Path diversity is an important index of the dependability and fault tolerance capability of networks. Most path diversity metrics are designed under an implicit hypothesis that all candidate paths for a divergence node are always equally chosen. This may not be true due to the effect of routing policy with consideration of cost or other factors. A new simple metric is proposed in this paper for path diversity measurement of end-to-end networks. An end-to-end network is firstly transformed into a sequence of cascading segments. With the assumption that only one of all available branches is chosen each time, a probability based metric is calculated for each segment, and the joint entropy is calculated as path diversity measurement for the overall end-to-end network. The metric is fitful for both single-hop and multi-hop topologies. A comparison between the new metric and other ones is given at the end.