Why aren't tiers more statistically based?

[D4RK_HUNT3R]

Here's an example of what I mean:

http://sports.espn.go.com/espn/page2/sportSkills?sort=total_rank#grid

Advanced techniques for sep. characters could add extra points, on top of things like weight and power. Some categories might be worth more than others.

This would make for a more quantitative analysis rather than qualitative, which IMO would make the tiers more for "everyone." The way Melee was done was with tournament results, but that could make for some faulty rankings (what if more people play character X than character Y, but in reality character Y is completely better?).

When the true pro scene arises, I think they should make it numerically based with actual research and statistics. And they should make a chart like in my example link -_- More professional and accurate IMO.

Sorry if I "offend" anyone.
*raises flame shield*



EDIT: BTW feel free to move my topic to the Tournament section if this is off topic somehow or it fits there better >_>

 

[dotpwn]

They make it this way because characters dont make a match, the players do. You fail to see that in sports everybody is the same character, "human."

 

[D4RK_HUNT3R]

Tier lists are never absolute in match determination, not with the current Melee one, not ever. Again, what if people just despise playing as Fox now because he's boring so everyone jumps on the C. Falcon train then C. Falcon gets ahead of Fox in tiers? (I know this is really false, but I'm using an example)

We need something a little more concrete than tournament players' results.

 

[Ripple]

They make it this way because characters dont make a match, the players do. You fail to see that in sports everybody is the same character, "human."

QFT

10chars

 

[Meta_Sonic64]

They make it this way because characters dont make a match, the players do. You fail to see that in sports everybody is the same character, "human."

Just as characters in Smash, some humans are better than others at doing things. Which is what this guy is trying to point out. Just because we are all humans don't mean we all have the same skill at the same extent

 

[Dragonbreath]

If you base them off simple character traits, then they aren't tiers. They're simply your personal valuations in the form of a list. Besides, with all the variables between characters, playstyles and personal preferences, it's impossible to mathematically prove one character as better than any other, let alone better than all of them.

 

[Dark Sonic]

If you base them off simple character traits, then they aren't tiers. They're simply your personal valuations in the form of a list. Besides, with all the variables between characters, playstyles and personal preferences, it's impossible to mathematically prove one character as better than any other, let alone better than all of them.

Here's your answer.
[/thread]

 

[D4RK_HUNT3R]

If you base them off simple character traits, then they aren't tiers. They're simply your personal valuations in the form of a list. Besides, with all the variables between characters, playstyles and personal preferences, it's impossible to mathematically prove one character as better than any other, let alone better than all of them.

It's not impossible, I just think they should try it out with every possible measurable attribute. If you looked at the actual example link, ESPN ranked them according to how many points they scored total. SSBB tiers could be like that too, with added Misc. points that include things like tether grabbing or Peach's floating.

They aren't just simple character traits. If you think that, then you have no idea how many things are measurable in this game. The tier list makers can adjust categorical significance by toning down the weight one category has. Think how many attributes could combine to produce a mathematically done tier list.

Hell, why don't they just make 2 tier lists: Tournament based and pure statistics based?

 

[SeriousWB]

How about, you try that out, then come back with your results.

 

[D4RK_HUNT3R]

How about, you try that out, then come back with your results.

Because it's too much for a single person to do, and I not only lack the game, but lack technical know-how to do a lot of the things.

 

[Same-Move Sammy]

How about, you try that out, then come back with your results.

I agree. I would like to see how this compares to subjective tiers.

 

[Zankoku]

I've worked with someone on Smashboards on creating a mathematical formula based on character matchups to calculate a tier list for Melee. It was half-decent, but it had one major problem:

It required completely accurate data regarding character matchups, and most of the experiments were based on the inaccurate phanna chart. As it went down the list, lower tiers had more misrepresented matchups and things started looking stranger and stranger (especially with Kirby's wtf 4-6 rating against Sheik).

Point is, building a tier list based on tournament results rather than empirical data that may or may not be accurate is faster and much, much more likely to succeed in the long run.

 

[John5]

I think there are too many variables that are not concrete enough for character tier list to be based off stats alone. Like, edge-guard capabilities, recovery, keep-away ability, etc. Tournament results are probably the most accurate way to go.

 

[D4RK_HUNT3R]

I've worked with someone on Smashboards on creating a mathematical formula based on character matchups to calculate a tier list for Melee. It was half-decent, but it had one major problem:

It required completely accurate data regarding character matchups, and most of the experiments were based on the inaccurate phanna chart. As it went down the list, lower tiers had more misrepresented matchups and things started looking stranger and stranger (especially with Kirby's wtf 4-6 rating against Sheik).

Point is, building a tier list based on tournament results rather than empirical data that may or may not be accurate is faster and much, much more likely to succeed in the long run.

Then that means you failed to weigh everything correctly. You should've expanded the characters' points to where some categories weigh less than others, so Sheik's points are more numerous than Kirby's.

Yes it's much harder than the traditional way, but don't you think it's worth it? I'm sure there are quite a few people with nothing better to do and would painstakingly do this. If you don't think so, you don't pay attention to some of the other crazy stuff gamers have compiled.

 

[Zankoku]

We weighted everything fine. The data input was what was flawed. Not only did we need a formula that took everything into account based both on matchup and overall "popularity," we also had to work on the assumption that we had completely accurate data on each character to start with. That's pretty much impossible with Melee's roster of 26 characters. Tell me, would it be even more possible with Brawl's roster of 37-ish?

 

[SeriousWB]

At one point you would have to be subjective on the matter, how can stats decide whether a higher or lower weight is better? Other than tournament results, which is the standard method anyway.

 

[House M.D.]

I like this idea a lot. Here are a few thoughts

1) The end result would have to resemble the tournament based tier list. Even top players do not agree on the tier list (ex. m2k and ken think sheik is the best), but they all probably agree that characters are generally where they should be (gimpy knows bowser sucks).

2) It might be better not to think of it in terms of individual matchups, because then it would get too subjective. Rather, let's take fox's ability to shine spike. This is very effective against certain characters (ex. falcon) but not others (pikachu), so when accounting for this ability, we would quantify it as effective against (say) half the characters, without worrying about who those characters are.

3) To qualify the last point, it may be useful to classify characters in a few ways. For example, certain things are very effective against fast fallers in general but not floaties. So perhaps the list could list a characters effectiveness against classes of characters

This would be a pretty big project requiring a lot of discussion and work, but I'd be happy to help once brawl comes out.

 

[House M.D.]

*one other point

of course this would be subjective. statistical measurements are only as good as the thought that goes into their design. this would just be alternative tier list, an approach to the problem using different tools than the sbr uses.

 

[Pye]

Hell, why don't they just make 2 tier lists: Tournament based and pure statistics based?

This was actually done at one point. One of the math wizzes on the boards created formulae that took into account as many play variables as he could think of and calculated "power levels" for each character. I'm still only a CEJEP student, so I didn't quite understand the workings of his math, let alone being able to repeat it for you, but I do remember that he took into account matchup strength as one of his variables, for each character.

I believe Shiek came up on top by the end of it. He had several versions of the list and altered his math a few times as he came up with better ways of calculating things or more variables to take into account, and Shiek was always on the top of the list. I think the reason for that was because her ridiculous **** matchups (Bowser, for example) with most of the lower tiers boosted her score significantly.

That's why I don't believe a statistical tier list would be a good method. There are too, too, too many variables to take into account. Also, if you think about it, a statistical tier list would have to take into account relative abundances of characters in tourny play anyway, otherwise you get a result like Shiek in top tier. It may not be incorrect...but it makes a false assumption: that in any given tourny match, a player has an equal chance of encountering any of the 26 characters. So Shiek is the best character in an environment where she can encounter a Bowser, a Ness or a Mario just as often as a Fox or a Falco, but that's not the case.

Like I said, I'm no math major, but I'm quite adept for math, and I have a feeling that, the more and more variables you added to a statistically calculated tier list, the more it would resemble our current empirical tier list.

 

[Pye]

EDIT: I apologise about the double post, my computer exploded

 

[Same-Move Sammy]

I agree, but if...

the methods were easily presentable, showing including all assumed weights
the methods involved factors that gave higher priority to more recent data and more serious tournaments
there could be confidence ratings (these would be useful on rarely-used characters)

...we might be able to develop a system that performs decently well and works in other fighting games. If people were able to tune the parameters on their own and compare results. Will we come up with something better than the debate-way of creating tiers? Maybe not, but it may be good enough for a master's thesis.

 

[Zankoku]

Actually, Wesley fixed that by lowering the effect that low tiers had on power ranking. It did this in real-time by figuring which characters would be more representative in a tournament and thus to the metagame at each point.

I don't remember exactly, but I'm pretty sure that Falco was #1 with Sheik at #2, since Falco won against everyone and Sheik had a largely favorable matchup against the majority of the cast even in the (2006-2007 tierlist) middle and high tiers.

 

[D4RK_HUNT3R]

Popularity? Formula? "Accurate data"? What are you talking about?

Are we even on the same page? Look at that ESPN link again, and that's closer to what I mean.

 

[D4RK_HUNT3R]

Why are matchups needed to be taken into account when doing a mathematical tier list? Can't you apply the same character to every other?

 

[eLantern]

First of all, baby steps need to be taken before there can be character match up charts. Those first steps should be that tons of data need to be collected, double checked and confirmed. There is a thread that I recently started that is attempting to gather character data. Hopefully as time goes on the data collected will be built up enough that formulas can begin to be figured out in an attempt to represent... as D4RK HUNT3R mentioned "a statistical tier list" or we can find out if it backs up the eventual agreed apon tournament based tier list. I like HUNT3R's idea of getting Pro Smashers input on which characteristic are more valueable and adjusting the formula accordingly, but all of this is possible only with more hands on time with the game.

That data thread is...
http://www.smashboards.com/showthread.php?t=146671

 

[jellis186]

Dog, I can tell you this now. I am a colligate level Triple/Long/High Jumper (Bests: TJ-51'3",LJ-24'11" HJ-6'8") and I can tell you that in no way is the speed involved in high jump (listed at 6.13) comparative to the speed in long or triple jump (6.75). And the analytical skills for long/triple (3.13) should be lower than high jumps (2.88). And to top it all off, to place both long and triple jump as the same event is, for lack of a better word, moronic.

What I'm saying is, the list is flawed. Experts can attempt to measure as much as they want, but there is still some bias, there is no magical formula to measure the best of the best. Empirical data is by far the best way to come up with a tier list especially for a metagame this intricate

 

[Meta_Sonic64]

Do you mean tiers by like
Speed: shows the list of all the characters and Sonic is at the top
Recovery: sames as Speed except ROB or Pit is at the number 1 spot
Power: DDD

and etc.

 

[LunaEqualsLuna]

Not possible to mathematically determine the effects of weight, jump height and move speed on a character.

You CAN use statistics and win loss data to emperically create a tier list, but i highly doubt its worth the effort and in the end you would probably just end up deliberately (or inadvertently) 'forcing' the stats to agree with the established tier lists anyway. Since a statistically based tier list doesn't take into account the most important data, which is the ACTUAL game itself.

You would be like, 'oh noez the formula says pichu is top tier!' something must be wrong and you would have to tweak the formula to make pichu come lower.

It will still end up subjective. Don't misapply mathematics... it usually just ends up wasting your time =D

 

[Meta_Sonic64]

Not possible to mathematically determine the effects of weight, jump height and move speed on a character.

Really, b/c even Sakurai noted Sonic to have the fastest move speed, not only that, but we have a video featuring the speed of every character. It is possible.

 

[Same-Move Sammy]

You CAN use statistics and win loss data to emperically create a tier list, but i highly doubt its worth the effort and in the end you would probably just end up deliberately (or inadvertently) 'forcing' the stats to agree with the established tier lists anyway. Since a statistically based tier list doesn't take into account the most important data, which is the ACTUAL game itself.

You would be like, 'oh noez the formula says pichu is top tier!' something must be wrong and you would have to tweak the formula to make pichu come lower.

It will still end up subjective. Don't misapply mathematics... it usually just ends up wasting your time =D

You can apply mathematics however you want. Forcing the model to fit what people think the tier list should be like may be inevitable. However, if the same algorithm (with game-nonspecific parameters) is successful in other fighting games as well, it would be interesting.

 

[iron blade]

The player using the character's abilities determines the tier. By your logic, a Bowser player would automatically be better than someone like Fox because he's stronger and heavier.

 

[Arteen]

Really, b/c even Sakurai noted Sonic to have the fastest move speed, not only that, but we have a video featuring the speed of every character. It is possible.

Keyword, effects. You missed the point.

 

[Tristan Fattest of All Kids]

If you like it that way, then go play football.

 

[Ced The Lad]

I don't think comparisons of sports to one another is essentially the best analogy to make here. This would only be accurate if we were trying to associate how technically sound we needed to be to play melee and compare it to Street Fighter, Guilty Gear, Power Stone, Tekken, etc.

I think a better analogy is to use some type of ESPN ranking that focuses on teams in one sport.

Now, you and others argue that you want character traits to determine the tier list. I equate this to ranking teams on paper only. Do you see where the problem lies here? It's too subjective depending on who you talk to. You'd have people saying something like "Golden State Warriors are the best running team in the NBA, so they're the best". An even better example are the Detroit Tigers of the MLB. Just because they added a big bat to their already dangerous lineup, we already see people hailing them as one of the top teams in the league. They have a good chance to, but how do we know for sure? What about matchups that don't favor them? (analogy of counter characters) What about ballparks that limit home runs thanks to the weather? (analogy of stages)

What any respected rank list in any sport does is rank teams based on performance. Regardless of the Patriots loss to the Giants (hooray!) their record indicates that they were the best overall team this past season. Similarly, a tier list should rank the characters' overall performances in as many competitive tournaments as possible. This is really the only accurate way to make any sort of rankign whatsoever.

Character potential is too random to measure on its own, but for those looking for such a list, a credible tier list made from tourney results will give you some idea as to what characters do have more going for them. And even though character potential can be somewhat measured by a ranking, remember that these things end up changing from year to year; especially when the game is less than 5 years old.

 

[shortwanabelaker]

so i was thinking about this a lot.. and a statistical tier list wud be really good id say.. based on potential and not not usage. there are a LOT of categories to rate (not rank) characters and everything wud be extremely theoretical, but i think its possible.

 

[Meta_Sonic64]

Keyword, effects. You missed the point.

So basically we don't know how the speed effects affect Sonic?

 

[LunaEqualsLuna]

So basically we don't know how the speed effects affect Sonic?

You cannot QUANTIFY how speed affects how strong a character sonic is in MATHEMATICAL terms...

Dude did you get dropped on your head as a baby or what
how can u repeatedly misunderstand like this...

 

[Meta_Sonic64]

You cannot QUANTIFY how speed affects sonic in MATHEMATICAL terms...

Dude did you get dropped on your head as a baby or what
how can u repeatedly misunderstand like this...

1. You can do a scale of 1-10 and Sonic's speed will be 10
2. You can do a scale for speed in word such as Very Fast-Slow

 

[LunaEqualsLuna]

1. You can do a scale of 1-10 and Sonic's speed will be 10
2. You can do a scale for speed in word such as Very Fast-Slow

are you trying to troll to be funny or you seriously don't understand what i'm trying to say?

If sonic has a speed of 10, and Bowser has a speed of 1, but bowser has a strength of 10, and sonic has a strength of 1.

Which one is better?

The point is you cannot mathematically say what ----> EFFECT<----- (NOTE THIS KEYWORD) that the speed has on how good the character is. is having a speed of '10' ten times better than a speed of '1'?
Is it 5 times better? you cannot mathematically define that. It will have to be subjective just like the current method of determining tiers, which is pretty much the only way to sensibly do it.

 

[Zaza]

There are so many reasons why this "more statistically based tier list" wont work that I barely know where to begin.

You want to create a list that is based off of hard data so that it is more concrete. I get that. But how are you going to convert those numbers into a ranking system and still preserve the concreteness of indisputable data? You can't. You would have to create a mathematical function to do so and there is no such intrinsic function in existence. You would need to create one yourself, and thats where subjectiveness comes in. We could all make our own function independently and no two would be the same. Not only that, I can guarantee that creating an accurate function would take an unimaginable amount of work. Seriously. I don't even want to go into the details of how hard it would be but trust me.

Here's the kicker though. There's no way to verify that the function you create is correct except but comparing it to the tier list based off of tournament results. Because the real tier list is the only trustworthy standard.

Besides, the tier list is very good at what it does. Not only that, it is the best conceivable way to rank characters. It really is.

It was a noble thought to create a new and better list, and its understandable that you didn't have the knowhow to realize it can't be done. But it cant be done.

I may not know as much about smash as some others, but I know my science and statistics.