pressure in the old model is against the observed landslide phenomena and hydraulics theory: (3) The old model didn't considered the percentage of going through infiltration water. Finally, the paper sets up a new slope stability model and critical value calculation equation of water-start landslide. The model and the equation are formulated according to easily slip over dip slope and chair-shape slope characters. Then, it brings forward the method on how to acquire parameters of the combined internal friction angle and the percentage of going through infiltration water as follows: (1) The back-calculation parameters of landslide; (2) The method of observation and statistic about the dip angle of slip penetrative discontinuities without hydraulic pressure and the percentage of going through water; (3) The experiment of the shearing and simple hydrogeology; (4) The method of geophysical exploration.