(First posted June 12th, 2009 on Tetris Concept)
Preface: When I say four-state, SRS, or symmetry, I'm talking about rotations from this table. When I say ARS, Sega, bottom-row alignment, etc., I'm referring to this table. More specifically, when I say bottom-row alignment, I am referring the property of most pieces in Sega tables to have at least one block in the lower-most row of a piece's bounding box in all orientations for a given piece.
Although I'm sure many would like to start a write-up on the problems with SRS's kick table, I think the logical starting point is with the basic rotations. The way basic rotation is handled has a big impact on the ultimate usability of the rotation system. The first thing to talk about in regards to rotation tables is how SRS handles symmetry. The TTC justifies symmetry by asserting that it is ultimately more intuitive than a rotation system that features either bottom-alignment or asymmetry like ARS. However, I believe that SRS is ultimately no more intuitive ARS in low gravity situations, and is in fact slightly _less_ intuitive. (ARS is also a clear winner in high gravity situations, but that necessitates discussion of wall kicks. I'll get to that in a little bit, sit tight.)
Although the claim that symmetry is ultimately more intuitive sounds pretty reasonable, I don't believe symmetry is something that can be handled in Tetris without introducing some odd behavior. Symmetry can be implemented quite logically for L, J, T, and O, but how do you handle S, Z, and I? The point around which the brain would logically want to call the center for each piece falls _between_ blocks. You can't rotate around this point without throwing the piece out of alignment with the grid. SRS deals with this by rotating the whole piece around a block _near_ the center. Since these pieces have an offset center of rotation, they wobble between four states of rotation -- upper horizontal, rightmost vertical, lower horizontal, and leftmost vertical. This isn't the most intuitive way to handle things. We can only immediately see whether it is vertical or horizontal and can't see the block that is the center of rotation. In fact, our brain only ever wanted to see two rotation states in the first place! I think Sega had it right in 1989: make a logical compromise. For example, lining up the "point" of a vertical S or Z piece with the center column of the horizontal one makes a lot of sense. The I-block is a little bit harder to deal with, but picking a column that is close for the vertical orientation works out perfectly fine. If it sounds wrong to be arbitrary about that, why not give _just_ the I-block that wobble? S and Z feel more awkward since their center of rotation is placed on an arbitrary block, but the I-block has its center of rotation place quite close to where it would be in reality anyhow.
Saying that it looks more natural with symmetrical rotations is also a silly argument. Don't _wall kicks_ look unnatural? Why not get rid of those? How about line clears? Naive gravity? I think it is important to remember that Tetris is an abstract game. Things like "looking natural" don't ultimately matter. Past that, we rarely ever see the pieces in free space anyhow! Since pieces spawn in the vanish zone, we'll be looking at the ghost piece for a great deal of the time. The funny thing here is that the behavior of the ghost piece matches ARS's bottom-row alignment. There is more of an argument for bottom-row alignment of L, J, and T to match the ghost piece for the sake of consistency of appearance than there is against it because it looks "unnatural."
I'd say that bottom-row alignment is more elegant in providing useful functionality. With pieces L, J, and T in SRS, a wall kick is _necessary_ to go from their flat-side-down orientations to their vertical ones. (Figure 1) On the other hand, ARS gets that rotation for free as part of its basic rotation. (Figure 2) Ultimately, this makes bottom-row alignment more versatile under high gravity. We can slip pieces under notches quite cleanly like this, whereas rotation would fail or they'd be pushed out by a wall kick with a symmetrical rotation table. (Figure 3) The real treat here is that by giving a bottom-aligned rotation a floor kick, you've given it more options than a symmetrical one. You get the useful twists without any inconsistencies in wall kick rules, and you can still get the benefits of a symmetrical rotation table should you get your piece "totally" stuck. (Figure 4) By using a bottom-up approach, you can have a more flexible rotation system without having to make those useful twists an exception by adding new wall kicks.
Finally, for reasons beyond my understanding, SRS's kick table is nearly incomprehensible. If the design philosophy for SRS was to be intuitive and natural, they certainly forgot about that here. The only thing that would be easy to understand about these kicks is that rotating will probably let you climb over previous inappropriate placement decisions and ugly mistakes with ease. The only way these wall kicks provide a challenge is by being frustratingly difficult to understand and thwarting hopes of playing quickly and efficiently. ARS's wall kicks, however, can be described in a few sentences. (The wall kick order is one right, one left, fail. With one extremely rare exception, L, J, and T won't kick if their center column is blocked after basic rotation. I-block could not kick in TGM1 or TGM2, but can kick up to two spaces away directly away from either the wall or floor in TGM3. T also has the ability to kick one space up if the horizontal kicks fail. The floor kicks for I and T reset lock delay and only work once per piece entry.) While ARS kicks are predictable and aligned with only one axis, SRS frequently pulls pieces diagonally and in directions that were not expected. If nothing else, SRS needs some sensible, straight, predictable kicks. I'd also advocate the final set of kicks being much less lenient than SRS. Max gravity should not be a matter of persistently rotating to hop over problematic elements of the stack. If there isn't a more significant difference between low and max gravity than "in one, you have to rotate more to get things where you want them," why even bother getting that fast in the first place? I believe that the set of wall kicks in ARS provides a very fair challenge. Since the rules are clear, you never feel like you were cheated because the piece didn't do something you thought it should have done. If ARS's kicks are "too hardcore," how about something like DRS?
PS: This is quite awful, haha. This is worse.