Of all the ideas you can bring to the table they will always sit in one of two camps:
First, the "known" world...
We start in the mundanely familiar, the places where diners need serving and crops need to be grown. As a player you take on an occupation that's available to you in real life, you might not have had the occupation but you essentially know what it involves. Communicating game mechanics within these settings is easy, because the framework of the tasks is already familiar, so much of the information is already known to the player. After the immediately known there are things that are familiar to us through popular media, specifically novels, films and television. These are the settings of military action, organised crime, espionage and hospital drama. They exist as real settings but we are often only familiar with them through stylised genre pieces. To us the versions of these places and occupations that are portrayed in popular media are the ones we know as real and these have become the ones aped in games. Driving games are, perhaps, the ultimate expression of this concept, everybody at least thinks they understand how to drive expertly at speed.
Then we have the realms of fantasy, including the supernatural, swords and sorcery and science fiction. Each of the sub genres in these settings have their own conventions: Elves are thin and beautiful, spacecraft manoeuvre like planes without gravity, sunlight is a pain for vampires, even it it only makes them glisten like diamonds, it's still an enemy. Games even have well accepted abstract conventions of their own, where enemies are killed by jumping on them and things can re-spawn infinitely. The most popular of these conventions is that things come in 3s: You start with 3 lives, keys come in 3 parts and, most importantly, 3 things of the same kind react together and disappear.
Games set in these established territories hang on known information, using knowledge you can reasonably expect the player to have at least some awareness of - you might have to communicate the finer details, but the broad strokes are already in their cultural data banks. However, all the infinite possibilities that we haven't pulled into our cultural domain live in a trickier place to pillage:
The hinterland of fail...
The hinterland isn't only populated by new worlds and novel mechanics though, it's populated by well understood mechanics that are used in a new ways and dressed in new fiction. In Shatter, for instance, we implemented an overheat mechanic on the bat, which prevented continuous use of the "blow" functionality. Overheating is a commonly used and well understood feature, often used on both weapons and vehicle turbos. However, it's not normally seen in brick breaking games. The team tried to communicate that the bat was overheating but sample players simply didn't make the connection between their actions, the on screen prompts and the effects. We knew this because players just held down the "blow" button regardless of any cues being given. Unable to communicate the mechanic the team chose to cut their losses and sought a different solution to the persistent blow problem.
Knowing when to quit the hinterland is a skill that can only be learned through experience, but the decision to go there (or not) in the first place is one that any developer, no matter how green, can make easily. Look closely at the time you have available to you, do you really have any time to waste? If the answer is no then stick to making known mechanics work in a familiar setting. Tuning a game to make it fun is hard work all by itself and with familiar elements you at least know they can be tuned. But with new ideas there's a hurdle that must be overcome even before you get to the tuning stage: The thing to remember about new ideas is that they are hard to communicate to the player. The closer a setting or mechanic is to something that's well understood, either through real life experience or popular culture, the easier the ideas are to convey.
That's not to say the hinterland should be ignored though, it' a fertile territory, full of opportunities and if you have the time you should explore it with gusto. Just remember that anything you find there must be dragged back and made a part of the player's known world. And this is not an easy job, it might well be one of the hardest in game development, because to make something part of the known world you have to be able to explain how it works and what it means. The hinterland of fail can and should be explored, it's possible to find rich rewards there but it is a sink, soaking up time and resources with impunity. Before you go there make sure you have the required time, manpower and usability testing available to navigate it, and even then, have a plan ready in case you still can't get your message across, because if you can't communicate something to player, it doesn't matter how clever it is, it's still a failure.