At this point I am enjoying creating and tweaking scenarios more than actually playing a new game. I don't care for RRIII so have reverted to RRII only. I will not create a scenario that I haven't won in expert mode while testing. I prefer shorter, especially 10 year, scenarios rather than testing it for hours and hours. I am not a big fan of city or territory connections for a winning condition. I usually strive to make my scenario for good replayability. One way I frequently do this is to eliminate regions and set all cities to 100% available industries. This places the industries at random all over the map and gives the AI players a better chance to haul demanded cargoes. I do as many things as I can to make the AI players competitive but not dominate, which is many times seems impossible. They generally depend on passenger traffic so limiting that and/or disabling relevant managers can help. They do haul cargo but seem to pay no attention whether demanded or not. Having a welcome bonus for the human player can help even the odds with the AI players. If one of the winning conditions is hauling a certain type of cargo then I will almost always also have another winning condition requiring a meaningful quota of lifetime industrial profits. Without the latter anybody can move lots of the required cargo to just any city whether they demand it or not. Another thing I have been doing for quite a while is using distinctive labels on the cities which designate their size which makes it easier for a player to figure out the best place to lay their first station. I use lower case followed by a * for villages, lower case for towns, first letter capital for cities, all caps for metropolises, and all caps with spaces between for the megapolises. I try not to have cities close enough together so that a large station can capture both. On many existing maps this entails moving or removing cities. I am also careful to not put sizable cities next to each other, especially in a string which makes the game too easy, and these tend to become primary AI starting targets. I'm not much into random events or plain news events though I do occasionally use them. I also prefer revising a fictional territory map over a real country one. This makes moving or adding cities easier as needed without messing up an existing country infrastructure. On occasion I have used an existing country map, renaming the map and all of the cities so I can move them around as I see fit. Coming up with new city names is tedious but works.