Recently, there have been plenty of "innovative" games. Fracture's terrain deformation. TimeShift's...well, time shifting. Mirror's Edge's perspective and art style. Portal's portals. Braid's time controls. EndWar's voice commands. All these games have been met with various degrees of success. So what factors make an innovation a success or failure? In short, what's the difference between an innovation and a gimmick?
I think we can all agree that Portal and Mirror's Edge are both pretty different from most games in the first person genre. But why is Portal considered innovative and near-perfect while Mirror's Edge is almost universally considered just a worthwhile rental? Let's examine 3 differences:
1) Format: Portal was a 3 hour, $20 game that you could get as part of a package. Mirror's Edge is a ~6 hour, $60 stand-alone game.
2) Setting: Portal is set in an experimental testing facility. Mirror's Edge is set in an urban environment.
3) Hype: Before release, there were a few videos showing off the portal mechanic that got people pretty curious about the game. Mirror's Edge released plenty of videos showing off the mechanics, along with many of those ESurance commercials.
Let's do a thought experiment. What if Portal had been executed under the same circumstances as ME? Imagine Portal as a $60, ~6 hour game - so about twice as long as it actually is - and it was set in an urban environment. How well would the portal mechanic fare in those circumstances? Could the designers have thought up enough puzzles to make the game last 6 hours? Would the puzzles in Portal even translate well in the urban setting? What technical issues would plague the game due to the setting, such as draw-distance and NPCs? And what if the pre-release marketing had been as extreme? What if half the puzzles were ruined in videos, and GlaDOS's manic-depressive nature was revealed before anyone even played the game?
My argument is that under those circumstances, people would not have liked Portal so much. The $60 price tag would turn people off already, since 6 hours is pretty short for a full-priced game. Furthermore, it would be tough for designers to milk that single mechanic and keep it interesting through out the whole 6 hours. The testing lab environment also allowed the designers to do all sorts of random stuff, like the energy balls, the platforms, and the Companion Cube. It would be a lot tougher to include such things if they had to make sense in the urban setting. And lastly, with all the pre-release marketing, the viral pleasant-surprise factor would be gone.
The fine-line from gimmick to innovation is not crossed by the idea, but rather by the execution. It has to be put in proper context: if the innovation does not produce enough truly novel experiences for a 6-hour game, then you should make it 3 hours. Or even less. Fracture's terrain deformation proved to be interesting for about 30 minutes - the game should've then been 30 minutes long (or just not made at all). Otherwise, players will get bored, and you will waste time. If you make the innovation only a part of the game, then the rest of the game must be exceptional compared to others in the genre. ME's running wasn't enough, so they introduced melee combat and some shooting, both of which were seen as half-assed attempts to fill out a pretty sparse game.
In conclusion, incremental innovation (which all of these are) should be introduced subtly. A single new mechanic will not alone give your game the legs that it needs for long-term success. So you should either scale your game down accordingly (Portal, Braid), or add it to a game that's already strong as-is (HL2's gravity gun, Gears of War's cover system). Since ME's novel mechanic is pretty reliant on its novel controls, I think it should've gone the Portal route and become a smaller game, where every minute spent with it could be made fun and interesting.