A statistical fit of an algorithm to "calibration data" gives parameter values for a "probabilistic decompression model." Our objective is to prepare a simple model that will estimate risk of decompression sickness (DCS) in air dives. We develop a logistic regression model using calibration data from carefully controlled experimental dives recorded in the U.S. Navy Decompression Database. We exclude saturation dives, which can have very long decompression times. For most depths, our model's prescriptions for 2% probability of DCS avoid the experimental DCS cases without mandating excessive time at decompression stops. Our model indicates that the long decompression times prescribed by some previous probabilistic models are not necessary. Our model cannot be used operationally because it cannot calculate depths and times at decompression stops; however, there is general concurrence between our model and prescriptions of a deterministic model known as the VVal-18 Algorithm; this supports the adoption of theVVal-18 Algorithm for operational use on decompression dives.