On Brawl Tiers

[Roller]

lets not make this a race thing...

But sonic loves to race...

 

[klapaucius]

The Intristic Value of Quantitative Assortment in Virtual Combat Situations, Part One

by

Klapaucius, Ph.D, M.B.M, MSCN, AWoL

The fourth iteration of the acclaimed Super Smash Brothers series, subtitled Brawl, is a computerized conflict simulator for the Nintendo Wii: a technological synergy of incredible versatility, capable of almost anything: its major host, an aggressive and enthusiastic Frenchman, even proclaims that you can "eat with" your Wii.1 to a regiment of intimate acolytes, some of which praise the game for its excellence in having a lot of things, and being slowed down relative to the previous installments (much like other cerebral diversions, such as chess2), and some of which engage in denigrating criticism of its unique properties; for example, the positive buoyance of the game's gravitational affinity, and its lack of advanced techniques (the most popular is to "wave dashes", a process which involves the evasion of nonexistent projectiles while airborne, causing an avatar to slide across the ground is a way which, to the objective spectator, seems like an extremely fast version of the character's natural movement. Many of the heavily organized conflicts which coalesce into Brawl's infrastructure have been solved when one participant started waving one or more dashes.)
Another common expression of dissatisfation is the incompleteness, or "brokenness" of many of the characters' anatomy. This belief pertains mostly to two of Brawl's characters. The first is known as the Solid Snake, a snake which has been granted human form by God3 and as a result became an agent of espionage for the American government, has extremely flexible limbs. This grants him access to the "up-tilt", a combat technique which consists of the Solid Snake causing his legs to become perpendicular to his torso, thus causing massive damage to its target. A normal human attempting this would incur severe damage of the groin muscles, and probable infertility. Also, the Solid Snake can use his Satanic abilities3 to summon infernal explosions at will, even while sliding.
The second is the Meta Knight. The Meta Knight's name loosely translates into Latin as "knight of changes", and this denotation is remarkably apropos. All of the Meta Knight's "special", or "B-rated", techniques involve transforming: into a combusting, invincible cyclone, into a torpedo-like living projectile, into an invisible backstabber, and in the case of his upward B-rated technique, a winged glider capable of mutilating its opponent. (This glider is one of a class which Smash fanatics respect over all else: a class capable of extreme and inevitable harm, or "****". Consequently, the Meta Knight is considered a classical "rapist".)
The Meta Knight is "broken" in that its already strange physiology--that is, a cerulean spherical body with atrophied limbs (Together with another powerful contestant, "Kirby", they are known as the "balls of Brawl.") it is believed to be missing an internal organ which all Smash contestants are required to have: the "hit-box", a box which allows the body to be physically harmed and, speculatively, may play a role in reproduction. Larger characters, like Zero-Suited Samus, the Princess of Peaches, and the Captain of Falcons, have the largest (and most vulnerable) boxes, while Meta Knight is thought to be entirely boxless. This is actually a misconception: Meta Knight does, in fact, have a box, but is is so disjointed and atrophied from years of disuse that is no longer functions.
If the evidence is to be believed, certain participants of the Brawl are more capable of achiving victory than others. This gives rise to a practice prevalent among "Smashers": the systematic catagorization of Brawling contestants in order of potency. These are colloquially known as "tiers". This report is designed to be a conclusive study of whether or not these systems are actual and effective, and where each "Brawler's" niche is in this methodology.

End of Part One

In Part Two (still awaiting completion): The exact science behind "tiers"; the metacognitive facts regarding furry space creatures, and the recontextualiation of the Coin Launcher.


Citations:

1: http://wii.ign.com/articles/703/703593p1.html

2: http://michaelchessyip.blogspot.com/2006/06/help-yourself-chess-news-and-resources.html

3: Book of Genesis, chapter 3, verse 14

 

[da K.I.D.]

The current tier list doesn't take into account the fact that some characters are played by more people. The more folks play a given character, the more likelihood of that character being played by a SKILLED person. The more skilled people play a char, the more likely he or she or it is to win a tourney. A divisive factor (i.e. the number of current mainers for that character, or perhaps the percentage of smashers maining that character) would bring the results back to something realistic.

This might also put a bit of a damper on character over-play; with the current system, the more people play a character, the more people are GOING to play that character.

Example: Meta knight is intrinsically popular by the release of Brawl. MK attracts many good players in addition to the rest. MK wins more tourneys due to having more skilled mainers. MK tops the tier list. More people play MK so they, too, can win. More good players switch to MK along with all the others. MK wins MORE tournaments. MK STILL tops the tier list. Et cetera, et cetera, et cetera.

why is it that everyone seems to think that this is an infalible arguement against tiers...
if i may id like to make a counter-example:
MK, Snake, adn sonic all attract players of all skill levels to play them. MK and Snake win more tourneys than sonic.... Why? not because more people are playing them, because at the beginning, i have to believe that just as many, if not more people played sonic than Snake and MK, but because of their actual character advantages. (i.e. them being better characters) people inherently recognising that they are better characters, either drop sonic, pick up MK or both. and thats how MK and snake win tourneys and top the tier list. tier naysayers propose that MK and snake are only the best because thats what people have heard, and than they choose to play him. i however propose that they have to be the best for that very reason. i also personaly believe that the number of people playing a character should affect their tier ranking.

snake and MK are not the best because everybody plays them.
Every one plays them because they are the best

other characters still have a chance. one extremely skilled player destroying a tourney or in some cases friendlies... or EVEN youtube videos... can impress middle-class players into getting deeply into that character, then after much practice attracting more players to that character by placing high in a tourney.

it's not a very good chance, but it's still a chance.

keep in mind that at this point and with the smash community's access to youtube, that scenario has happened with every single character in the game. why do you think everyone realises sonic is good now?

 

[Terios the Hedgehog]

By these results, Sonic would be king, and Jiggly would be dead last. That's kind of interesting.

OMG! Pilgrim's theory is CORRECT! Sonic IS the king.

The fourth iteration of the acclaimed Super Smash Brothers series, subtitled Brawl

Ummm.... wha~~~~?

 

[da K.I.D.]

i also have the say that the formula currently being used by the smash community was not a fluke because it has been used, and correctly as well, for 3 games and 10+ years. saying that our methods are incorrect is a baseless assumption. saying that theres a more efficient way to do it is completely different

 

[Pilgrim]

i also have the say that the formula currently being used by the smash community was not a fluke because it has been used, and correctly as well, for 3 games and 10+ years. saying that our methods are incorrect is a baseless assumption. saying that theres a more efficient way to do it is completely different

Which is what I did. Also, define "correctly used" ? Baseless assumption? It's statistical law. Underspecified models are less valid (i.e. less accurate) than overspecified or appropriately calculated models.

 

[popsofctown]

@"more players means better character"

It's true that better characters do get more players. however, that's not the only factor, so you can't depend on who gets the most players.

Captain Falcon is a great example. We all know that Captain Falcon is one of the worst characters in this game ( a stats method could precisely quantify how much he sucks ). But when we pick a character, there are factors besides whether that character is good. Many people main Captain Falcon because he's the manliest character in the game, and he's cool, and when he does his up B right he yells "Yes!". None of these affect gameplay, but make him more popular.
Popularity contest would tell you Captain Falcon is better than Jigglypuff, but if two guys practiced each character for two months, or even two years, i'm pretty sure Jiggs would win.

@ "this would include lower levels of play. Tier lists are for highest levels of play"
The data could be gathered from only the highest levels of play then, problem solved. If you can't identify the spots where highest levels of play are happening, then you can't make a list the old way either, so i'm assuming we can spot those places out.

@"the old way worked good enough for ten years"
Never sacrifice good for even better.

 

[Leprechaun_Drunk]

2. In ALL areas of life, statistics are proven to be more accurate than subjective opinions. Not spot on, but always more accurate. Watch TV. There's an election coming up. Do you see the people on TV saying: "here's the guys from the CNN Back Room. They really really know there stuff and know who's going to win."
Very rarely something like that happens, but 9 times as often, they tell us what the most recent poll says. Polls polls polls. Because polls are statistics and statistics are more accurate.

tell that to the electoral vote system.

on a serious note, I fear that this idea will fall into the pit of other awesome ideas that would generally benefit the smashing community (competitive and otherwise) that will not be implemented into the game, no matter how logical, intelligent, or worthy they are. However, I would support such a change in the means to which we devise a tier list.

 

[popsofctown]

Dude, don't get me started on the electoral college

I don't know though, maybe this will get adopted, considering how many times "facts" has been repeated in the Brawl-Melee debate and the Ban MK thread (it went like 14 pages long in a thirty minute gap between moderator checks, pretty much, haha. You guys should go read it.)

 

[Leprechaun_Drunk]

if, by some miracle, these new methods are implemented, it would take a massive amount of time and effort before conclusions would be made. If this brilliance is not to die out, a team of highly intelligent smashers who understand these concepts should be conceived and given the task of working it all out. Only problem is, I don't think anyone is as qualified as the OP...and i doubt one person (especially someone fresh out of Grad school with a wife) will be able to do everything on his/her own.

 

[AlphaZealot]

Statistically speaking a pawn is the weakest piece on the board, yet, a black pawn on b2 is incredibly powerful, or several pawns in a strong formation (versus pawns stuck as islands). Technically, a bishop and a knight are of equal value (same tier) yet, in an open board the Bishop is stronger, in a tighter more confined board the Knight is better.

Tiers, as they are currently devised, measure more popularity and skilled player character selection, as opposed to inherent character atributes.

Trying to rate characters based on attributes (speed, strength, what have you), simply doesn't work. How would you rate Diddy's bananas? Better or worse than Pits arrow? Most would say bananas are better than arrows, but in the specific Diddy-Pit match up arrows happen to be more powerful, and in Diddy-Falco versus Pit-Falco, Pits' arrows likely are better, yet in Diddy-Marth versus Pit-Marth, Diddy's bananas are likely better. There are simply to many variables and to many oddities that influence any given match up that you can't look at one character and say "hes faster so he should win". This doesn't even touch how match ups get swayed by levels and the rules of a tournament.

At the end of the day, if you say one character is statistically (inherent character attributes) better than another, that character still needs to be used in a tournament and place in such a way that proves it. This is taking your science one more step. Its showing that something is true both on paper and in practice. Science, often times, works in theory but not in practice, and tier lists happen to be based on practice, not theory (well, minus tier lists created prior to 2005 or so).

 

[Terios the Hedgehog]

Plus going by STATS Falcon is pretty good. XD

 

[Pilgrim]

Statistically speaking a pawn is the weakest piece on the board, yet, a black pawn on b2 is incredibly powerful, or several pawns in a strong formation (versus pawns stuck as islands). Technically, a bishop and a knight are of equal value (same tier) yet, in an open board the Bishop is stronger, in a tighter more confined board the Knight is better.



Trying to rate characters based on attributes (speed, strength, what have you), simply doesn't work.

Read the posts again you misunderstood the variables I am interested in.

 

[Devil7]

wait since when did brawl have a tier list. Right know its all speculation and aparently most of don't know what it takes to make a tier list to begin with.

 

[popsofctown]

At the end of the day, if you say one character is statistically (inherent character attributes) better than another, that character still needs to be used in a tournament and place in such a way that proves it. This is taking your science one more step. Its showing that something is true both on paper and in practice. Science, often times, works in theory but not in practice, and tier lists happen to be based on practice, not theory (well, minus tier lists created prior to 2005 or so).

Yeah, i think you misunderstood. When Pilgrim said attributes, he meant all the collective differences that aren't cosmetic. His proposal is, in fact, going to be based off of practice, not theory at all.

@Leprechaun Drunk
Finding someone qualified enough to crunch the numbers is an issue. That person having enough time to crunch the numbers is not that big an issue. It's powerful software. You enter the data. You click the test you want. The computer processes while you go do something else for 5 hours, then you come back and it gives you results.
The only issue we have is not being qualified enough to know what in the sam hill the results mean, or if we picked the right test.
The main thing Pilgrim doesn't have time to do is probably data entry. Laymen can enter data. We can ask him what data we need to collect, we can ask him what rows and columns to enter the data in, and we can ask him which test to run. Answering that won't take up too much of his time.

EDIT: if Pilgrim says he'll do that, i'll volunteer to crunch the data.

 

[betterthanbonds9]

i lol'd at arguing with a BRer about making a tier list, i really did.

 

[popsofctown]

He's a backroomer, but I'm very very certain he misunderstood the OP. Based off the way he responded, he seemed to be disagreeing with an idea that i disagree with as well, but wasn't put forth in the OP. Until we are talking about the same thing, we aren't arguing.

 

[AlphaZealot]

Alright lets take a step back.

In as simple terms as possible, what is the step by step process you propose for creating a tier list.

 

[Pilgrim]

Okay very simple:

And the computations will take less than 20 minutes to compute (not 5 hours, unless you are back in the 1990’s), and there is open source (i.e. free) software packages available that can compute all the necessary statistics. The only hard part will be people recording the data in a uniform manner—which can be easily solved if the community makes a commitment to this— and delegating a person to enter the data that is reported from tournaments. You could use excel to collect the data. Making columns with the data described below and your rows would be individual matches.


For each tournament that happens you would need the following— which will likely be contained in a simple tournament roster comparable to the one provided in wii.

End result of Tournament Ranking over all characters played (i.e. does not matter if it is triple elimination and they used three characters, just what did they place in the end [e.g. 1st)—this will be entered multiple times for the same person per tournament)

Character used in match up by the individual and the opponent faced (this will be double entered with each character having a placing and as the primary character)

Win or loss for each character used

Number of contestants at the tournament for this match

An ID number for the tournament (could be arbitrary—like naming the 1st tournament 1 and the 226 tournament 226)

An ID number for the participant at the tournament (this could be arbitrary too—like 1 through 42) This is in case an individual uses different characters or the same character and controls for player attributes that may be unique to the individual.

Simply using the data above would greatly increase the accuracy for tiers and even provide a relatively simple tier list for character match ups. If you wanted to spice it up you could enter years of experience or total hours played on wii console. or other variables that may predict performance such as controller preference— however this is not necessary, it would just increase the accuracy.


A very simple model using logistic regression and multilevel modeling could provide a relatively easy equation for this data. It would greatly increase your statistical accuracy above and beyond that of mean differences.

I could send an excel sheet with examples of how to enter this data if that is not simple enough.

 

[DanGR]

Haha, this reminds me of something: http://www.thiemeworks.com/garden/archive/19980514.html

Anyways, I'd volunteer to put some of the numbers together if it meant there would be a more accurate tier list than there would be without my help.

 

[DanGR]

Simply using the data above would greatly increase the accuracy for tiers and even provide a relatively simple tier list for character match ups. If you wanted to spice it up you could enter years of experience or total hours played on wii console. or other variables that may predict performance such as controller preference— however this is not necessary, it would just increase the accuracy.

The only problem with this is that some people learn faster than others. There's too many variables.

Haha, this reminds me of something off topic, but oh well: http://www.thiemeworks.com/garden/archive/19980514.html

Anyways, I'd volunteer to put some of the numbers together if it meant there would be a more accurate tier list than there would be without my help.

 

[Opfer]

I think there's still a misunderstanding going on here. Correct me if I'm wrong but this is how I see it:

What Pilgrim is talking about is processing data similar to what's going on in this thread, just in a more sophisticated way. The way it's done there right now doesn't, for example, take into account how often a character was actually used vs. how often a character won, but only the absolute number of wins a character had, thereby being greatly influenced by chracter popularity.

As far as I can see, this has nothing to do with the Tier List the backroom releases (which is, as far as I know, not based on tournament statistics, but instead on discussion about the character matchups). I don't think he ever suggested using statistical methods in order to replace this discussion (i.e. using statistics to analyze the indiviidual moves of each character or anything like that).

So I'm all for using more complex methods for processing tournament result data. It's probably not a lot more work, and a better view on the statistics can't hurt either.

Also, "The Intristic Value of Quantitative Assortment in Virtual Combat Situations, Part One" by Klapaucius is full of epic win, I wonder why nobody commented on that yet, simply awesome, man .

 

[Pilgrim]

The only problem with this is that some people learn faster than others. There's too many variables.

Those would primarily be used as covariates to decrease error variance-- which could only help the results-- not detract. As stated before you will never have perfect accuracy, only better-- and better is better than, well, worse.

 

[popsofctown]

Those would primarily be used as covariates to decrease error variance-- which could only help the results-- not detract. As stated before you will never have perfect accuracy, only better-- and better is better than, well, worse.

So, by that you mean, the sophisticated methods are powerful enough to account for everything just a bit better?

 

[betterthanbonds9]

it just hit me, doesn't this account for too many easy first round wins? (ie, pro vs scrub)

OR does it also detract from the big tournaments that would be filled with pro vs pro and if they all used "god tier" characters then that could only detract from the characters rankings because they'd assumingly be highly used?

just trying to figure out more of the situations this is applicable

 

[popsofctown]

whoa, that post was a bit confusing. The more advanced stats methods won't create bias, they only remove bias. Their might be some things even the good methods can't account for, however, i think you're trying to ask if the advanced methods would help with those, but your post is hard to read.

 

[betterthanbonds9]

whoa, that post was a bit confusing. The more advanced stats methods won't create bias, they only remove bias. Their might be some things even the good methods can't account for, however, i think you're trying to ask if the advanced methods would help with those, but your post is hard to read.

yeah, i couldn't figure out where to put the question mark........

the thing im asking is if this accounts for popularity by assumingly decreasing the role it poses in this tier list then wouldn't the big tournaments that in melee were filled with marths shieks falcos, and foxes end up decreasing their high placing in tiers?

and the first part shouldn't have been hard to read...if a pro beats a scrub, would it really matter to an extent what character the pro used? It could give a win to a character that was insignificant overall once the pro got later in the tourney.

 

[kr3wman]

yeah, i couldn't figure out where to put the question mark........

the thing im asking is if this accounts for popularity by assumingly decreasing the role it poses in this tier list then wouldn't the big tournaments that in melee were filled with marths shieks falcos, and foxes end up decreasing their high placing in tiers?

and the first part shouldn't have been hard to read...if a pro beats a scrub, would it really matter to an extent what character the pro used? It could give a win to a character that was insignificant overall once the pro got later in the tourney.

In a true tier list, you compare every characters' matchups and compare it overall with all the other character's overall. In a tournament-based tier list, you compare every characters' matchups against only the high tier character. Jesus, it's not like its complicated.

 

[popsofctown]

yeah, i couldn't figure out where to put the question mark........

the thing im asking is if this accounts for popularity by assumingly decreasing the role it poses in this tier list then wouldn't the big tournaments that in melee were filled with marths sheiks falcos, and foxes end up decreasing their high placing in tiers?

and the first part shouldn't have been hard to read...if a pro beats a scrub, would it really matter to an extent what character the pro used? It could give a win to a character that was insignificant overall once the pro got later in the tourney.

If we'll stick with the melee example, a tournament full of marths and sheiks wouldn't make marth's and sheik's rankings go down, just because they were uber popular at that tournament. However, when a marth or sheik emerges victorious, the win is less significant for that characters reputation than if one marth emerged victorious amongst 99 Marios. The advanced stats methods would pull from that tournament, ok, Marth is good, 10 pts. But Peach just won a tournament over here and there was only three Peaches. 20pts. etc. That's something we could try to account for now, roughly, but if we used some powerful Stats software, it will do that stuff automatically.

And for the pro v scrub matches, it can also account for that if we collect the data. If we collect player names, then the computer will go through and say "ok, Azen is awesome. there's a 96% chance he will win a match, altered by where he plays, who he plays against, and which char he picks."
It will collect "Ok popsofctown is total bullcrap. He wins like 15% of his matches. This is altered by who he plays against, what char he picks, and where plays."

So if i play Azen, there's no telling what character he's going to beat me down with, but the computer will make the character matchup turnout less significant since Azen started out with a 99% chance to win off of skill.

 

[JackieRabbit5]

this really needs to be done badly...i'm very curious to see some results

sounds like it would be much more conclusive, and it would only get more accurate after more tourneys and data

it wouldn't even have to make tier lists composed by "backroomers" obsolete...it would be good to compare with

 

[popsofctown]

currently they make the tier list partly off of statistics, party off of opinion. The statistics parts could be more accurate.

 

[betterthanbonds9]

If we'll stick with the melee example, a tournament full of marths and sheiks wouldn't make marth's and sheik's rankings go down, just because they were uber popular at that tournament. However, when a marth or sheik emerges victorious, the win is less significant for that characters reputation than if one marth emerged victorious amongst 99 Marios. The advanced stats methods would pull from that tournament, ok, Marth is good, 10 pts. But Peach just won a tournament over here and there was only three Peaches. 20pts. etc. That's something we could try to account for now, roughly, but if we used some powerful Stats software, it will do that stuff automatically.

And for the pro v scrub matches, it can also account for that if we collect the data. If we collect player names, then the computer will go through and say "ok, Azen is awesome. there's a 96% chance he will win a match, altered by where he plays, who he plays against, and which char he picks."
It will collect "Ok popsofctown is total bullcrap. He wins like 15% of his matches. This is altered by who he plays against, what char he picks, and where plays."

So if i play Azen, there's no telling what character he's going to beat me down with, but the computer will make the character matchup turnout less significant since Azen started out with a 99% chance to win off of skill.

okay, that's the main thing i wanted to know.

I've never been adverse to this idea, i just always questioned 2 things: 1. implementation and 2. times when this wouldn't work

it's pretty obvious where i've questioned each, but i do still question how many people will send that info in, but it'd be cool.

And now for the last question which has been posed many times before and never answered:

Right now, there is no tier list so there is no such thing as a character's popularity being altered by the tier list. Now, sure there are stuff like ankoku's thread and some not legit tier lists, but that's a different story. Obviously now there is no subjective list and it is up to those with the ability to win tournaments to win without knowing all that's possibly going to be known 4 months from now. So what would this tier list prove at the moment? IMO this would be a better project to be done either: between tiers to document who's doing well (could also show the effect of tiers when compared to multiple time periods) and as a general matchup chart to show who REALLY has the advantage or if it's neutral.

currently they make the tier list partly off of statistics, party off of opinion.

finally somebody acknowledges this