Key Concepts

Key Concepts
- - -


Reaction Time

Human reaction time determines how Melee is played. It, more than probably anything else determines the actual level of risk attached to any given action in neutral.

While a personal peak reaction time is genetic, it’s better to term that hard-coded peak as a reflex. In this context, a reflex is distinct from a trained reaction. In Melee, how quickly your knee jerks when tapped doesn’t matter. How well you’ve trained your brain to press Z upon recognizing tech in place does. That is to say, your personal trained reaction time is going to depend on training, on your level of focus, and on time of day/sleep/diet well before your DNA comes into play.

Reaction speed is accelerated with the presence of other sense information i.e. you will react faster to a move with a clear sound effect than to one that is silent.

The other relevant concept in regard to reaction time is Simple vs Complex Reactions. In short, if there’s only one possible stimulus to react to (Simple) then your reaction time will be better than if there are multiple possible stimuli (Complex). Accounting for console lag, you can reasonably expect your simple reaction time to hover around 16 frames or so, maybe better, maybe worse. You should test yourself to see. Complex reactions are a bit more complex. The more options possible, the slower your reaction will be.

However, you can often stagger reactions to make a complex reaction simpler. For example, if you are reaction techchasing, rather then reacting to one of the four possibilities at the same moment, you can stagger them according to the time needed to punish, resulting in a sort of setplay in which you react to “tech in place y/n,” then “missed tech y/n,” and finally “which way are they rolling?” This kind of staggering will render a borderline impossible complex reaction a manageable (if still difficult) series of simpler reactions.



Frame Advantage

Frame Advantage is basically “which character can move first?”

There are two main uses for the terminology. First, how hitstun or shieldstun lines up with endlag determines how safe something is on hit. The basic metric is if something is better or worst than -7 frames. If an attack is -7f on hit then it’s shieldgrabbable. Obviously Shine OoS and other OoS options have other values to consider. CC works similarly and normally has better frame advantage than a shield.

Second, you can use frame advantage to describe “which character can move first” in neutral situations. For clarity’s sake I call this Relative Frame Advantage. As characters move around they commit to lag. If one character wavedashes but the other doesn’t, then the character that stood still is not in any lag, whereas the wavedasher has a small amount. That small frame advantage may or may not significantly alter what following options are safe/unsafe because you get that much extra time to do what you like at no risk whatsoever. This concept gets more complicated— but is still intuitively appreciable— if you consider that space takes time to cross. Similarly to how an opponent can’t contest your startup etc if they’re at significant enough frame disadvantage, they also can’t contest you if there’s so much space for them to cross to get there. In regard to frame advantage, space = time.



Unreactable Zone
If we consider space = time with your reaction speed, we get a concept that I call the Unreactable Zone. This is easily illustrated with Fox’s nair. Let’s say that you have a hypothetical average reaction time of 20f. Fox’s nair has a startup of 4f, +3f of jumpsquat. That means that Fox can potentially run/travel for about 12f and nair before you can react to his movement. If you boot up debug mode and look at how far that is, Fox’s reach at 20f is the edge of the Unreactable Zone. If you’re inside of that zone then you can’t react to Fox’s movement before getting hit. If you’re outside of it then you can.

The interesting thing about the Unreactable Zone is that, because of Relative Frame Advantage, it’s fluid. That is, if fox is in 13f of lag from wavedashing, then the unreactable range is cut by 13f, giving you that much more leeway to play around it. We do this kind of mental calculation constantly, so it’s good to be aware and deliberate about it.



Reads/Reactions
There are several ideas to keep in mind in regards to when you can/should guess and when you should use your reactions.

Every time you guess you run the risk of guessing wrong (Frootloop says "Tricking someone always loses to not getting tricked."). Every time you guess wrong you open yourself up for punishment. As wrong guesses stack up your opponent’s winning chances dramatically increase. If you want to win a tournament you have to win a large number of sets in a row. In order to do that you need to be winning a lot more than 50% of exchanges. This is not the case for reactions. You can’t react wrong, you can only react slowly— and if you’re playing well you’ll be able to notice that the reaction is slow in time to stop yourself from overcommitting. This renders reaction punishes much more consistent, i.e. more winning, than reads. All in all it’s generally best to heavily favor reaction punishes over reads.

That being said, there are really two kinds of reads. The first kind is just a raw guess. The second kind is a simple reaction, a sort of half-way point between a reaction and a guess. An easy example is puff reading a roll. Puff can guess that option and timing and press downB as if she had her eyes closed OR she can choose to treat it like a simple reaction, get into position, and press downB on confirmation of the roll. If the opponent does anything else then puff won’t be able to punish and it’ll likely reset to neutral, but this way puff can cover a common option with little to no risk attached to guessing wrong.

Finally, reads/reactions are again complicated in an interesting way by the Unreactable Zone. If you’re at or inside of the Unreactable Zone, you obviously can’t react to all actions or movements. As such, you frequently (but not always) have to guess at what will happen. You have to abuse your better understanding of the MU/the other player/game mechanics/etc and use better risk reward than your opponent to come out on top. If you’re outside of the Unreactable Zone then reads are almost always superfluous and stupid. There’s no reason to guess if you can just use your eyeballs.



Risk Reward

Risk reward is easiest to illustrate using reads but the principle carries over to other parts of the game. Simply put, you want to optimize the ratio of what could go right for you vs what could go wrong. A lot of that is determined by endlag/the duration of moves— A WD uptilt in neutral is safer than a raw upsmash because it has less endlag and is thus much harder to whiff punish, especially on reaction. Unless the upsmash has dramatically better reward attached, the uptilt has better risk reward and is the better play. Similarly, moves that have better disjoint or are for some other reason safer/less likely to trade generally have better risk reward.

The risk attached to a move is usually more or less obvious, but the rewards can be ambiguous unless you’re very familiar with your character’s punish game. Maybe the reward is just two hits, maybe it’s two hits into a 50/50 gimp. It’s hard to say without charting some stuff out. This is one of the reasons that it’s more efficient to develop a robust punish game before focusing on neutral.

Keep in mind too that any commitment costs time and has an opportunity/frame advantage cost. That cost matters a lot more often than you'd think. Doing nothing is usually better than spending that time committing to something low risk low reward.

All in all, risk reward a bit of a complex calculus, but is largely intuitive and definitely a fundamental issue. If a given situation is usually going to work out in your favor we call it EV+ (expected value positive). Stacking up EV+ situations is synonymous with stacking up winning chances.



Mixups

Proper use of mixups is best illustrated by contrasting it with the popular (bad) usage.

Popularly, mixups often refer to options that can be used seemingly at random in neutral i.e. things that you can do. “It’s good to do this sometimes.” This isn’t exactly wrong, but it’s definitely unhelpful. Properly, a mixup should be designed to punish specific options in common scenarios. If you do a mixup on shield then you should have written down on a piece of paper what they can do our of shield, what you can do to cover those options, and have struck out the superfluous/overlapping punishes. This way your mixup is a) informed b) always has really good risk reward and c) deliberate. It will ingrain good, relevant habits and stack winning chances.

I think that using a random “good” option at a random time is a bad mindset. It fosters desperation and can stand in the way of understanding. Instead, I think that you should be trying to proactively cover/punish options, preferably with good risk reward. The better your risk reward, the more situations that you understand to be EV+, the more winning chances you accumulate. Obviously you shouldn’t telegraph your intentions to the extent that you’re predictable, but if you actually understand the options/mixup scenarios that you’re attempting to punish then that’s not an actual concern. This is a distinct and strictly better point of view from “Mix it up. it’s good to do this sometimes,” and leads into the following article, Melee Principles.