Traffic monitoring using Bluetooth MAC addresses has been intensively studied for a decade. Nevertheless, estimation of Origin-Destination volume is still challenging, because of the unstable nature of the detection. With the aim to reveal the factors affecting the detection probability of MAC addresses from moving Bluetooth device, this study conducted a series of driving tests to collect detection samples under various scenarios. The data was then utilised to develop a Logistic Regression Model to estimate detection probability considering the installation positions of the scanner and the driving speed of the experiment car. The results agreed on the contribution of distance and angle between the scanner and the car, as well as the car speed, as reported in preceding studies. This study further identified the contribution of driving direction, and height and timeout duration of the scanner. The proposed model successfully estimated the detection probability with reasonable accuracy.