It tries to boil the problem down to a 'discomfort' level, where what really matters is how long I think I can comfortably stay on the road, how fast I can cross the road, and how long I expect to wait to cross the road.
Let's assume all sections of the road are equally good for crossing (no pedestrian crossings etc). The road will then have a 'maximum angle of attack' which is a function of how fast I am travelling and how long I can stay on the road. I will walk in roughly the same pattern as in the link HOWEVER I will never cross at an angle greater than the maximum and I will potentially cross earlier or later depending on gaps in the traffic; it's faster to keep walking then to wait for a gap in traffic.
The main shortcoming in the model is that there is no upper limit on the angle of attack, and so it is easy to find example situations that are unrealistic.