It has been said that F. J. A. Hort, in constructing the text of the Westcott & Hort edition, simply looked for the readings of B and followed those.
This is just about precisely backward. Hort did not start from some anonymous text and then start looking for ways to correct it toward B. Rather, he started from B and then looked for places where it should be rejected. In other words, he used B as a "copy text."
It is curious to note that the copy text (also known as a copy text), one of the fundamental devices of most classical textual criticism, doesn't even seem to be mentioned in most manuals of NT criticism. Simply put, the copy text is the starting point for an edition. An editor, after examining the various witnesses, picks a particular manuscript as the best source and then, in effect, collates against it looking for places where a better text presents itself. As G. Blakemore Evans puts it in the textual introduction to the Riverside Shakespeare, "an editor today, having chosen for what he considers sound reasons a particular copy-text, will adhere to that copy-text unless he sees substantial grounds for departing from it" (p. 37).
This, we should note, does not mean slavishly following the copy text. Hort didn't follow B closely; a good editor will be open to good readings from any source. But the copy text is the starting point. One follows it in the absence of reasons to depart from it. So, for example, one would tend to follow the copy text spelling of various proper names, or on points of Attic versus non-Attic usage, or on inflected versus non-inflected Semitic names. And, of course, in the case of readings where the canons of criticism offer no clear point of decision, you follow the copy text. It gives you a fallback if you have no other grounds for decision.
Note that this is in strong contrast to most methods of Eclecticism. Eclectics generally don't start anywhere; they have to decide everything -- even such trivialities as spelling variations -- from the manuscripts or from some external reference. It's a lot of work for slight reward -- and it arguably produces a rather inconsistent text.
Now we should note that the Copy Text notion arose in situations with very few witnesses -- e.g Shakespeare, where there are never more than three independent witnesses, usually not more than two, and occasionally only one. However, the idea has been successful enough that it is now applied to texts with far larger numbers of witnesses -- e.g. Chaucer, where some passages have as many as 75 witnesses. There is no inherent reason why the method could not be applied to the NT as well.
Of course, if one is to choose a copy text, there is the question of which copy text. This is rendered much more complicated by the nature of New Testament witnesses: Most of the important ones, the papyri and uncials, lack accents, breathings, punctuation, and spaces between words. Should one adopt a copy text which includes these features (in which case it will be much more recent than what are usually considered the best witnesses), or choose a text with the best text apart from readers' aides? Or even choose one text for the text and another for the aids?
If you prefer the Byzantine text, it probably isn't an issue. Others will face a harder choice. Personally, I would incline to take the best text, while allowing for the possibility of a text with more reader aids.
On that basis, I would suggest the following:
Gospels: B. Or P75 where it exists, but consistency argues for using B throughout. There are no other real candidates. is mixed and rather badly copied, and every other copy except D has Byzantine mixture. (Of course, if you prefer the Byzantine text, you can have a copy manuscript -- probably E or perhaps W.
Acts: Again, B. Although there are proportionally more good manuscripts, none can claim superiority over Vaticanus.
Paul: Now this one is complicated, as there are fully four reasonable candidates: P46, B, , and (improbable as it sounds to list a minuscule) 1739. Nonetheless, I would argue that 1739 is the best of the choices. The best texts -- at least in my opinion and that of Stephen C. Carlson; compare also Zuntz -- are P46, B, and 1739. But P46 is very incomplete, and also contains a much-too-high rate of scribal errors. B is better on this count, but it too is defective. Adopting 1739 gives us a very good text, complete, and supplied with accents and breathings. The other alternative, , will appeal primarily to those, such as the UBS committee, who believe in Alexandrian Uber Alles without noting that the quality of the different types changes from corpus to corpus.
Catholics: Here again we have several options: B, P72, , A, C, and 1739 are all possibilities. P72 is probably eliminated by its incompleteness and its errors plus its wild text of Jude. A is the head of the main branch of the Alexandrian text, but while that is the largest group, it does not appear the best. C would have a strong case if it were complete -- indeed, if it were complete, it would be my first choice -- but it's too fragmentary. Textually, stands almost alone; so does B, whereas 1739 heads a large group. Ultimately, I would say the choice comes down to B or 1739. I would incline very slightly toward B.
Apocalypse: Here again we have four choices: A, C, , or P47. The latter is eliminated by its fragmentary state. isn't a particularly good text. C may well be the best text, but it once again has too many lacunae. We must choose A almost by default.
Those who wish to examine a text built on the copy text concept, a PDF of a text of Philippians (with critical apparatus, introduction, and commentary on significant variants) is available here.
We should recall, however, that the copy text concept can be applied to more than just the text of the New Testament. An edition of one of the versions might well be founded on a particular copy text (and some have been -- e.g. the Hopkins-James edition "The Celtic Gospels," an attempt to recreate the early Vulgate texts used in the British Isles, is based on Codex Lichfeldensis). So we should probably enumerate points to be considered in choosing a copy text.