The load–displacement behaviour of timber joints is nonlinear, yet simplistic assumptions are often made in the design phase, leading to an idealized elastic-perfectly plastic representation of their response. This highlights the need for more refined models to accurately describe their mechanical behaviour. An example of modern timber joints is represented by steel-to-timber joints that employ laterally and axially loaded self-tapping screws, common in large timber structures. In this paper, first, the impact of different parameters of influence: the test set-up, the screw length, the friction, the torque, the load-to-screw axis angle, and moisture cycling on the load–displacement of joints with screws under applied lateral-tension load, was studied. Then, 10 to 20 tests were performed for every load-to-screw axis angle, for a total of 60 tests, to describe the shape and the variability of the load–displacement curve. Based on this experimental data, an empirical-probabilistic model using the Richard-Abbott analytical expression for describing the load–displacement curve was developed, and the model parameters were determined. A comparison of slip-modulus values at ultimate limit state derived from the empirical-probabilistic model with values prescribed by Eurocode 5 suggested that the current definition may be inadequate for steel-to-timber joints with screws. Beyond this contribution scope, the empirical-probabilistic model constitutes an input for multi-fastener joint models, which can subsequently be incorporated into the reliability analysis of timber structures.