Let's talk about Melee inputs.

More specifically, the Inputs / Minute (IPM) statistic you see in SCL post-set stats and in the Slippi desktop app when looking at your replays.

First thing to note: this statistic is very often compared to Starcraft's Actions / Minute (APM) statistic which is used to indicate how "fast" a player is.

In Starcraft, an "action" is very discrete:

1. Tell units to go here
2. Select units
3. Start making a marine
4. etc

In Melee, what is an "action"?

I specifically decided to move away from this word because right now we do not have any way to quantify it.

What we do have, however, is all of the buttons and analog inputs the game has read. So let's analyze those instead.

Button presses are extremely obvious: press a button, +1 input.

Stick inputs are harder. There's a million ways you could count them but @Kadano and I decided a long time ago to look for transitions between zones.

Each transition: +1 input
Returning to deadzone doesn't count

The triggers in Melee are also analog. Currently the analog input should count +1 input when going past the 30% pressed threshold.

Triggers also have a button input all the way at the bottom which counts for 1 input like any other button.

Now I said "should count" in the previous comment because in preparation for this I realized there was a bug in the input counting logic causing the analog portion of the trigger to never count.

This means light trigger presses currently do not contribute to IPM.

This will be fixed in the next release.

Potential issues/inaccuracies with current methodology:

1. Mashing/rolling the control stick can lead to many "undeserved" inputs
2. Bottoming out a trigger as fast as you can counts as 2 inputs (one analog, one digital)
3. Is releasing a button an action in some cases?

I mention this because I'm well aware that the current methodology is not perfect. It is only *somewhat* usable in comparing different Melee players.

It is completely inadequate, however, for comparing (in a direct sense) to numbers obtained by other games. Please don't do this

All that said, let's take a look at the input breakdown for one of Ginger's games against Mango. Using the (hopefully) bug-free calculation, Ginger clocked in at 656.8 IPM.

Ginger made:
834 joystick inputs
36 c-stick inputs
406 button inputs
73 analog trigger inputs

So obviously you can see that the joystick is the main contributor to IPM.

That said, even without ANY analog inputs, Ginger pressed 198 digital buttons per minute.

No matter your qualms, it's quite clear that Melee players do press a lot of buttons

As to why the IPM values from HTC Throwdown in 2015 seem to be lower than recent values, I'm really not sure.

There's definitely a bug in the IPM calculation logic from back then but I would have thought it would lead to higher IPM.

The 2015 logic: https://github.com/JLaferri/HardwareEnhancedMelee/blob/3945bcdd15a7cb4a83342d4e7cb78c7e57bb3d59/Arduino%20Due%20Source/EnhMeleeDueFirmware/EnhMeleeDueFirmware.ino#L515-L532

As a reminder, everything involved in the computation of stats is open-source.

If you have an idea for an improvement, join our Discord, talk to us about it, and work on a PR.

Here is a link to the logic which counts player actions in a game: https://github.com/project-slippi/slippi-js/blob/master/src/stats/inputs.ts#L80-L117