Dispelling game speed myths: BlazBlue vs. Guilty Gear

With the release of BlazBlue: Chronophantasma for home consoles in Japan, there's been a glut of people importing the game and there has been something of a minor revival of the scene. The series has never had a particularly strong following in the United States; anime games are already fairly niche, and Guilty Gear was seen as superior by those who played them.

In particular, the major complaint levied against BB is that it feels massively slower than its predecessor. While there is some truth to that, common ideas for why it's true are actually incorrect, and I want to detail them here, as well as offering potential other ideas.

First let's look at hitstop. This is the big one, the most egregiously false complaint about the two games. For some reason, people think hitstop in BB is a lot longer than in GG. At least since Continuum Shift II (released in May 2011 for home consoles), this has not been true. Hitstop values were actually identical between the two games. And with the release of Chronophantasma, Blazblue actually has less hitstop than Guilty Gear does.

Attack attributes tables, courtesy of the Dustloop wiki

A quick primer for those unfamiliar with the inner workings of the attack system: in both games, all attacks have a specified "attack level", and moves with the same attack level have the same frame data in terms of how long the target is put into stun state. Individual moves have different knockbacks and can induce other states that are not covered by these tables. And of course there are individual exceptions.

A level 0 attack in BB is equivalent to a level 1 attack in GG and so on, in that jabs are level 0 in BlazBlue and level 1 in Guilty Gear, etc. What we see from these tables is that in terms of how fast moves in combos or pressure must follow each other, BlazBlue is actually faster--by a large margin in some cases. Its hitstop is a full three frames shorter.

We also see that attacks themselves cannot actually be slower in BlazBlue--since overall hitstun is actually lower, moves cannot, in the aggregate, be slower, or else they would not combo. However unlike GG, BB often makes use of the ground in combos, as there is no OTG penalty, and wallbouncing induces hitstun, allowing more liberal use of slower moves in combos.

Another oft-cited reason that BB feels slow is that combos are longer and do overall less damage; this is mostly true. The game's hitstun and damage proration systems work such that combos often last longer in real-time than in GG. Additionally, supers tend to last much longer and be more cinematic.

One caveat, though, is that the Guts system in GG makes combos earlier in the round appear to do significantly more damage than they do in effective damage. A fairly optimal Millia combo can take many characters from 100% health to 40 or 50--but due to the Guts system, this may only be one fourth or one third of the character's actual effective health. BlazBlue has no such system, but characters do have different health values; the highest health character has the lowest character's health plus over a third.

It's also worth noting that throws are much slower in BlazBlue, at approximately 13 frames. Guilty Gear has one-frame throws, but lacks the throw reject miss system that prevents mashing throw break in its younger sibling.

But really, the biggest, actual reason that the newer series is slower is the characters actually move more slowly. Guilty Gear has a gravity system: different characters fall at different speeds. This is still true, but to a much lesser extent, in BlazBlue. But overall, the average falling speed in GG is much higher. In addition, stages in GG are not nearly as wide, and characters run and airdash significantly faster, making the game feel more mobile and interactive overall.

To give some actual quantitative numbers, the average jump duration in frames in GG is 39.16, and in BB it's 40.34. While in a direct comparison of two characters that single frame doesn't make much of a difference, it can contribute to a more sluggish feeling. To give an idea of this, even when playing on GGPO, with a 2-frame buffer you can still feel that the gameplay isn't as crisp as local, it's very noticeable but generally inconsequential. Halve that and you have something that can at least contribute to a feeling.

Even more intensely, the average backdash duration in BB is 22.08 frames, and only 16.07 frames in GG. Air forward dashing is two frames slower in BB, air backdashing a full four frames slower. And since the stages are wider, this covers even less effective space. It is unfortunately not a simple thing to compare numerical run speeds, since the actual values used in each game are calculated differently.

So we've established that a huge reason for the game speed difference is in character mobility, but I want to add another idea, which is less quantitative: game situations change much more frequently in Guilty Gear than in BlazBlue. By "game situation", I mean classifying an instant of the game as being in neutral, one character pressuring another, and one character comboing another.

We've already established, informally, that combos are longer in BlazBlue. In addition, the higher character mobility and overall space coverage in Guilty Gear (look at Anji's Fuujin, for example; that thing covers like half of the entire visible screen) means that less time is spent in neutral. Which leaves pressure. The existence of faultless defense and its fantastic behavior as a kind of pushblock, combined with overall faster jumping and backdash speed, means that it longer pressure strings are generally not viable in Guilty Gear. Contrast this with BlazBlue, where barrier (though improved in BBCP!) is not nearly as effective as getting people out, and movement is somewhat more sluggish. Watch a Litchi working her pressure string to an unfortunate cornered opponent, or Carl or Relius.

Throws work this way too. Most characters in GG cannot generally combo after their throws; almost every BB character can.

So overall, Guilty Gear changes state much more often and much more quickly. This and the greater mobility in both time and space are the reasons the game feels so much faster. Combos being shorter is true, but not significantly so. Hitstop is actually higher than in BlazBlue by a large margin.

So then, how did BBCP come to be felt as so much faster than its predecessors? By changing the second factor. I mentioned above that barrier received a boost in the new revision so that it's a better pushblock; this makes longer pressure strings less viable. The addition of Crush Trigger as a guard break mechanic means the threat is much more present than with the guard primer system that existed before, meaning a pressure -> combo change can happen more frequently. Hitstun proration has been attached to real-time combo duration, reducing overall combo length. And the reduction in hitstop shortens combo and pressure strings further.

All of this serves to decrease how long the game goes without a significant change in state, leading to a faster feeling. Without a huge paradigm shift in how characters move, especially relative to the size of the stage, though, the game will never be as fast as GG.

I give this rating system three and a half thumbs liked

Rating systems come into play in a lot of a person's daily life. Hotel ratings, vehicle crash-test ratings, Yelp reviews, all kinds of systems exist for the express purpose of concisely expressing the quality of something.

I've recently been thinking about different kinds of rating systems and their pros, cons, and trends, so here I'll detail a lot of what I've come up with. Specifically, I'll be investigating tier lists, five stars, out of 10, and like/dislike.

For the fighting game fans, tier lists are the primary method by which characters in fighting games are rated. In a tier list, elements are placed into roughly equivalent categories based on their perceived quality. Importantly, tier lists usually assign letter grades much like the American school system, which is the absolutely critical part of the system.

People have an inherent idea of what kind of quality deserves what kind of grade. Characters deserve an A if they are overall very strong with no major weaknesses--exactly like a student would deserve. S-tier characters are beyond even that, having some ridiculous strength that sets them apart. On the lower side, B characters have notable weaknesses and C characters are really just unfortunate.

The major benefit of this system is that it provides a context common across all fighting games that doesn't really need to be explained: everyone knows what kind of performance warrants what kind of grade, so the cast can be shifted up or down relative to another game and this still accurately expresses that the cast in one game is overall stronger than in another game. It also allows the expression of minor, but notable, variations by giving characters +/- grades consistent with their strength.

Notably this approach doesn't scale particularly well when rating arbitrarily large numbers of things. With only five real grades to consider, you end up with something very similar to a five-star system. Which of course, needs no introduction; it's the kind of scoring many online retailers and movie critics use.

Depending on what the system is being used to rate, common practices can differ on what deserves five stars. Product reviews often use five stars to mean "worked as expected". Movie critics reserve their five stars for the best of the best; genre- and period-defining films. Importantly, there is no real universal context in which a five-star system is based, so people don't have an inherent bias toward what the system should represent. Unfortunately, it ends up often being reduced to a binary like/dislike system by people who want to change the moving average score by as much as possible in the direction they feel it should move.

YouTube uses a like/dislike system, and actually moved to it from a five-star system a few years ago. It's very useful in expressing aggregate opinions across many people; it really doesn't matter if 1000 people rate a video between 1 and 5, you can get the same information by having them choose whether they felt overall positive or negative about it. In general this system isn't very good at comparing elements, but websites like reddit have molded it into a pretty good shape to do just that.

Let's look again at what the previously-discussed systems are used for. Tier lists are used for representing an easily-tractable dataset across a handful of categories given a common background context. Five-star systems intentionally remove the context, and are used typically to indicate the quality of a single element compared to a perceived average, rather than against any other set of elements. Like/dislike systems are a way of crowdsourcing that same indication.

So what do we use to represent a single person comparing many things? The out of 10 system works well here. Used on MyAnimeList.net for example, this system rates elements on a scale from the integers between 1 and 10, usually not allowing fractional scores. This system is extremely good at creating sorts of elements. The five-star system usually assumes a perceived average, but the out of 10 system is better at producing that average by taking an individual user's ratings. This allows you to have a good idea of the person's likes or dislikes, and also provides a lot of freedom in how you want your scale to work.

Many people, influenced again by the American school system, often set a 7.5 as "average", consistent with a C grade, and rate higher or lower based on this. So you get a sort of bell curve centered around 7.5, rendering the lower scores relatively useless. You could also take the more general approach and set the average to be 5, for whatever definition of average you'd like to use, and then distribute around it. But many people are averse to this, since 50% is a failing grade.

A friend of mine and I, curiously, take very similar approaches on opposite sides of that center. I make the assumption that the shows I watch and rate on MyAnimeList are, in general, better than the aggregate of all shows. If I watched and rated every single show, my average would be a 5. But since I, in general, watch things I consider better than average, my average score sits at 6.4. Conveniently this also lets me express the categories I like, which are the following:

10 - Otherwise a 9, but significantly influenced my perceptions or opinions in some way
9 - Extremely enjoyable, no significant weaknesses
8 - Great
7 - Pretty good, but not very notably so
6 - Decent to fairly good, OR high-quality but not appealing to me, OR mediocre-quality but highly entertaining
5 - Mediocre, OR good-quality but not appealing to me, OR bad but highly entertaining
4 - Weak, but with a potential redeeming quality
3 - Lacks any redeeming quality whatsoever
2 - Absolutely godawful
1 - Offensive to my sensibilities

Note that I dedicate more ratings toward things that I enjoy less than my average. I feel that the things I enjoy fit nicely into those four ratings, but I need extra divisions to accurately represent how awful some of the things I could watch are.

My friend takes the exact opposite approach, setting a 4 to "decent", and anything worse than mediocre gets a 1 or 2. He reserves his 10 slots for extremely small numbers of shows he considers to be the greatest things ever created, and likes to emphasize subtle differences in the things he does like.

That's the beauty of this system: given a single person's set of rated elements, it is easy to determine the strategy and distribution they use, and adapt your interpretation of the list based on that. A clear weakness is that separate rating sets cannot be directly compared, however.

Mild Injury of the Author

Death of the Author is a 1968 essay by Roland Barthes that has had profound effects on the discipline of literary criticism. Its actual, formal precept is that tying a literary work to an author imposes limits on the work and in how it can be interpreted, when the work should have no such limits.

In common usage today, "death of the author" refers more to the idea of declaring an appeal to authorial information to be invalid. If a reading or interpretation of a work relies on information about the author, then it is an invalid interpretation.

I have long since taken issue with this common usage. To deny that a literary work has an author is, first and foremost, to deny the human achievement of the work. All stories we read are written by humans, and can only be consumed through that viewpoint--we certainly cannot accurately imagine ourselves as a dog, or a highly advanced alien spaces, so humans are absolutely essential to the work.

Denying the author also leads to several cases of apparent contradictions which I will describe below.

But that isn't to say the whole concept doesn't have merit. It is extremely important to understand that, being a human, no author has full control over his or her own unconscious self. What the author may say about a work he has written could very well be false. And, most importantly, abstract meaning of a text relies on the interaction between reader and text rather than author and text, and so no reader need believe that the author's interpretations are any more valid than his own.

This image is occasionally circulated as a way for students to vent their frustration about literary analysts searching too deeply in text. These students make the error that the author is in full control of every aspect of himself, when the state of depression of the character may have unintentionally motivated the choice of a saddening color because it matched the theme. Strict usages of death of the author would deny analysis of why blue may be a saddening color--because that exists due to the cultural memes the author has internalized, which cannot be used in a death of the author analysis.

To use an example anyone reading this blog post would likely be familiar with, take Neon Genesis Evangelion, directed by Hideaki Anno. Anno has repeatedly stated that the Christian imagery seen throughout the show, of his own design, had no particular meaning. It is important that we do not take this at face value; Anno is known to have been extremely depressed during the production of Evangelion, and it is very valid to find meaning in the Christian imagery as Anno lashing out at God in frustration.

A less morally-acceptable example would be in Orson Scott Card's Ender's Game, controversial due to some interpretations of the book as Hitler apologia, building up Ender as a faultless genocider. Understandably, Card has denied this. But again, authors do not have full control over their unconscious selves, and they of course may also lie. Analysis of literature is incomplete without incorporating the author in some way, but despite the existence of "author" in the word, the author is not an interpretive authority on it.

But Death of the Author would deny this information entirely, and contrary to Barthes' goal, imposes a limit on the interpretation that can be had of a work.

Arthur writes a novel. In his novel, the narrator, also the protagonist, grows up in an orphanage after his parents die when he is young; he is occasionally visited by an uncle, his only living relative. The uncle is mentioned only in passing. When the protagonist grows up, he has difficulty forming intimate romantic connections; critics suggest that the protagonist was molested by his uncle and that memories of the events have been repressed.

Arthur, seeing the reviews, is shocked. That was not at all what he had intended; the uncle served as an unintentional role model for the protagonist, who learned emotional detachment due to this uncle not having any love for him. So he writes a second edition of the novel, this time removing the uncle entirely, and emphasizing that the protagonist's emotional issues stem from a lack of family.

This situation poses problems for Death of the Author. It is simple enough to deny authorial intent and still get meaning out of a work, but that intent has now been turned into an official re-write of the text. Therefore, to deny the author's influence, one must dictate that the two editions are completely separate stories and have no relation to each other--something that is clearly false. A story is just an author speaking in an official capacity as the author; denying the author entirely leads to problems like this one where the same story must be considered as two unrelated ones.

Another problem can arise, as well. Alice is a prolific author, praised by social equality types as writing stories that feature strong instances of oppressed people; she is particularly known for her protagonists being a symbol of the feminine and overcoming masculine oppression. To expand her horizons, she writes a new story, from the perspective of the masculine.

But she fails. For whatever reason, she cannot effectively write from a masculine perspective, and her protagonist once again ends up as a symbol of the feminine, agreed upon by critics all over. Death of the Author limits the analysis of the work that can be done under the assumption that the protagonist is a symbol of the masculine. Without a hint of what the author intended, the very idea cannot be found, as the symbol of the feminine is so strong. Particularly, it would declare the masculine reading invalid as it is unsupported in the text.

But clearly, there is useful analysis and interpretation to be done from the perspective of a failed masculine. Interesting parallels can be drawn between an attempt at the masculine and what it ended up as. But by completely disregarding the author, this analysis is impossible. To add to that, if a critic hears of the author's intent and then performs a text-only analysis based on it, he or she has still failed Death of the Author, as the analysis was motivated by the author's existence.

To conclude, understanding that the author is not an interpretive authority on their work is critical to a complete analysis of literature. Even factual statements rather than interpretations about events in a story are not necessarily completely at the whim of the author; they may misremember a detail and say something happened that is directly contradicted by the text. When this happens within the text itself, we usually call it a "plot hole," and other details of the text are used to fill in information to correct the contradiction. If this is infeasible, then it is a damning indictment of the work.

But to completely destroy the author entirely limits both the human achievement of the text and the human reality that shaped it. What aspects of the author a reader may find useful are up to him or her--and indeed, the reader may decide to use no aspects at all. But axiomatically declaring that the author cannot be used is highly stifling.