So recently, Broski made a video where he had two drive rush "input eating" issues in quick succession:
https://youtu.be/vexNGBEhQvQ?si=2TDgBfcx3Oe1_rTb

Skip this section if you don't care to fully understand the mechanic:


This confused me because my understanding of general drive rush "input eating" issues can be classified into two forms:

Priority Resolution
  • If two inputs are simultaneously valid at the same time, the game has to choose a single action to output on the next frame. The priority resolution in this game is:
  • Taunt > OD > SA > DI > Throw > Parry > Special Moves (more detailed priority within special moves) > Command Normals > Normals (Kicks/Higher strength buttons have more priority)

Example of an issue caused by this:

  • The screen freeze lasts for 10 frames
  • Let's say you input stLP then MP 7 frames after as Luke, all within the drive rush screen freeze because you intended to do your stand jab target combo. The game will output stMP because it can only pick one of these actions to come out on the next frame and higher strength buttons have higher priority.
  • This does not match with the player's intended behaviour, but there is no "fix" for this. Fighting games are a two player game where the opponent's actions will affect the player's, which is where the technique Option Select is born, you just have to live with this as long as drive rush has a non-reactable screen freeze as a part of its mechanic
"Shifting the Timeline" of Inputs
  • The screen freeze essentially shifts your input ~10 frames into the "past" compared to when you intended to do an action

Example of an issue caused by this:

  • You are spamming crMP with Luke and you are using the 4 frame buffer to get the move to come out as soon as possible (pressing MP within the last 4 frames of recovery of your last action)
  • Your opponent drive rushes 10 frames before your move recovers, and you input another MP on frame 7 of the drive rush screen freeze. If the screen freeze didn't happen, you would've input MP 3 frames before your move recovers, which is valid for the input buffer, and your next move would come out at the end of recovery. But, because the game stalled, you are still inputting MP 10 frames before your move recovers, which is NOT valid for the input buffer, so your move does not come out
    • This gives the impression that your input was eaten, but it wasn't, you just inputted an action in the middle of a move (when an action isn't valid) instead of at the end of a move when you intended
  • An addendum will be at the end of the video to visualize this, go to the end of the description for potential fixes for this

A common issue people will cite, related to this, is your jump not coming out. This is because directions/movement is a non-bufferable action. The player can in general easily solve this by just holding the jump button for a split second longer. When you choose to jump, you are choosing to relinquish all control of your stick for almost a full second until you land. So if a Guile is full screen and you want to neutral jump a fireball, and there is a chance he will drive rush right before the fireball would have connected, to mess with your inputs (which is a legitimate offensive strategy in the game), you can just hold up for very slightly longer and you're OSing if the Guile drive rushes or not. This also applies to parrying projectiles, you will be in blockstop/blockstun for a bit, so no reason not to just hold parry for ~10 frames instead of tapping for the shortest possible duration.

This is not a new concept in fighting games. We always do things with our inputs to mitigate the risk of something bad happening. Techniques such as double tapping exist to combat low to no buffers, we input walk forward fireball as half circle to not get dp, and we input moves with back/down back to be blocking in case something goes wrong to name a few. This is just another instance where it's up to the player to adapt to the game if you want to get better and have greater consistency. Adding a buffer to jump would lead to other issues regarding priority resolution as mentioned above, especially with sloppy inputs containing multiple jump directions, adding a buffer to directions is not the answer.


Broski's examples did not necessarily look like a problem stemming from the two classes of issues mentioned above, and there was a common variable with trying to do an action after Aki's light lash, so I decided to investigate the replay. Turns out, there was another common variable in these two examples, he input his button frame perfectly on the first frame of Cammy's drive rush screen freeze (frame 4 of parry).

Through further testing, I found out that during this single frame, the input buffer does not exist for some reason (likely a bug). So, the drive rush screen freeze does not eat inputs, it eats the input buffer to be exact (for a single frame). Note that this is extremely hard to replicate because you need to both be within the buffer window of an action AND frame perfectly press your input on the first freeze frame of drive rush. This likely does not happen that often, but may occur sometimes during reversal/minus frame situations such as in Broski's video.

To be clear, if you were in neutral and tried to check drive rush and failed, it was 100% your fault. There is absolutely zero input eating in this scenario.

Regarding the "Shifting the Timeline" of Inputs problem, you could potentially raise the buffer during drive rush to help with this, but this would also increase the window for Priority Resolution issues. For example, if you raised the buffer by 10 frames (the length of drive rush freeze), there could potentially be a giant buffer of 14 frames where you might hit 2 inputs at the same time and get something unwanted. Also, there are situations where this "problem" could be helpful. If Broski's heavy buttons came out in these scenarios, he would've gotten counter hit by Cammy's jab in both scenarios. So, the "input eating" actually saved him from making a bad decision trying to check drive rush during minus frames with a heavy.

I do think that the game would benefit by increasing the input buffer by 3-5 frames in these scenarios, however, I can't actually know how much it would affect things without extensive testing on a patch with that addition.

Edit

Pub: 15 Jan 2024 23:23 UTC

Views: 141