What they are doing is called facial verification, which compares images to each other to deduce they depict the same face. It is not true facial recognition per se, as in the computer sees a picture and 'knows' it's you.
But by pushing the limits of facial verification, it differs little from actual 'recognition'. It is like someone seeing your face in different settings and recognising it is the same person.
Now not just your grandmother knows it's you
Now how accurate is this? According to the brief synopsis Facebook put online: "Our method reaches an accuracy of 97.25% on the Labeled Faces in the Wild (LFW) dataset, reducing the error of the current state of the art by more than 25%, closely approaching human-level performance."
To lessen the annoyance of Facebook creepily suggesting tags, you can at least turn the 'tag suggestion' feature off in the Timeline and Tagging portion of your Facebook settings.
In the meantime, perhaps you should reconsider when you're about to upload your drunken selfies.