Two ways of comparing these are:
- How they influence "inputs per tetromino."
- How they influence surface stability.
IPT simply measures how many moves, rotations, hard drops, holds, and soft drops were needed on average per tetromino placed. Players generally perceive needing less inputs to be a more efficient way to play.
At first glance, middle Tetris column (MTC) stacking seems to save some button-pressing because the I-piece doesn't have to go far. In reality, MTC stacking is one of the worst methods when it comes to IPT. The reason is how it affects the number of inputs of other tetrominoes.
With side Tetris column (STC) stacking, some placements will naturally require no movements and little to no rotation.
These types of placements are good for IPT. However, they are no longer possible with MTC stacking.
The player must move these tetrominoes away from the spawning position. This creates a need for more inputs. For this reason, STC stacking is more efficient as far as IPT goes. For more information about this concept, check out Evaluating the 6-3 Split.
Surface stabilityThe best surfaces are able to fit any Tetromino. With good surfaces, a player is able to avoid making gaps and holes. By splitting the surface into two islands, MTC stacking reduces the total number of placements possible that do not create gaps or holes. To show this, let's consider a typical stack for both methods. We will compare how many placements each allow that do not create gaps or holes. Each example will contain 40 filled cells to allow for some surface bumpiness.
Example Side Tetris Column:
(43 total placements)
Example Middle Tetris Column:
(40 total placements)
Despite both surfaces looking as much like the other as possible, the MTC stack lost three placements. It may not seem like much. However, the less stable the playfield, the more those extra placements count. With a larger sample size, I'd expect MTC would be even worse than this particular example on average.
Where did those missing placements go? The T-, J-, and Z-pieces all lost one placement each.
For an explanation, let's think back to a concept explained in the Upstacking article: surface connectivity. The MTC stack has two disconnected surfaces: where column 5 meets column 6 and where column 7 meets column 6. The STC has only one disconnected surface: where column 9 meets column 10. All three of these lost placements are due to the lost connection of where column 4 and 5 would've created a step against column 6.
This is not to say that there are no other ways to evaluate MTC and STC. Other factors include, for example, high gravity play and creating skimming options. That said, I would recommend STC stacking to a newer player before experimenting with other methods.