Veril
Frame Savant
Our base goal is simple: find, compile, analyze, and make available every single numerical piece of information about this game. Given the immense scope and complexity of this aim, we will need to subdivide research into individual projects and sub-projects. We will work primarily on brawl, but also (on a voluntary basis) on a stable brawl mod (B+ is stable until the end of September, BBrawl is stable iirc, B- is... lol, P:M is unreleased), melee, or SSB. For now I'm going to focus on vBrawl until that is fairly well organized and on track.
I. Character Data Repository:
The frame data currently available, at best is accurate but nowhere near complete and at worst is riddled with errors. PSA and frame advance provide us with the tools to fairly easily verify all the non-derived qualities of a move. Now complete data is not necessary for the vast majority of players, but it will be for us. For every frame of a move we need to know what is happening, what are the exact properties of the hitboxes, is there special iasa, how can it be canceled, what is the momentum effect by frame, as well as the obvious stuff like startup and First actionable frame.
There are definitely extraneous bits of information that we use. A good example would be "duration" which is just the first actionable frame +1. The old smashlab did a pretty poor job of implementing standardized terminology and actually improving research methods. Seeing as I've had to create a whole bunch of new terms, and how at times the wrong choice of term was made (boost smash never replaced DACUS, people still call it a Dacus and its annoying to have to remember both). The standardized format for frame data that people have claimed was created by the early smashlab is trash. Fortunately I've been working with data, with help from some knowledgeable people (mostly early on) but almost entirely on my own for the last year. So I'll type up a prototype for the ideal data formats (lab vs. public which doesn't have to include anywhere near as much info and should be streamlined).
A lot of data can be derived mathematically. Some things we simply are not at the stage where this is possible though. You can find advantage on block pretty easily, but the only on hit advantage that could be calculated simply would be a non-tumble inducing hit with only bkb. That's just hitstun mod * bkb value. Easy if you ignore ∂W bkb, kbg, angle, damage, character, techable frames based on DI or SDI, etc etc etc
Finding even the complete raw advantage formula for a non-tumble inducing hit is fairly time consuming, and it will only be possible to map out the character variance for these once several moves of the sort have been completely evaluated. For it to be accurate and relevant you have to manually confirm the hitstun growth points across the cast from 0-300% (ideally, I've never done that for B+ cause its not realistic for most moves, but it is useful from a research perspective). This requires a ton of tests in frame advance but there are a lot of tricks to do it faster. If any of you have seen my throw data threads, those generally only tested Adv at 0, unless it was a character I really was interested in (6.0 lucario up-throw omg what a nightmare). If I didn't find easier ways to do this than PSA and counting frames exclusively I would have had to do somewhere around 2000 tests in frame advance. Thankfully I had the formula for how throws function framewise (given to me by Magus when I was starting out, <3 Magus fo evah). So I did a ton of it with excell and only had to do a few hundred tests, which were a lot faster than the ones the rest of smashboards has been doing cause of the timer-frame conversion chart I made. Kaylo and Hotgarbage know about that as do several others I've spoken to. (gotta remember to post that here)
Whoa tangent. Anyway we need to find the adv of every move on every character between 0-300% at all degrees of staleness. Really we have to do this for a large group of moves and then produce a formula. Not some janky crap based on like launch speed. Knockback constants, damage, angle (esp something like the sakurai angle), and character physics are the variables to consider when determining hitstun. We also need the trip % at all % and this needs to actually take into account the sakurai angle. One of the mistakes I've seen is people claiming there is a 0% trip rate on Snake's f-tilt. Wrong. It has the sakurai angle, which can induce trips at sub-tumble %.
For every move, on every frame, we have to know exactly what is going on. The first phase should be collecting the player research that appears good and fact-checking it and mining PSA (easy and Blind is giving me them all converted to text so I'm gonna post them in here for easy access). Then we should work on moves that have static knockback and find the advantage on those. From there move on to throws (which work differently from other attacks and I'm DEFINITELY the expert on at this point so I'll be running that subproject) and finding the exact % all moves will induce tumble on all characters. That's important obv cause you can't DI prior to tumble.
tbc...
from the staffer shack:
Projects:
1. Collect, analyze and distribute frame data (for vBrawl, Melee, B+, BBrawl and B-, PM once its released) in a form that is complete and comprehensive as well as in an alternate player-friendly simplified format. We want an unparalleled level of detail, notably frame by frame hitbox images with corresponding hitbox information for every move, more detailed block adv and on-hit advantages.
2. Codify new formula and improve upon existing ones. This requires an overhaul of methodology but I've got a ton of support so it'll be ok.
3. Spread awareness of how to do smash research most effectively to improve the data mining ability of the community.
4. Redo the "official terminology" so that it actually reflects reality (ex. Dacus is used more often than boost smash, yet the smashlab decided to remove DACUS as the official term >.<)
5. Explore alternate means of presenting data. More on this when I've got pretty pictures ;p
6. Expand upon and fully explore what DMG started with his Planking info project. This could be huge and really significant if done right, and it absolutely requires a team.
7. Stage evaluation and testing to help inform TOs and the sbr/mlg: some of the stage hackers I've worked with (notably Frozen Popo) are just so far ahead of the curve compared to most of the community its INSANE.
8. Guides. Better guides.
I. Character Data Repository:
The frame data currently available, at best is accurate but nowhere near complete and at worst is riddled with errors. PSA and frame advance provide us with the tools to fairly easily verify all the non-derived qualities of a move. Now complete data is not necessary for the vast majority of players, but it will be for us. For every frame of a move we need to know what is happening, what are the exact properties of the hitboxes, is there special iasa, how can it be canceled, what is the momentum effect by frame, as well as the obvious stuff like startup and First actionable frame.
There are definitely extraneous bits of information that we use. A good example would be "duration" which is just the first actionable frame +1. The old smashlab did a pretty poor job of implementing standardized terminology and actually improving research methods. Seeing as I've had to create a whole bunch of new terms, and how at times the wrong choice of term was made (boost smash never replaced DACUS, people still call it a Dacus and its annoying to have to remember both). The standardized format for frame data that people have claimed was created by the early smashlab is trash. Fortunately I've been working with data, with help from some knowledgeable people (mostly early on) but almost entirely on my own for the last year. So I'll type up a prototype for the ideal data formats (lab vs. public which doesn't have to include anywhere near as much info and should be streamlined).
A lot of data can be derived mathematically. Some things we simply are not at the stage where this is possible though. You can find advantage on block pretty easily, but the only on hit advantage that could be calculated simply would be a non-tumble inducing hit with only bkb. That's just hitstun mod * bkb value. Easy if you ignore ∂W bkb, kbg, angle, damage, character, techable frames based on DI or SDI, etc etc etc
Finding even the complete raw advantage formula for a non-tumble inducing hit is fairly time consuming, and it will only be possible to map out the character variance for these once several moves of the sort have been completely evaluated. For it to be accurate and relevant you have to manually confirm the hitstun growth points across the cast from 0-300% (ideally, I've never done that for B+ cause its not realistic for most moves, but it is useful from a research perspective). This requires a ton of tests in frame advance but there are a lot of tricks to do it faster. If any of you have seen my throw data threads, those generally only tested Adv at 0, unless it was a character I really was interested in (6.0 lucario up-throw omg what a nightmare). If I didn't find easier ways to do this than PSA and counting frames exclusively I would have had to do somewhere around 2000 tests in frame advance. Thankfully I had the formula for how throws function framewise (given to me by Magus when I was starting out, <3 Magus fo evah). So I did a ton of it with excell and only had to do a few hundred tests, which were a lot faster than the ones the rest of smashboards has been doing cause of the timer-frame conversion chart I made. Kaylo and Hotgarbage know about that as do several others I've spoken to. (gotta remember to post that here)
Whoa tangent. Anyway we need to find the adv of every move on every character between 0-300% at all degrees of staleness. Really we have to do this for a large group of moves and then produce a formula. Not some janky crap based on like launch speed. Knockback constants, damage, angle (esp something like the sakurai angle), and character physics are the variables to consider when determining hitstun. We also need the trip % at all % and this needs to actually take into account the sakurai angle. One of the mistakes I've seen is people claiming there is a 0% trip rate on Snake's f-tilt. Wrong. It has the sakurai angle, which can induce trips at sub-tumble %.
For every move, on every frame, we have to know exactly what is going on. The first phase should be collecting the player research that appears good and fact-checking it and mining PSA (easy and Blind is giving me them all converted to text so I'm gonna post them in here for easy access). Then we should work on moves that have static knockback and find the advantage on those. From there move on to throws (which work differently from other attacks and I'm DEFINITELY the expert on at this point so I'll be running that subproject) and finding the exact % all moves will induce tumble on all characters. That's important obv cause you can't DI prior to tumble.
tbc...
from the staffer shack:
Projects:
1. Collect, analyze and distribute frame data (for vBrawl, Melee, B+, BBrawl and B-, PM once its released) in a form that is complete and comprehensive as well as in an alternate player-friendly simplified format. We want an unparalleled level of detail, notably frame by frame hitbox images with corresponding hitbox information for every move, more detailed block adv and on-hit advantages.
2. Codify new formula and improve upon existing ones. This requires an overhaul of methodology but I've got a ton of support so it'll be ok.
3. Spread awareness of how to do smash research most effectively to improve the data mining ability of the community.
4. Redo the "official terminology" so that it actually reflects reality (ex. Dacus is used more often than boost smash, yet the smashlab decided to remove DACUS as the official term >.<)
5. Explore alternate means of presenting data. More on this when I've got pretty pictures ;p
6. Expand upon and fully explore what DMG started with his Planking info project. This could be huge and really significant if done right, and it absolutely requires a team.
7. Stage evaluation and testing to help inform TOs and the sbr/mlg: some of the stage hackers I've worked with (notably Frozen Popo) are just so far ahead of the curve compared to most of the community its INSANE.
8. Guides. Better guides.