IMO, Rogue is a surprise deck in order to counter general aspects or highly played decks in the metagame.
The balance of rogue vs. meta is interesting. Metagame shifts are not only caused by new sets coming in and old ones rotating, but with the rogue playing community with superb deck building quality and ability to detect and define the states of current metagame. Rogue is, in the cases where the cardpool isn't changing, the birth phase of the deck. Not to be confused with plain bad decks that didn't exist before. The things are dead from the start.
The loop goes: Rogue -> Archetype -> Old archetype (the type of deck which has been established, but cannot compete with the current archetypes due to meta shifts). I personally think that this is a loop. A deck which has been an old archetype CAN go back to being rogue and start it's way towards the archetype again. This just means that the deck should have gone completely under the radar, as in not taken serious part of the competition anymore.
An important part of rogue is also the surprise factor. This should be noted when looping old archetypes back to rogue phase - if your opponent knows the whole deck you're playing card-by-card, you'll lose 10% in the matchup. It's like when Plox gained it's Healers - something unpredicted, causing the deck to do better countering the fast, medium-damage decks by increasing longevity of the powerlock. The deck was played by 4 people team in Finnish 2009 nats, winning all 3 age groups.
I don't think you need to use the cards no one has ever even glimpsed before in order to make a rogue deck. In fact, it's usually important to use the proven cards or engines for some extent. Still, you should be able to forget everything when starting to build a rogue deck - no fixed lines, no staples (I don't mean that your deck should not contain staple cards when it's finished - just that every card would have to earn it's place in the deck in question, not just because it's played in every other deck means that it should automatically be included in 4s). No limits. This makes a rogue building an extensive process. And to think that rogues have to be rebuilt for every tournament. Try going 3 tournaments, say week away from each other, in the row with the same exact rogue list and win - first is easy, second is hard and third is somewhere near impossible - at least, if you win the third one, you'll no longer be holding a rogue deck in your hands as it should be recognized as a part of metagame already due to being something but a rogue deck - a well built deck that is showing potential to be the next shift of the metagame.
Of course, there are a lot of bottom table rogues. There can be plenty of reasons - the consistency is off, the meta-prediction wasn't accurate enough, or the concept of rogue is unclear. With the concept, I mean that rogues are aimed to counter the currently played decks. Rogue isn't random binder card.dec. Rogue isn't tier3.dec. Rogue isn't something that can be built without certain mindset and knowledge about the meta.