Taking Moveset & Data Analysis to the Next Level

Remzi · May 19, 2009

So after a little more than a year since Brawl's NA release, it's safe to say that we have covered all of the basic topics that are generally discussed in a fighting game. We've talked about tiers, we've talked about every matchup, we've talked about character weight, we've even broken down the frame data of almost every move in the game.

The Problem:

However, we STILL have even the most knowledgeable players saying "X character's dtilt outranges Y character's ftilt," even when attempting to speak in the utmost detail.

Don't see the problem there?

A typical frame data comparison of two moves will consist of: "X attack's hitbox comes out in 5 frames, while Y attack's hitbox comes out in 8 frames. Therefore, the Y attack must activate 4 frames earlier than the X attack in order for it's hitbox to unleash first."

Why can't people say X Character's dtilt is an EXACT amount longer than Y Character's ftilt? I think you see what I'm getting at here...

The Solution:

We need to establish a universal measurement for length. This will help us to measure:

-Hitbox lengths
-Stage lengths
-Platform lengths
-Knockback
-Dash lengths
-Jump Height
-Grab release distances
-More

However, applying such a number to all of those things listed above is easier said than done. To start, we would have to apply an arbitrary number to one thing that has a definite length. For example, lets say that FD measures 200 "____s" in length. You may be wondering why I left that area blank. Well, I can't think of a good word so I'd like for people who respond to this thread to drop some ideas for a word we can use as units. We don't want a word that is used for real life measures (i.e inches, meters) as that may cause confusion. We want a new, arbitrary word that sounds good. Anyways, if we use the length of FD as a base number to refer to everything else, we can establish an accurate and consistent form of measurement.

So assuming we did use 200 as the number to apply to FD's length, if we broke FD's surface into 200 equal pieces, one of those pieces would be our unit of measurement.

If we accomplish this, we can apply that to all other measurable things. For example, Marth's fair is 7 ___s long (not really, I haven't measured).

It will be a difficult and lengthy project, but imagine how useful it would be in the end.

Thanks for reading!

Thoughts? Ideas for a unit name or FD's measurement #? Volunteers to help?

Post here.

ndayday · May 19, 2009

Exactly how would you measure this? I mean, it's not something really measurable like frame data, you would have to go into pixels, and unlike frame data, it would be insanely hard to try to get an exact amount of the range of an attack. For example, character A's dtilt is...say, 56454 pixels long, but it might have a range of 90 degrees to 20 degrees. (which would be an approximation of its own) Say you found a unit, lets call them binches for now. (Brawl Inches) Say they're, oh, 735 pixels. Using the previous dtilt example, it would measure to be 76.808163265306122448979591836735 binches. That would only be the length....you would have to find an exact area (or pretty darn close to it) for the entire hitbox of the attack, which means you would quite literally have to measure over for every pixel of the attack. Here is a picture:

Assuming that the hitbox is all the blue, and the red is the character, how would you find the range? Assume every pixel is 1 binch, then tell me how you would go about comparing that to the hitbox of an other attack.

In other words, I don't think the idea is needed or even possible to figure out easily.

Nic64 · May 19, 2009

Ideas for a unit name or FD's measurement #?

measure all attacks hitboxes by how many pixels they are? <_<

I think you'd have to have a scale where you simply determine that X is longer than Y and Z is longer than X so Z is two <unit> longer than Y or something.

CRASHiC · May 19, 2009

http://www.smashboards.com/showpost.php?p=5579925&postcount=9

Try this, but have it more structured, having each hitbox shown at the same point on the stage, at the same zoom pixel per pixel. Range is then compared by overlapping images.

SuSa · May 19, 2009

"A typical frame data comparison of two moves will consist of: "X attack's hitbox comes out in 5 frames, while Y attack's hitbox comes out in 8 frames. Therefore, the Y attack must activate 4 frames earlier than the X attack in order for it's hitbox to unleash first."

What does this matter if X outranges Y.....if your Jigglypuff and use a bair, and I'm Marth and I use fair - I outrange you.

Also would distances be measured from the furthest hitbox? Because if Marth is right next to jigglypuff - he wouldn't outrange her by anything, while if he's at his furthers point he outranges here by quite a bit. However there are countless inbetweens.

EDIT:

How did Mr. Eh do those hitboxes... that's what I'm wondering..

Remzi · May 19, 2009

NAKEDeDeDe said:
Exactly how would you measure this? I mean, it's not something really measurable like frame data, you would have to go into pixels, and unlike frame data, it would be insanely hard to try to get an exact amount of the range of an attack. For example, character A's dtilt is...say, 56454 pixels long, but it might have a range of 90 degrees to 20 degrees. (which would be an approximation of its own) Say you found a unit, lets call them binches for now. (Brawl Inches) Say they're, oh, 735 pixels. Using the previous dtilt example, it would measure to be 76.808163265306122448979591836735 binches. That would only be the length....you would have to find an exact area (or pretty darn close to it) for the entire hitbox of the attack, which means you would quite literally have to measure over for every pixel of the attack. Here is a picture:

Assuming that the hitbox is all the blue, and the red is the character, how would you find the range? Assume every pixel is 1 binch, then tell me how you would go about comparing that to the hitbox of an other attack.

In other words, I don't think the idea is needed or even possible to figure out easily.

I know what you mean, I thought of this already. I probably should have detailed it in the OP but w/e.

Take a move like Marth's fsmash for example which has good vertical and horizontal range. Rather than giving a number for EVERY exterior point on the hitbox, we could give a number for the furthest horizontal extent of the hitbox all around, the bottom exterior, the highest point, and perhaps a few others in between. This would give us the key data for each move, while still allowing us to get an accurate approximation for the pieces of the hitbox in between.

As for how the length would be measured, you would compare the edge of the characters normal standing hurtbox (or standard aerial position hurtbox for aerial attacks) and measure the distance from the furthest horizontal point on said hurtbox, to the points I listed in the paragraph above this one.

Sorry if that was unclear, just say the word and I'll type a better explanation.

Susa: We would measure it from the furthest point. So lets say Marth's fair is 15 binches long at its longest point while Jigg's bair is 6 binches at it's longest point, we could establish that Jiggs would be unsafe for a length of 9 binches.

san. · May 19, 2009

There are also holes/openings in hitboxes. Some hitboxes start at one end and finish at the other. some character's attacks can get through the quick "open" spots of the other characters more/easier than others.

SpongeBathBill · May 19, 2009

I suggested using pixels before and got laughed at because of how hard it would be to measure -.-

ndayday · May 19, 2009

san3711 said:
There are also holes/openings in hitboxes. Some hitboxes start at one end and finish at the other. some character's attacks can get through the quick "open" spots of the other characters more/easier than others.

This too. It's not a matter of whether it can be done or not, (because it can) but I think it would be way to hard to come to a precise number, especially when you take into account stuff like hitboxes inside of characters and disjointed hitboxes. To measure it, you would have to go:

Htibox #:
Binches in at least 8 directions (north, south, east, west, NW, NE, SE, SW)

and draw your conclusions from that. This way, you could say that a characters range for a move (for the sake of simplicity, lets make it a single hit move) is about 7 binches larger in the horizontal direction then another single hit move.

I'm interested in the concept...but another thing is the problem of figuring out exact hitboxes. Is there a method to getting the data?

Vyse · May 20, 2009

CRASHiC said:
http://www.smashboards.com/showpost.php?p=5579925&postcount=9

Try this, but have it more structured, having each hitbox shown at the same point on the stage, at the same zoom pixel per pixel. Range is then compared by overlapping images.

You have to be wary because just because an attack may have a measurable length in term's of the character's model, it may not be an accurate depiction of the 'hitbox'.

Snake's U-tilt is a good example.

Another example that comes to Mind is Melee Gannondorf's running grab, which reaches behind him.

SamuraiPanda · May 20, 2009

It would be quite nice to have a hack that shows a grid or something over the pause screen that would allow us to measure distances by just counting. Of course, it would help to have the game pause completely zoomed out from the stage to normalize the measurement.

thrillagorilla · May 20, 2009

SamuraiPanda said:
It would be quite nice to have a hack that shows a grid or something over the pause screen that would allow us to measure distances by just counting. Of course, it would help to have the game pause completely zoomed out from the stage to normalize the measurement.

Do you mean like fixed camera mode? Wouldn't it be theoretically possible to do it in that mode?

Another thought about hit-box measuring, though. Is there a character with a consistent hurt-box to test it on? All of the character's hurt-boxes move on a frame by frame basis, even when "standing still". On a side note, I would suggest using small blocks from the custom stages to work out your measuring units. It is easy to line up your test character's hurt-box with the beginning of each block, and easy to approximate where the attack will be at its greatest length/height due to their consistent structure.

Ack, almost forgot. One more question if you would allow. Would this data include sliding from craqe walking and studder stepping?

Vyse · May 20, 2009

thrillagorilla said:
Ack, almost forgot. One more question if you would allow. Would this data include sliding from craqe walking and studder stepping?

I wouldn't imagine so. The actual size of the hitbox doesn't change because of the step/slide, just positioning.

Remzi · May 20, 2009

SamuraiPanda said:
It would be quite nice to have a hack that shows a grid or something over the pause screen that would allow us to measure distances by just counting. Of course, it would help to have the game pause completely zoomed out from the stage to normalize the measurement.

As long as you take the snap shot that you are working with ON FD, this is not necessary. You can divide FD into X amount of segments (depending on what we decide to make the #). What you are saying can be used to measure the other stages, though.

This is for the number idea, however, not for the grid idea which would be almost equally useful.

thrillagorilla said:
Ack, almost forgot. One more question if you would allow. Would this data include sliding from craqe walking and studder stepping?

No, though we could measure approximate stutter step lengths as well and tack those on.

Vyse said:
You have to be wary because just because an attack may have a measurable length in term's of the character's model, it may not be an accurate depiction of the 'hitbox'.

Snake's U-tilt is a good example.

Another example that comes to Mind is Melee Gannondorf's running grab, which reaches behind him.

I'm thinking we should make getting detailed pictures of all hitboxes a top priority then.

We could start by having each character board get ACCURATE diagrams of each hitbox, then we could turn that into straight numerical data.

Vyse said:
I wouldn't imagine so. The actual size of the hitbox doesn't change because of the step/slide, just positioning.

Yes, that's correct.

BleachigoZX · May 20, 2009

Bleachigo's, that is the ideal measurement of Brawl.

Browny · May 20, 2009

Just for people who want to know... Here's how I wrote my hitbox thread

http://www.smashboards.com/showthread.php?t=211629

djbrowny said:
Actually, the jigglypuff, ness and lucas boards have done the same thing. I have no idea how far theyve gone, but even more interesting is that none of them contacted me and asked for specific details in how I came up with these pictures...

It was made by using 4 pictures of all attacks (more for multi-hit attacks, uair had 8) showing the exact frame the attack hitbox came out, and the spark that resulted. For example, bair

That spark represents the maximum range I was able to get the bair to hit from, from behind. I then did that to all 4 directions for each attack, trying to take each screenshot at the exact same frame (extremely difficult, I couldnt manage it often). Then with each pit, I took a reference value of the range relative to sonics body. In the above example, lets say the range from the spark to Sonics glove was 10 pixels. I then measured the length of a reference item, using the platforms of battlefield. they were around 550 pixels.

Then in the picture I used in this thread, I measured the size of the platform (about 600 pixels) and applied a correction factor of (10/550) to find the relative range of the attack. This removed most of the errors involved in using different screenshots to measure the same attack, since its hard to keep all images at the exact same zoom level and orientation.

With that done to each direction of the attack, I drew an ellipse with the major and minor diameters equal to the calculated range of the attack and located it with each edge on the pixel I drew as the max range of the attack in any direction.

All these methods cause a build up in errors, I could expect each step to go maybe 2-3 pixels each way so it could get as bad as around 6-9 pixels each way, but its the best I can do by only using screenshots

Toby. · May 20, 2009

DJB you never cease to impress. Fancy stuff.

This sounds useful, although it would take an incredibly long time.

Dantarion · May 21, 2009

Or, reverse engineer stuff held in memory or the files on the brawl disk.

Tennet · May 21, 2009

Just put a ruler on your TV duh...

B!squick · May 21, 2009

Tennet said:
Just put a ruler on your TV duh...

The obvious problem with this is that not everyone has the same TV. Which means this method would only work if someone was willing to collect data for EVERY character's ENTIRE move set... are you volunteering?

Tennet · May 21, 2009

/sarcasm...

SSD · May 21, 2009

Why make up a silly name when we could simply call them units, which is exactly what they are?
I'm going to experiment with finding a way to measure things without having to resort to pixels.

ndayday · May 21, 2009

You could make the stage a grid, and apply the same distance squares for each stage. As for the name, units sounds good. Binches might bring people to conclude that the distance is one inch, which I would assume is not the case at all. And fancy stuff, DJB.

SSD · May 21, 2009

Using stages as a unit of measure, even if you divide them, is hard as the camera angle and distance change the way the stage looks. Each stage has a different "fully zoomed out" view as well.

Characters shift back and forth as they go through their idle animations. While testing the hitbox of Pit's uptilt, I manage to position him next to marth so that when Marth leaned forward, it would register a ping hit, but it would miss if he shifted back the slightest amount.

If we use something in the game to base a measurement system on, it needs to be something completely static that can easily be measured regardless of camera angle and position.

SuSa · May 21, 2009

Fixed camera mode (if you can pause, IDK how it pauses <_<) and sandbag possibly?

SSD · May 21, 2009

SuSa said:
Fixed camera mode (if you can pause, IDK how it pauses <_<) and sandbag possibly?

pausing works just like pausing in normal brawl. =/

Remzi · May 23, 2009

Good stuff guys, glad we have a good amount of interest here.

Goal #1: Detailed pictures of all hitboxes. I believe Shaya has all of them for Marth.

If we can get someone to do it for each character that would be amazing.

SuSa · May 23, 2009

I still don't understand how to get an accurate hitbox.

Because I'd be willing to do it for Snake. (Lulz.... I can already foresee utilt...)

uhmuzing · May 23, 2009

I believe that it would be helpful if we had a universal unit of length to measure hitboxes, but it would take a ridiculously long time and alot of painful effort. In the end, though, I think it would be well worth it.

Darxmarth23 · May 23, 2009

I don't know, but I hope t his helps:

Each character consists of a red bubble and yellow bubble. The red one is the characters hurt box, and the yellow one is the hitbox. Generally they overlap each other entirely. But during certain moves, the yellow hitbox passes the red hurt box causing it to be disjointed.

If we could somehow remove the texture of a character and have it stand as just red and yellow box's, maybe it could help you somehow measure the range.

And also: http://www.smashboards.com/showthread.php?t=232375

I don't know if the above link or paragraph helps....

Browny · May 23, 2009

SuSa said:
I still don't understand how to get an accurate hitbox.

Because I'd be willing to do it for Snake. (Lulz.... I can already foresee utilt...)

read post 16 of this thread lol...
it may not be 100% accurate, but its as close as anyones ever got

AMKalmar · Jun 3, 2010

I apologize for reviving this dead thread. Someone linked to it and I thought it wasn't as old as it was.

Orion* · Jun 4, 2010

this isnt worth the effort lol. its a **** load of work, for a very small reward

Dark Sonic · Jun 4, 2010

Wouldn't the visible hitbubble PAC hacks be perfect for this?
http://www.smashboards.com/showthread.php?t=254970&highlight=hitbubble

I'm confused as to why no one mentioned this yet

Orion* · Jun 4, 2010

the pacs are sufficient imo but we dont have a way to "measure" the hit bubbles. i dont think its necessary but w.e

Taking Moveset & Data Analysis to the Next Level

formerly VaBengal

stuck on a whole different plaaaanet

Smash Lord

Smash Hero

Banned via Administration

formerly VaBengal

1/Sympathy = Divide By Zero

Smash Ace

stuck on a whole different plaaaanet

Faith, Hope, Love, Luck

Smash Hero

Smash Ace

Faith, Hope, Love, Luck

formerly VaBengal

Smash Master

Smash Hater

Smash Master

Smash Champion

Smash Champion

Smash Master

Smash Champion

Smash Journeyman

stuck on a whole different plaaaanet

Smash Journeyman

Banned via Administration

Smash Journeyman

formerly VaBengal

Banned via Administration

human-alien-cig

Smash Champion

Smash Hater

Smash Ace

Smash Researcher

Smash Hero

Smash Researcher

Information

Network