Since the balance of roads is a sensitive issue, I made it in four versions:
v1.0 - default version
v1.1 - without villages in the central paths; this should reduce the chance of the AI "leaking" to the opponent, but it makes two out of three paths without roads and increases the length of the "non-stop" run to the opponent.
v1.2 - with multi-paths to the nearest villages; this increases the chance that a road will lead there, but also increases the risk of getting a one-way path.
v1.3 - combines both.
You can view the scheme in TEditor; I don't feel like drawing it. If anyone doesn't mind, please draw it.
In general, we need to generate all four versions and determine the optimal one. It's best to test it with numbers right away - statistics from 20-30 generations for each template are usually informative enough.
The content of the zones is typical everywhere. The guards in the central paths can be increased (if needed).