Quick Maffs.

Out of curiosity I counted actions in two ways from this loop:

Actions movelist-wise for one cycle of the loop: 9.

These are the actions that create a significant change on the screen, so to speak, regarding the character.

Style Switching isn't counted unless it's locked-off and used to change Dante's animation (as an alternative way to taunt, or changing his position to dodge the attack of an enemy).

Switching weapons isn't counted per se either, even if Dante's models change.

What's counted are moves from devil arms and gun weapons, enemy step (as it affects animations, positions, inertia, etc.) and taunts.

Actions inputs-wise for one cycle of the loop: 17.

These are the actions regarding inputs. Switching weapons, styles, lock-off->on, pressing face buttons and moving the left stick are counted. A backwards-forwards move would count as 2 actions (each direction=1 action), while a move just backwards would count as 1 action, even if you have to go back to neutral/leave the left stick (because it would be analogous to press and release the melee button or style button; I count it as one action to simplify).

One could define an efficiency quotient of sorts dividing the first variable by the second one:

9/17=0,529.

This means that for each action inputs-wise (or each input), we get on average 0,5625 actions movelist-wise (what we see on screen in terms of animations from the char). 

The opposite would be 17/9=1,88 => One needs to do 1,88 actions on average inputs-wise to get 1 action movelist-wise.

We could use EQ (for brevity) of different techs or loops to compare them.

Sequential inertial raves backwards would be a 3/6=0,5 (for one cycle).

Sky running a 6/12=0,5 as well (from sky star to a second flying guard).

A Full House - Rave loop (one cycle) would be again 3/6=0,5 as well. If backwards: 3/7=0,42.

Inertial Rain Storm forwards from rave and Sky Star (DMC4): 4/6=0,66.

Shotgun JC into Jealousy: 3/4=0,75.

Of course you also have exceptions like gun raves and a simple JCd rave into rave (one cycle), which would be 2/2=1.

We could also use them to see if proposed systems that remove style switching are really better.

Some random example from reddit:

Individually we can just compare the absolute values of the different techniques or loops. For example, the loop shown at the beginning would still require 17 actions inputs-wise (I added a made-up input for Dark Slayer holding guard+gun modifier, since this was proposed for DMC5).

This means this system doesn't provide an improvement for the loop, because the advantage of not having to press any style button is lost with the gun modifier, which is a con per se anyway, adding one more necessary input to simply shoot. Something that needs to be done asap in many situations.

One could, however, still calculate several quotients (not just count the absolute values) and use different criteria for a better visualization in the comparison.

Sequential inertial raves backwards: 3/5=0,6.

Sky running: 6/9=0,66.

A Full House - Rave loop: 3/6=0,5.

Inertial Rain Storm from rave and Sky Star: 4/5=0,8.

Shotgun JC into Jealousy: 3/5=0,6.

Gun raves: 2/3=0,67.

Rave JC into rave: 2/2=1

An obvious criterion would simply be the average of both systems:

- Normal Style Switching System's average EQ: 0,701.

- Proposed Alternative System average EQ: 0,69.

Another criterion could be the Standard Deviation, which tells us how dispersed are the values of a sample in regard to the mean. And it's calculated as the square root of the average of the squared differences between each value and the mean.

In our case it would be N-1 instead of N because it's a sample and not absolutely all the relevant techs/sequences (it's a type of correction for using a sample).

In any case it can be quickly found with this calculator:

- SD Normal Style Switching System: 0,225.

- SD Alternative System: 0,164.

The criteria could be then:

- Average: how many moves we get on average per input. The higher, the better.

- SD: it shows if the necessary inputs are more or less uniform, or if there are techs with very high efficiency (potentially very easy) and techs with very low efficiency (potentially too hard). Basically, how extreme are the values. The lower, the better (in this case the alternative system seems slightly better for this sample, even though the difference and size of the sample are too small to be very relevant).

Overall I don't think the alternative system would be worth it. Even if we consider ideal cases like Guard-Raves, because it makes some stuff that is already easy/relatively easy/doable even easier, at the cost of making other stuff that should be done quickly and easily unnecessarily slower or harder.

If we also take into account having to switch weapons with the d-pad, the necessary precision to JC into R2 (or R1) as Guard, etc., then it's even more obvious.

All this was just a quick example that isn't supposed to be too rigorous and there are many variables that we aren't considering.

For example, we could try to operationalize visual impact or relevance (functionality) of different sequences or parts of them instead of simple moves from the movelist. As example, a rave backwards would have a small multiplier for visual impact vs a regular starrave.

Or hand-gestures instead of raw inputs counted individually. For example, if you have jump on L1 and melee/style as face buttons, you still need one gesture with each hand for jump cancels. Meanwhile a typical X to Circle jump cancel requires one gesture with the thumb for both inputs.

Another example is how I move slightly my left hand to press at the same time Sword Master and the shoot button (on L1 for me) in this sequence after air trick:

Moving my hand slightly in the direction of my body makes me press both buttons without having to even think individually in pressing L1 with one finger and the right d-pad with another finger.

What about other combat systems...

If we considered other Action Games, the quotient wouldn't be 1 either, mainly because of directional inputs, but overall it would probably be higher.

The low efficiency in terms of inputs in DMC4/SE is a price we pay for having a system with such levels of interactivity. Basically, it's the best system to keep a significant quantity of moves, while being able to execute them in most situations (vs an extensive movelist where too many moves can't be used at any given moment).

Having to do almost twice the inputs vs executed moves from the movelist comes mainly from Style Switching and Devil Arms/Gun Weapon switching. But if we tried to remove them, to keep the same techniques and loops possible, we would need to sacrifice something else that would lower the quotient of efficiency anyway. Similarly to how the person from reddit had to add a gun modifier and use 2 triggers for 2 types of style actions.

Sometimes the alternative systems reach absurdity and make certain techniques or moves impossible to execute.

What about APM...

Not that relevant when comparing systems for the same game in this scenario. At most we could try to calculate minimal APM ranges (movelist-wise) for certain techs. If an APM threshold is very high for a certain loop, set-up or tech (or else it wouldn't work because the enemy would fall, for example) and close to human limits, then we should try to design the system in a way that the efficiency inputs-wise is very high for the mentioned loop or tech too.

Obviously only if the side-effects don't make other important techs way too hard. At some point some sacrifice has to be made... Inertialless raves are very hard, but if we removed style switching to increase the input efficiency just for them, we would mess up other more important things.

Something interesting about APM and other games vs DMC is that, despite how low is the efficiency inputs-wise for DMC4, the APM (movelist-wise), or what you see on the screen, is still probably higher. Talking about high level gameplay in both cases.

This is, again, because of how interactive the system is. Thanks to the different mechanics to cancel moves <=> weapon switch <=> style switching, the player isn't slowed down or plainly stopped by the system itself.

If you were curious, the APM movelist-wise of the first loop is 470 ~. If it was possible to do the loop during 1 minute, 470 moves would take place in that lapse of time including enemy step (313 without enemy step). The APM inputs-wise would be 886.

As I have said several times, APM is just one of the factors that can affect difficulty. There are sequences with lower APM that are harder than this loop because of other factors.

And enough rambling...