Just to introduce myself, I've been using ZZ since 2009 and achieved world class OH times with the method. If ZZ had a textbook, I'd most likely be one of the people to write it.
Personally, I think a fair comparison of ZZ-a and ZZ-b requires considering the user and his or her capacity to memorize and quickly load algorithms as well as his or her capacity to phase effectively. If you take out the user, I think both areas are too abstract to really warrant definite comparison.
I noticed gewinnste posted a longer post on page 21, so I'm going to respond to that.
EO-line takes about 0.8 moves more than the cross and so a tiny bit longer
Theoretically, sure. But, EO line is far more abstract than the cross. Optimal solutions may not be as finger friendly as good CFOP crosses. Some optimal EO line solutions are also extremely nonobvious.
3gen F2L in ZZ takes about the same amount of moves as CFOP-F2L (move restrictions in ZZ-F2L being made up by block building possibilities) but will actually be a bit faster because of no cube rotations and being 3-gen.
High level CFOP solvers often see their first pair during inspection. Doing ZZ effectively (getting that good move count) often requires finding a good first block. This can be difficult because ZZ offers twice the choice in terms of where to start (CFOP offers you one of four pairs while ZZ offers you one of four pairs and one of four blocks). I would say that the decision-making in ZZ is more complex. The move counts may be similar, but the process of solving and the flow is very different.
When I compare CFOP and ZZ, I usually like to think that CFOP is more directly productive and that ZZ saves its effectiveness for the end at the cost of an abstract beginning. CFOP solves 4 pieces after inspection, and generally requires solving two pieces at once four times. It is very organized. Modern CFOP is fast because the cross is straightforward and the solver aggressively solves pair after pair due to the lack of strain on decision-making. ZZ is very open-ended in the beginning of the F2L. But, once you get the first block, the rest of the F2L can be solved in one physical motion without stopping. Until then, the beginning of ZZ is very stressful. When someone says that ZZ F2L is "better" than CFOP F2L, I interpret that as they believe that eventually, ZZ solvers will be able to make decisions as quickly as CFOP solvers during the F2L so as to not slow down during the beginning of the ZZ F2L.
ZZ-b has some complications that you didn't specifically mention. For example, to phase effectively, you would have to phase using all four slots. I've never considered phasing seriously, but I'd imagine that phasing with all four slots seamlessly can get difficult. I've noticed that some people say that phasing is "lightweight", but I personally can't grasp phasing fluently without stopping using all four slots. Also, ZZ-b arguably adds another "look" in the phasing step. It's probably lighter than your average look, but it's arguably still a look.
P.S. I think that ZZ-a is just overkill (~100-150 cases for ZBLS + ~480 cases for ZBLL) and will likely not even be faster than ZZ-b because of the longer recognition times for the ZBLS and ZBLL cases.
People have already pointed this out, but to reiterate, the "~100-150 cases for ZBLS" you mention do not exist if you are doing ZZ-a. The whole point of ZZ-a is to do a regular ZZ F2L, which draws a fine line between the F2L and the LL. ZZ-a is just ZZ F2L + ZBLL. It's pretty simple conceptually. Part of the appeal of doing a normal ZZ F2L (which is what ZZ-a requires) is that you can aggressively solve your F2L without the added abstraction of phasing. To me, the phasing represents kind of an awkward speed bump in the solve. But, this could just be because I'm not completely fluent in phasing.
Generally
The reason I say that the comparison is personal is because both methods require capacity that I consider significantly above average. ZZ-a requires the ability to memorize a large number of algorithms while ZZ-b requires coping with another level of abstraction. If your strong suit is memorizing more algorithms than 99% of the cubing population, then ZZ-a might be better. If you don't fear the abstraction of phasing, then perhaps ZZ-b is the way to go. Making a comparison to determine which is better theoretically is a bit too abstract.
I think some of the people here were in the mood to argue because your comparison contained information that was not exactly correct (particularly, lumping ZBLS with ZZ-a). Personally, I believe that there are a lot of details in these methods that you can only appreciate once you've seriously studied them. Else, comparing the methods on a theoretical level isn't useful, since the discussion doesn't capture a lot of the things needed to make a fair comparison. I've been doing ZZ for a very long time, and I am well aware of many of its strengths and weaknesses.
Of course, you had prefaced your opinions with the "IMO". But, still, I guess some felt strongly about the issue because you are relatively unknown while they've invested significant time in learning this stuff. The whole thing about being relatively unknown shouldn't be too relevant, but it is here on this forum, realistically. Anyway, I hope the posters didn't give you a bad time. I think they are just very passionate about what they do.
