Roguelike is a game with a number of features, randomly working together to form a totally random gaming experience that even the developer can't explain.
When I started to plan Teemu's next version I had already started to use a system where a number of features is put to to do list. I think sometimes planning can go too detailed and the plan itself becomes the project, most likely to fail even in the planning process let alone when the game is implemented.
A short generic list of to do items is a viable idea, because that way you can limit the number of features to important ones and try to think how long it will take. Not only that but I have started to keep the initial feature as simple as possible, in the logic of start small ideology. That way you can both try to implement the feature sooner than later and also test it in real world.
We humans want to think that our ideas and plans are perfect, but more often they are not. That's why simple and fast sketch of the feature is something that can create a better gameplay even you didn't expect it to happen.