Our game focuses on exploring a city searching for clues and learning new skills to reach new locations to find something very valuable to the protagonist.
The game design is based on two ideas:
Choices through gameplay. Instead of explicitly making the action of "choice", the outcome of a level objective is what makes the choice. For example: Gary's Eevee evolves into 3 different Pokémon depending on your previous battles with him in Pokémon Yellow. It's not something you chose; it's something that happened while you tried to win or lose against him.
Real time actions. While you are doing an action (i.e. Jump over a fence), you may be interrupted by something or some NPC, and the resulting outcome differs a lot from the one expected by the player.
When we started testing the first levels, we noticed that the player may get bored exploring the streets of the city. The protagonist is a small sized animal (like an adult pet dog), and we wanted to make the scale of the city realistic relative to the main character, so in comparison the size differences of both protagonist and levels are huge.
So we thought of adding stealth and pursuit mechanics (the protagonist, as an animal, would have to avoid animal rescuers). Both of these mechanics seem to fit very well within the two main game ideas. But when it comes to implementing them, it becomes a huge hassle.
Firstly, the player will be forced to escape once a pursuit sequence happens, trying to find a good hiding place, and will not stop to think where they should go.
Then we have the problem that the game doesn't have a proper Game Over concept, but just bad endings, so we don't know how we should handle what will happen if the protagonist is caught. The only current possible "Game Over" situation is to starve (which on a normal gameplay would be almost impossible to achieve), and you'd have to start over from the very beginning because its considered a bad ending of the game. Getting caught might happen much more often, so if we do the same as for starting, it would be extremely frustrating for the player to play all the way from the beginning.
Another problem, which is more production wise, is that it may turn out to be so time consuming to implement the AI of each NPC that may chase over the player, since we'd want to include factors such as personality, number of encounters with said NPC and the time the player has been within that NPC's FOV, and make a variation of pathfinding for these kind of Al's (we have to do it manually since we're using a custom engine).
Also, we may have to rethink the design of certain streets and areas, which may alter the layout of the town, which is tightly designed after the branching story, the main axle of the game.
Is there a better way to include these stealth and chase mechanics, that minimize the occurrence of these problems? If not, would there be anotjer workaround?