We derived fat indices for sika deer (Cervus nippon yesoensis) in eastern Hokkaido, Japan, and estimated the probability of over-winter survival with a logistic regression model using fat indices. Kidney fat mass (KFM) appears to be an adequate index of wide range of physical conditions before the onset of severe nutritional stress. When KFM values fell below 20 g, femur (FMF) and mandible cavity fat (MCF) indices declined sharply. FMF and MCF were useful indices for detecting malnourished deer. A logistic regression model describes survival thresholds in two bone fat indices for calves (45%) and three fat indices for adult females (FMF = 25%, MCF = 30%, KFM = 20 g). These models are useful for estimating the probability of winter survival in Hokkaido sika deer.