Friday, May 20, 2016

Neutral and Frame Advantage

Neutral and Frame Advantage

When I was writing my initial brain-sorting on neutral and positioning (that this document is in spirit an addendum to), I was intrigued by the primacy of positioning but frustrated by its incapacity to describe mixups beyond generalized footsies. There’s an additional timing/movement-based decision mechanism that was a struggle to identify but I believe that Relative Frame Advantage reconciles Melee movement with FGC footsies.

- - -

Frame Advantage/Traps

In Street Fighter frame advantage refers primarily to the relationship of blockstun to endlag. If you hit a blocking character with a move with enough blockstun and little enough endlag, you can perform an action before they can. The greater your frame advantage, the better your options to combo or counter an opponent’s action out of blockstun. Naturally, a move that is negative on block can get you punished. Shieldstun and endlag work very similarly in Melee. Very few moves are positive on shield in Melee, but because our possible actions are restricted to a handful, we tend to look for things that are better or worse than -7, the threshold for a guaranteed shield grab. Anything better allows for a mixup with a jab/dash/crouch etc. designed to beat the grab.

Due to differing game mechanics not worth divulging here, these post-hit mixup scenarios work a little bit differently in Street Fighter but are called Frame Traps. Frame Traps are normally straightforward mixups based on how positive a move is on block. Generally speaking, the greater the frame advantage the greater the potential punish off a correct read. That being said, there is some amount of nuance to the system and this is where things get interesting. Some moves or spacings push the blocking character too far away to punish out of block directly. Let’s take for an example Yun’s light shoulder in SF4. Even though it is technically negative, it is safe on block because the spacing creates an interesting scenario where any immediate punish attempt will whiff. If the extra space/time for a slight walk forward needed to punish a poke exceeds its endlag then it can be considered to have effective/relative frame advantage. Although this advantage is indirect it is very real. After a light shoulder Yun can basically do whatever he wants (normally to wait and punish a potential whiffed punish or walk forward/jump) as if he had a frame trap himself.

This setup is very interesting but it isn’t actually that unusual. The only change from a normal frame trap scenario and this backwards sort of one is the additional space.
This space keeps Yun safe by artificially subtracting frames from his opponent’s relative frame advantage by requiring that it be used for movement.

It’s not really any different than had the two arrived at this timing variation from full screen as opposed to a block. What is important is that we are beginning to consider movement as a time commitment as vital as startup.

Druggedfox has previously expanded on frame advantage as it applies to playing neutral beyond whiff punishing laggy fsmashes. In an option vs option scenario, say Falcon’s nair vs Shiek’s ftilt, Falcon requires a certain amount of frame advantage in order to protect his movement to position, jumpsquat, and nair startup (a minimum of 11f for nair vs Shiek’s minimum of 6 for ftilt, so at least +6). In order to land a nair without fear of it being beaten by ftilt, Falcon must anticipate Sheik’s committing to some kind of lag (be it a whiffed move, a WD, a dash, etc) that will provide him with at least a 6f head-start but as soon as he recognizes it the nair is guaranteed to beat an ftilt*. Naturally, if Falcon does not have relative frame advantage then committing to the nair is unreasonably unsafe given the risk/reward.

*At this point it’s important to note that while it is impossible to discern exact frame advantage in the moment as competitive players we are very very very good at approximating it. Over the course of however many ungodly number of times we’ve experienced it our guts have learned to recognize when a shieldgrab is guaranteed or when a nair is unsafe. Examining the specifics will not only help us fine-tune our impulses but to use them to our pronounced advantage.

Consider how this relates to Opportunity Cost. If Falcon thoughtlessly dash dances, WDes in place, retreating nairs, or whatever other unhelpful rhythmic habit he’s picked up then the opportunity to recognize and to seize his +6f relative frame advantage to get a nair will pass him by.

Relative frame advantage gives us ways to reliably win even unfavorable option vs option fights. It should be the chief methodology in neutral after accounting for stage position. But relative frame advantage applies beyond specific move vs move scenarios. I will expand on two related concepts, Scrambles and Tempo.


For the purpose of this conversation, scrambles are situations that are a) unreactable and b) that neither player has prepared for. They are often accompanied by a harried feeling. One more common scramble is Fox accidentally undershooting a nair outside of grab range. The opponent can’t directly punish the nair but can make one decision at the same time that fox comes out of his landing lag. This is of course reminiscent of Yun’s backwards frame trap, but in this instance the relative frame advantage is not deliberate. Because scrambles are unreactable and unintentional any decision made within it will be made on impulse. You can look at scrambles with the impulse guidelines below, which should prove revelatory on several counts.

1) What a player does first he will do most often.
2) If you directly punish something, they will not do that again in the next instance (if the event passes unrecognized then ignore this rule).
3) Players will use ~3 options at most in any given situation.
4) Decisions are prompted by position/relative frame advantage.

This is exactly how Samus players get those magic grabs out of nowhere. Because the grab beats spotdodge (a common habit in scrambles) and dashing away might be restricted by timing/stage position a Samus player just has to confirm that their opponent does not jump out of scrambles. As long as jump is off of the table grab will beat most other options. Its success rate goes through the roof.

#4 is the real kicker, though. This is an observation that due to its nature isn’t precisely provable but I’ve paid very close attention over the past month or so, watched videos, and asked a few good players to confirm. I am now confident in stating that actions out of scrambles are impulsive and based on frame (dis)advantage, not the situation itself. A scramble is a scramble because it is a strange, unintentional situation requiring an immediate, habitual decision. This makes it difficult to recognize, so guideline #2 rarely applies and he’s very unlikely to mix up between prepared options. Yet we recognize that actions out of scrambles do vary. In minor scrambles maybe the Fox player will dash back, in major ones maybe he spotdodges. The decisions that are being made are responses to the amount of danger unconsciously assessed which we can measure in frame disadvantage. Usually the bigger the disadvantage the more dramatic the decision (e.g. light frame disadvantage = dash back, heavy frame disadvantage = spotdodge) but there is no difference between similar frame counts. If fox makes a habitual decision to spotdodge at an about -18f relative disadvantage after a shallow nair then he will make the exact same decision to spotdodge at an about -18f relative disadvantage after a whiffed dash attack. It’s important to note what that decision is (the punish on a Fox that habitually shields when very nervous is completely different from the punish on one that spotdodges) but understanding frame disadvantage, not the situation itself, as the motivator for the decision dramatically decreases the complexity of perceiving and punishing Fox’s habitual decisions out of bad situations.

Naturally, these instances of certainty will decrease in frequency as a player increases in skill i.e. familiarity with situations. It does however have incredible benefits beyond the occasional punish out of a scramble. Because relative frame advantage includes the amount of space that a character can travel in x number of frames, you can use these principles to track and to anticipate decisions made in neutral exchanges on the borders of reactable (~20f, which in terms of space varies dramatically by character). You may notice and even compel responses to soft/hard frame advantage by intentionally changing your levels of commitment. If they dash back every time then do it on purpose and take the stage. Etc. This points to a very interesting methodology for punishing not just a move choice but a timing and movement choice.


Tempo is a chess term that I’ve seen referenced in a Smash context but never sufficiently expanded on or married to melee outside of a few personal conversations I've had with Frootloop that inspired this entire exercise. In chess, every move costs exactly one turn unit of in-game time. If by way of some tactic you can achieve a desired effect in two fewer turns than with another tactic then you reap the benefit of having two extra turn’s worth of time, which is obviously incredibly valuable.
In Melee our commitments vary in duration, but they are all still time commitments. I’ve written about the how even a safe move can be costly in terms of opportunity. A retreating bair might go unpunished, but it costs so much time (jumpsquat+startup+endlag+landing lag) during which you cannot do anything else that you might be giving up an opportunity to grab etc. In addition, there’s a spatial cost. A retreating bair surrenders space to your opponent. This is space that they can use to dash dance as well as space that will take even more time for you to cross over again, should you need to in the future. Situationally, a WD away might achieve the same defensive objective as a retreating bair with a very small fraction of the time commitment. Managing your time commitments relative to your opponent (Including in relation to the space between you) is Tempo. In this way it is another iteration of relative frame advantage.

Using a tempo advantage to pressure the opponent into acting first is called Initiative. In a way you could see it as forcing them to throw in RPS on “scissors” while you have the space/time to wait until “shoot.” The clearest examples in Melee involve severe stage advantage being used as a resource for frame advantage. Consider Puff’s edgeguard on Marth below the stage. Puff sits on the ledge and simply reacts correctly to Marth’s recovery options. Because his options are limited to DJ, sideB, and upB, none of which are in this case very threatening, it is not very difficult to flowchart a series of reactions out of ledge invincibility to ensure a 100% edgeguard rate off of Marth’s compulsion to act first (i.e. before he falls too low and dies). Positional advantage manifests very similarly as initiative by restricting options and compelling the opponent to act first. With stage/frame advantage many mixups dramatically favor the center by way restricting/enhancing access to space to retreat. That is, if they need be played at all.

Many situations in Melee are flowchartable. Picture again Yun’s light shoulder. Yun gets effective frame advantage for connecting a light shoulder even on block. He can potentially use this opportunity to force his opponent into the corner, limiting their options and perpetuating his tempo advantage as initiative. If Yun plays well then his light shoulder can = initiative. Similarly, any situation that nets Yun a light shoulder on block can = initiative. Yun doesn’t have to wait around for initiative to be given to him, he can actively pursue it with light shoulder on block and confidence in his ability to consistently convert from there! Maybe he uses light frame advantage from waiting out of range of a poke instead of getting scared and whiffing an attack to then walk forward and force a block. It could be that simple.

Actively playing with tempo is managing your active threats, your active range as potential, and your vulnerable lag to engineer stage and frame advantages. While it is impossible to avoid commitments completely it is possible to orchestrate their placement. If you land further away than fox can run during landinglag + dash etc then you are safe to land. However, if you land in a space that fox can actively threaten then you land with frame disadvantage. If an approach is spaced such that a WD can avoid it then WDing saves valuable frame advantage relative to jumping. Avoiding habitual or superfluous movement rhythms allows you to better look for and take advantage of overcommitments. You can use a small amount of space to protect the startup of a jump+bair in order to use it to purchase more space. You can jump empty to use the threat of bair to purchase space without actually committing to a bair. All of these are active uses of time commitments and ranges to engineer ever greater advantages.

No comments:

Post a Comment