I'm inclined to disagree just a little bit on 'all you can do is go for that last 1%' actually. As far as competition goes, I know there's always the idea of hey, I want my game to be the best/most popular so it sells well and so it can be a project I'm really proud of, I understand that. However, despite all the perfecting of Metroidvania, I don't often find myself revisiting the old games I have for any more than 6-ish or more playthroughs. I want more new stuff in that genre. Even if it were a game that played the same, and just had different zones (or re-randomized buildings and caves, for instance), I would eat it up. The reason why, is because it's a genre I'm actually really really into. The people that are into that genre are probably going to try to check out all the competitors, if they can. I legitimately love exploring new territory, finding new stuff, and getting a bit better over time as I do it. That's a characteristic of AI War that I really really like, as well, and a big part of why I really play it. Sure, it's always the same basic stuff, but man I love finding that planet with an advanced constructor, ARS, and a couple fabricators or something. Just, that jackpot of stuff being in a different place every time is so cool to me.
I also want to ask about that last 1% actually. Define the last 1%. Again, going back to metroidvania, it's not as if each one is the same game. Look at the most recent DS games in Castlevania, and you'll see Dawn of Sorrow, Portrait of Ruin, and Order of Ecclesia. In basic mechanics, they're pretty similar, but there's always this one huge thing that sets each one apart from the last. Dawn of Sorrow allows you to learn (via drops) one attack from literally any enemy in the game. Portrait of Ruin gives you two characters and the puzzles associated with that. Order of Ecclesia makes every single attack a spell, and allows you to combine every spell with any other spell. Sure, those are little differences in the mechanics of the game, different ways to spice up the game a little... but who says that those can't be fleshed out into some kind of really unique game? Metroidvania/puzzle is something that kind of exists elsewhere. I think the soul system in Dawn of Sorrow could have used a lot of refinement. Portrait of Ruin could try more puzzles, separate the characters, improve the AI, any number of things. Order of Ecclesia could allow you to combine more things, or add a gem or rune to spells to make them more unique in addition to all of the combos. All of a sudden, you have games that are very fundamentally the same, with a very different feel from each.
The real problem there, I guess, is that we're talking making a game unique once the very basic mechanics have already been perfected. To a gamer like me, that would guarantee everybody my money. I'm a bit of a connoisseur of the little differences like those, especially in genres I'm really really into like metroidvania. I definitely don't feel like there's any 'perfect' one, either. I don't feel like they're hunting for innovation, they're trying to make sure the game doesn't get stale, and succeeding. Is that the 1%? Or, is the 1% strictly limited to the basic game mechanics?