Breaking Bad Habits in Fighting Games

Patrick Miller
7 min readMar 27, 2023

[This essay was funded by my generous Patreon supporters. If you liked this and want to see more, please consider joining the crew!]


Really appreciate all the work you do.

I’m primarily a KOF player and I have a problem. Yashiro’s 214P is his main combo starter from lights and heavies. It’s also really unsafe on block. I practice confirms in training mode, but during matches I autopilot into the move in a blockstring or drop the confirm and I get blown up.

The confirm drops are an execution issue, but what are some ways that I can try and correct this (and other) ‘autopilot’ habits in my play? The easy answer is “simply do not do the special move,” but that doesn’t feel very actionable.


YashirOh No

This is an excellent question, and also a great excuse for me to write a bit about parts of the learning process that I have talked about often on stream but haven’t really covered here. This is some really cool stuff that digs into some nuts and bolts of how we train to play at such high speeds, so thanks for sending it along!

Your intuition that “Do not do the special move” doesn’t feel particularly actionable is right on the money; it’s a high-value tool, so of course you’re going to use it. “Use the special move less” might get you slightly better short-term outcomes (fewer attempts -> fewer failures -> fewer punishes -> fewer Ls), but it’s not easily actionable either and doesn’t really address the issue at hand. “Use the special move less” treats the special move as a risky bet, but it shouldn’t be as long as you’re not autopiloting it.

BureBa deserves an anime adaptation.

When it comes to breaking bad habits, especially autopiloting habits, I find it much easier to frame the goal as a positive action statement instead of a negative. “Don’t do” or “do less” doesn’t give you something to do instead, and so all that usually happens is that you do the thing anyway, but feel bad about it because you weren’t supposed to do it. So we’re going to break up this autopilot by finding some branching points where you can focus on doing the 214P when you should be doing it, and do other stuff instead when you shouldn’t be using 214P. But first:

What is autopiloting?

For those who aren’t familiar with the term: “Autopiloting” refers to the experience of doing something in a fighting game seemingly without thinking about it. Sometimes this is a good thing! Strong players often cite the value of developing a strong autopilot as an important part of consistently getting out of pools in major tournaments; if you’re strong enough with your character to just run over most average players without having to dissect their individual quirks and tendencies, you’ll be less mentally fatigued in the later rounds. But more often people experience autopiloting when they’re watching helplessly as their hands do an absolutely awful thing that their brain knows is not a good idea.

If you’ve been reading my other essays, you may be familiar with the concept of an “OODA Loop” that I use for describing the cognitive work that players perform while playing a fighting game. The short version is that we can describe the real-time thinking that the players are doing as a series of phases in a “turn”; we Observe the situation on screen, Orient that observation into a readable gameplay situation, Decide on what to do based on that situation, and then Act out our decision, then Observe the outcome and do it all again.

Within the framework of an OODA Loop, “autopiloting” is essentially an issue created by failing to orient properly; you may be literally looking at the screen, but the information you’re taking in isn’t refined enough to yield different decisions besides “Do the special move and hope it works”. So in order to level up this part of your game, we’re going to do a little research to break down the big situation into a bunch of smaller ones, drill them, and then look for key points to focus on getting the special move out when you want it.

Breaking up the autopilot

We’re going to start you out by doing some homework. Check your Yashiro combo routes to make sure you’re doing the right ones for you; I’d say do whatever routes give you the most time with the least execution challenge, and if there are optimizations you want to make, focus on them after getting comfortable with the easier ones. Next, make sure that you’re using the right routes for each situation where you’d be landing a hit; if strong Yashiro players usually use different buttons for a footsies confirm than they would after a jump-in or a short hop or a meaty or whatever, there’s probably a good reason for that, so try it out and see if it makes things easier.

What this process should do is help break down the autopilot by adding more nuance to your situational understanding. Maybe your route needs to look different in situation X vs situation Y because one situation is easier and slower than the other; maybe you can find some easy option selects in certain situations that can simplify the challenge so you don’t even need to react to anything at all.

I don’t play enough KOFXV to know what his routes look like, but I took a quick look at his Dreamcancel page, and it looks like most of the combos into 214A give you 2–3 light hits to confirm into, and the combos into 214C give you at least two hits to confirm into, which should be plenty of time to visually confirm the hit/block. So now you need to figure out something to do on block that will still give you an advantage of some kind, like canceling 6B into his 41236A, which is safe (if minus) and mixing that up with 41236C if you want to catch someone mashing. A quick consult with KOF-playing homies MechaMacGyver (aka Xiao Guey) and Killey validated these options, and they also suggested that for strings starting with close heavies you can simply end the string on a command normal to keep things safe if you have to.

Once you’ve got your situations mapped out, use them as drills to practice hit confirming. Set up the situation, set the training mode bot to Random Block, and practice doing 214C on hit and 41236A on block. Grind them out in sets of ten like you’re practicing free throws or something. Do a couple sets, see how many you got right, check to see if there’s anything you can change about how you move your hands or look at the screen, and do a couple more sets. Try it the next day and see how your results change over time.

While you’re drilling, try to identify the point of no return — that is, the latest point in a situation where you have to choose to special or not to special. Once you’ve found that point, pick a spot on the screen that you’re going to look at to get the information you need to confirm the special input. Maybe you’re looking at a hit or block spark, or a certain pose in a hit reaction, or even a UI element; for example, high-level SF players will often use the stun gauge to hit confirm because the feedback is quicker and clearer than the actual character interaction, and Guilty Gear players will react to the Burst Gauge emptying faster than they will the actual Burst animation. Try looking at different parts of the screen and see how they affect your success rate. When you find a spot that you feel you can react to, keep on drilling it. Your eyes and hands will begin to treat this as part of the situation, so your brain won’t have to be as involved.

Over time, you will likely see overall improvement over the first couple weeks, then you’ll hit a plateau that is higher than where you started across the board, but some situations will be harder for you than others because they have tighter timing or a heavier mental stack load. You will also probably find that your success rate is actually lower in a live match than in training mode, which is normal, because a live match gives you a lot of things to think about besides just doing this specific thing. There will be some situations where you are more confident and others where you are less confident, and you’ll have to decide whether you want to a) do something less risky in those situations, b) grind it out and hope there is room to continue to improve your success rate, or c) just accept the risk and prepare yourself to eat punishes sometimes.

Doctor of DDTs MechaMacGyver suggests holding 214C only for guaranteed punish situations where you know your hits will land before you press them, so you may want to focus on those situations as the highest-value ones before trying to add riskier situations to your game.

Hope this helps! Thanks for reading.

Cohost, Twitter, FGC.Network Mastodon| read my words

Twitch | watch me play

Patreon | support my work


-patrick miller



Patrick Miller

a little bit miyamoto musashi, a little bit yoga with adriene.