So for a while I just ignored Oracle. To me it seemed like a Spy that didn't even give +Action. It seemed pretty weak. But with the release of Catacombs, I've been re-assessing that. Because Oracle and Catacombs function pretty much exactly the same way, except Oracle messes with your opponent's deck. I actually played a game the other day where both were in the kingdom, and I ended up getting both of them, to my benefit.
Really, this, to me, is part of a larger paradigm shift for me where I'm trying not to think of cards as "bad," which, while perhaps not intentionally, often sinks into our heads when we rate cards, because cards at the bottom of the list will invariably be thought of as not good. Scout and Counting House are not bad cards, and they can shine brilliantly in the right situation. They're just not amazing cards.
So what I try to remember when looking at a kingdom is to not think "Oo, that's a good card, oo that's a bad one" but to think "How can I make these cards work together?"
Here's a game I played against a bot:
http://dominionlogs.goko.com/20120924/log.5057ee40f93b9113b758b6b5.1348505874535.txtSpy, normally ignorable, came in very handy as the only cantrip on the board, allowing me to King's Court it for draw and Actions, and thus play (or King's Court) multiple Barons (read: one Baron and a Band of Misfits pretending) per turn. It did also help that the bot ignored Lighthouse for too long, and I was able to Rabble it into oblivion.
And this is why I like the Dominion Strategy articles - it emphasizes "how can I make this work?"