Meta Knight Officially Banned!

[fkacyan]

I dunno, there have been scenarios, at least in my case, where I handily defeated some multi-mainers' not MK, then got ***** by their MK for the remainder of the set.

There are actually quite a few players in NYC where this occurs to them; not just to me, actually, so idk...

Has it ever occured to you that maybe, just maybe, the higher echelon of players in NYC is just that much better than you?

 

[Tommy_G]

And replied to the video.(whether on YouTube or here) No need to nearly double posting to tell us. Oy.

I don't appreciate your trolling good sir.

 

[Dr. Tuen]

forgot to respond to this... but plz remember to let me know, I feel it will get completely buried wherever it is posted.

I feel I didnt explain my point enough though and I still want to know how your approach would deal with this.

I think that MK is too good, but not good enough that he is bannable. We have seen it countless times before, the best MK's get beaten by all sorts of characters, sometimes multiple times over many meetings. This is totally independent to his dominance however which is undeniable and reason to be banned. This is simply because they best players will more often than not, use metaknight. If someone out there has serious determination to win and has the skill/practice to back it up, why would they use anyone but MK? Sure theres the 'MK mains wont know my characters matchup properly' argument but that only goes so far.

Whether MK was technically 1% more viable than snake overall, or 100% more viable, he is always going to attract the best players. So now we have two issues, MK has an increased dominance in a feedback-loop scenario where the best players use him, so the next best players will follow suit in order to beat the best players. Then there is MK's natural strengths which will artificially inflate MK mains placings in tournaments, simply because he is the best character and this would happen to anyone.

The point is, how far beyond the next best character, is MK? You will never know by looking at his dominance alone. Because that reason applies to the best character in ANY game. What options do we have to analyse how good MK is, if it is impossible to remove those two influences on his dominance? How can one do this in an unbiased tournament situation, the instant you attempt to measure this by introducing additional constraints, you have biased it already and people will argue this all day, and fairly too.

Hence my overall standpoint on this, whether MK is truly broken or not is irrelevant. Trying to prove this beyond reasonable doubt... I dont see that as being possible. By all means, show us your work but I wll be highly skeptical, the same I was with John#'s data even though I fully approve a ban, Ill say right here that his data proves nothing, its an opinion with biased numbers attached to it. What does matter, is what the actual players of the game want. If you dont listen to the community, you have absolutely no right to dictate rules in a self-appointed position of authority.

So its that simple. To me, both sides have nothing but biased opinions and data which can easily be manipulated to give the results you want (I know all about these tricks). What separates them, is what the actual players want. If the vast majority all come to the conclusion that MK is harming the scene and detrimental to it in all aspects, then you must accept it. If anti-banners in a position of authority deny that on the basis of them being 'wrong' without any data to back it up which is utterly impossible for them to provide which cant be countered with equally skewed 'data', then they should have no say on the matter.

Yeah, I'm still dealing with some of these interesting interactions. The best we can do is to try and look at each factor individually, and if possible show interactions.

The three things I'm going to look at are frequency of appearance in the top 8, frequency of usage in the top 1000 players, and repeat player appearance.

Frequency in the top 8 - This gets at the "dominance" issue. If MK appears in the top 8 more than everyone else, and on top of that his results average higher than everyone else's, then he could be considered "dominant". Given this is frequency data and not ranking data, this will allow for a more robust analysis technique.

Frequency of usage in the top 1000 players - This gets at the "over-centralization" issue. Granted, the previous point can address this too. Anyways, we've already got an idea of who gets used more than who, but there are statistical metrics utilized in research that can determine if the jump in usage is considered "significant". This gets a bit difficult to utilize though, since some assumptions must be made.

In standard frequency analysis, you assume all choices are perfectly equal and go from there. In ANY fighting game, balanced or not, this will turn up a significant result. Even if MK were involved in a countering triangle (similar to Fox>Marth>Sheik>Fox in Melee), comparing him to Ganon as if they were to be used equally would turn up a significant result.

To fix this, you can apply what can be estimated to be a "standard" for this type of ranking distribution. So we'd say maybe there's a linear relationship between tier list position and popularity. You may even go as far as using an exponential curve, but great care must be taken if doing that. I believe the popularity in smash IS an exponential curve, which means one could be fit to our data. If you fit the curve to the data and THEN compare the frequency data, nothing will come out as significant. So it's real dicey, to say the least. I may just present multiple formats for people to make their own assessments.

Repeat Player Appearance - This is the most important part to me. Smash attracts a younger player base than other fighting games and that has a significant effect on attendance because of economic concerns. But let me put forth a completely unreasonable hypothetical situation. Say I, in my "project zero" efforts, become the best Zero Suit the wooooorld has ever seen! And win Apex. Oh I wish, right? Winning my first national. :-D. Do you think my chances of attending another national goes up or down? I'd say up. Way up!

So to me, this says there's a kind of win-barrier. When you start placing or winning at big enough events, it facilitates your attendance for future events. So there may be a lot to be said about people who are able to attend large events multiple times. They could be the ones capable of breaking the "win-barrier". I may be able to look at those who have attended X number of events and subsequently look at their character distribution. It could prove to be insightful.

That said...

Estimated Time of Completion - I was bold in saying "four days" before. It turns out, I want a lot more data than I thought. Like... an entire year's worth. Of SWF AND AiB. So now I'm not sure when this will be done. It's a lot of data mining, to be honest. I will, however keep you in mind when publishing the results. Though it couldn't hurt to friend me here or on AiB (I'm also "Tuen" there) to make sure I remember :-p.

Thanks for your interest in my work!

 

[Tommy_G]

Over-centralization is not good criteria for banning a character.

If no characters were changed, and everyone just happened to play Captain Falcon, would that mean he's bannable too? simply because he's played often?

 

[fkacyan]

Tuen is a cool guy, he's like me but with patience and motivation.

I shall watch your progress with great interest.

 

[Black Mantis]

If no characters were changed, and everyone just happened to play Captain Falcon, would that mean he's bannable too? simply because he's played often?

but captain falcon has bad matchups and even ones (unlike mk)

 

[John12346]

Overcentralization alone isn't enough of a ban criteria.

However, when it's combined with the many other criteria set up for the ban, it holds quite a bit of merit.

Has it ever occured to you that maybe, just maybe, the higher echelon of players in NYC is just that much better than you?

Perhaps, but I wasn't speaking just for myself there. It's happened to plenty of good players here in NYC too, not just the mere decently skilled me.

 

[Exceladon City]

I bet you lose to good players who do practice. It's not realistic to win even with MK without practice.

Nope. He took second seed in his pool at SiiS6 and he only lost to Shugo in his pool.

Pretty sure that OS just learned people instead. At least here in Ohio he does particularly well against the tops of our state without practicing (even though he's in KY now).

^This.
I remember him not playing for almost a year, getting a weeks worth of practice in before SiiS3 and placing like 13th.

Then, winning a tournament months later after not playing the game since then. Beating Infern and Y.b.M. as if he had never left.

Over-centralization is not good criteria for banning a character.

If no characters were changed, and everyone just happened to play Captain Falcon, would that mean he's bannable too? simply because he's played often?

When the over-centralization is damaging the state of the game then it's a problem. MK is literally "Pick me or lose." With the exception of the few players that outplay MK mains below them, it's pretty much that.

There's no real way around him, you have to plan accordingly. You have to play each set like you're gonna be playing against MK, regardless of what character you're facing at that point in time. Every stage you pick or ban determines how much closer you are to fighting a MK.

 

[Battousai780]

Is it really true that in order to have a smash tourney stickied on SWF it must use Unity ruleset?

 

[Hippieslayer]

Over-centralization is not good criteria for banning a character.

If no characters were changed, and everyone just happened to play Captain Falcon, would that mean he's bannable too? simply because he's played often?

Nominated for worst allegory of the year!

 

[Steam]

MK isn't a case of "pick him or lose"... not even close. But I will say that the most important thing to have in your toolbox is a decent MK.

however he's always treaded that line of being too good... and I think our changes to the ruleset are the reason for this. I agree with the ban for the sole reason that I (and others) want to see what a MK banned metagame actually looks like. The metagame will just continue to be stagnant with MK legal as his presence kinda stuffs any other progress since everything is focused on him :|

 

[infiniteV115]

If it's a Brawl-only tourney, then yes, it must use either the Unity Ruleset or an approved Experimental Ruleset (which falls under Unity but you get the idea)

 

[Kink-Link5]

Melee you can do whatever the **** though.

 

[Dr. Tuen]

Tuen is a cool guy, he's like me but with patience and motivation.

I shall watch your progress with great interest.

I look forward to publishing the results! If it ever gets done... sometimes I think I bite off more than I can chew given that I already have a job alongside my PhD research >.>.

Overcentralization alone isn't enough of a ban criteria.

However, when it's combined with the many other criteria set up for the ban, it holds quite a bit of merit.

Yes, exactly. And it should be noted that particularly extreme results should be noted too. That's why I'm going to the lengths that I am. I tried an over-centralization analysis with the data that I do have and... well... I broke statistics. The level of significance achieved is something I have never witnessed in my entire life! So I chose not to report it before having more to back it up... I don't want to sensationalize the public just as they start taking a serious look at my work.

 

[John12346]

Hah, you get just as excited as I do when some crazy numbers pop up in statistical work.

 

[fkacyan]

Math of all kinds is awesome (It's my major!) but I've learned the hard way that stats mean nothing to most Smashers. Or didn't. Maybe they do now! I might have to revise my motivation.

EDIT: Tuen, what was the levelof significance?

 

[Dr. Tuen]

Math of all kinds is awesome (It's my major!) but I've learned the hard way that stats mean nothing to most Smashers. Or didn't. Maybe they do now! I might have to revise my motivation.

EDIT: Tuen, what was the levelof significance?

Take this with an entire salt shaker of salt, the I MUST finish collecting my new data set to check this again.

But for those who understand p-values in statistics, I got a p-value of ZERO. To put this in prospective, Microsoft Excel can hold numbers as small as 1E-308.

This was via an Pearson's Chi-Squared test on the character distribution in the top 1000. This is also using the version of the test where you assume even distribution amongst the 36 character.

I did a post-hoc (second test, pretty much) analysis on each of the 36 factors to see which ones contributed the most to the result. Even after stripping away 30 characters, I get a p-value of ZERO.

So yeah. You may see why I really want to re-check that. It's a result so fantastically absurd, it must be checked twice.

 

[fkacyan]

A P value of... Zero...?

P values? I check that.

 

[John12346]

I'm a bit rusty on my statistics, but a p-value of 0 means that there's no possible chance that whatever conclusion you drew happened by mere chance, correct?

 

[Dr. Tuen]

I'm a bit rusty on my statistics, but a p-value of 0 means that there's no possible chance that whatever conclusion you drew happened by mere chance, correct?

Yep, in essence it's a 100% statistically significant difference between the groups being compared. 0.05 ~ 95% significance, 0.01 ~99% significance, and so on. I've seen things as low as 0.001 and 0.0001 before, but nothing like this.

 

[Tommy_G]

Nominated for worst allegory of the year!

Nominated for the most useless post of the year.

No proof to disprove my argument or to say anything intelligent, not to mention not knowing what allegory means.

I can't believe there aren't more infractions given for ridiculous posts like this...the mods are probably pro-ban and will only give infractions to anti-bans.

 

[DMG]

Now you're dipping into dumb dumb land

Like for real you might as well just stay home and not even post anymore.

 

[fkacyan]

I'm a bit rusty on my statistics, but a p-value of 0 means that there's no possible chance that whatever conclusion you drew happened by mere chance, correct?

Yep. A lot of statistical tests are pretty strict as to the amount of chance they allow before the test can't be used for anything (p<.05 or p<.005 are very common), but in the years I've been doing things with numbers I've never hit upon an actual p-value of zero. I've gotten very low numbers that might as well be zero, but...

 

[Elyssa Xey Hexen]

What null hypothesis were you testing that you got a p value of zero?
Oddly enough I learned this from my genetics class, so I might be a bit off using the null hypothesis for this question lol

 

[Dr. Tuen]

What null hypothesis were you testing that you got a p value of zero?
Oddly enough I learned this from my genetics class, so I might be a bit off using the null hypothesis for this question lol

The null hypothesis states that the distribution of character choice is indistinguishable. So you wouldn't be able to use statistics to differentiate between MK, Diddy, or Snake... if they were equal-ish. I tried this for the whole cast at first (pretty unrealistic, given the difference between the top tier and the triforce tier), and as per follow up tests I stripped away characters to see which ones were contributing the most to the significant result. Even when the top 6 are considered, MK still causes a p-value of zero. It's crazy >.>

 

[Raziek]

Can you put that into non-statistical terms, Tuen?

What exactly does the p-value of 0 mean?

 

[Elyssa Xey Hexen]

Interesting to see statistics put to use like this. Perhaps I should have taken a stats course at sometime to know more about the methods. Oh, well. Not required even for a math/physics major.

 

[link2702]

reason i'm a musician not a mathematician....y'all lost me a long time ago with the "p value of zero" or w/e

 

[rm88]

Over-centralization is not good criteria for banning a character.

If no characters were changed, and everyone just happened to play Captain Falcon, would that mean he's bannable too? simply because he's played often?

If everyone just happened to play Captain Falcon, pretty much every character in the game would be viable. That's kind of important.

Let's not be silly. Everyone doesn't "just happen" to play MK, there's a reason why he's so overused.

 

[Elyssa Xey Hexen]

If everyone just happened to play Captain Falcon, pretty much every character in the game would be viable. That's kind of important.

lI can sense a "ban A tier" argument developing from this lol

 

[rm88]

I personally don't believe anyone in A tier deserves banning, not even close. Those 3 characters are perfectly manageable for a lot of characters, none of them are "universal counters".

 

[Ripple]

I'd ban Falcon if he were over centralizing even if he wasn't the best

 

[rm88]

Still, it's a situation that has yet to happen; a low tier character with tons of bad MUs dominating the tournament scene of any game.

 

[John12346]

I've always felt that overcentralization alone wasn't enough of a criteria of a ban, but rather one of the factors. I mean, if Falcon was used as much as MK, but took only, say 2 or 3% of all tournament money because of it(which is extremely unlikely), I wouldn't really be clamoring for a ban right off the bat.

On this one specific criteria, I feel a bannable offense would preclude being both overused and taking too much money from tournaments, not really one or the other necessarily.

 

[rm88]

The thing, MK is not overused "just because". It's not like everyone said "cool! I've always loved MetaKnight" and started using him. He is relatively easy to learn and has pretty much no bad MUs, while being a really safe character with almost no downsides. No other character is comparable to that description, that's simply unfair.

 

[LKratos]

Yep, in essence it's a 100% statistically significant difference between the groups being compared. 0.05 ~ 95% significance, 0.01 ~99% significance, and so on. I've seen things as low as 0.001 and 0.0001 before, but nothing like this.

If you're using chi-squared, the variables have to be independent correct? I'm pretty sure that not only are characters not evenly distributed (it would be more like a normal distributive curve with MK at 50% and tiers descending outward), I'm pretty sure character picks aren't independent of each other (they factor in matchups against other characters, low tiers MUs against high tiers, etc.), but I'm pretty sure even tournament results aren't completely independent (people may see previous results and base their picks on that).

I was about to ask for your alpha, but then I realized it doesn't really matter with a p of 0 XD

But yeah, assuming independence and even distribution, I could see where MK winning so much could get a p of 0. But I'm not terribly good with statistics, so I'm not sure exactly how to go about testing this. I've never tried a 1000 test 36 variable significance test before lawl.

 

[Thino]

Well congrats anyways on scaring away the top players the ones that provided us with good shows on livestreams while reaffirming your casual community.

good shows is subjective, what you consider a good show, might not be the same good show as everyone else

plus I'm much more okay with this decision being influenced by tournament attendance , rather than stream quality

Yeah, I'm still dealing with some of these interesting interactions. The best we can do is to try and look at each factor individually, and if possible show interactions.

The three things I'm going to look at are frequency of appearance in the top 8, frequency of usage in the top 1000 players, and repeat player appearance.

Frequency in the top 8 - This gets at the "dominance" issue. If MK appears in the top 8 more than everyone else, and on top of that his results average higher than everyone else's, then he could be considered "dominant". Given this is frequency data and not ranking data, this will allow for a more robust analysis technique.

Frequency of usage in the top 1000 players - This gets at the "over-centralization" issue. Granted, the previous point can address this too. Anyways, we've already got an idea of who gets used more than who, but there are statistical metrics utilized in research that can determine if the jump in usage is considered "significant". This gets a bit difficult to utilize though, since some assumptions must be made.

In standard frequency analysis, you assume all choices are perfectly equal and go from there. In ANY fighting game, balanced or not, this will turn up a significant result. Even if MK were involved in a countering triangle (similar to Fox>Marth>Sheik>Fox in Melee), comparing him to Ganon as if they were to be used equally would turn up a significant result.

To fix this, you can apply what can be estimated to be a "standard" for this type of ranking distribution. So we'd say maybe there's a linear relationship between tier list position and popularity. You may even go as far as using an exponential curve, but great care must be taken if doing that. I believe the popularity in smash IS an exponential curve, which means one could be fit to our data. If you fit the curve to the data and THEN compare the frequency data, nothing will come out as significant. So it's real dicey, to say the least. I may just present multiple formats for people to make their own assessments.

Repeat Player Appearance - This is the most important part to me. Smash attracts a younger player base than other fighting games and that has a significant effect on attendance because of economic concerns. But let me put forth a completely unreasonable hypothetical situation. Say I, in my "project zero" efforts, become the best Zero Suit the wooooorld has ever seen! And win Apex. Oh I wish, right? Winning my first national. :-D. Do you think my chances of attending another national goes up or down? I'd say up. Way up!

So to me, this says there's a kind of win-barrier. When you start placing or winning at big enough events, it facilitates your attendance for future events. So there may be a lot to be said about people who are able to attend large events multiple times. They could be the ones capable of breaking the "win-barrier". I may be able to look at those who have attended X number of events and subsequently look at their character distribution. It could prove to be insightful.

That said...

Estimated Time of Completion - I was bold in saying "four days" before. It turns out, I want a lot more data than I thought. Like... an entire year's worth. Of SWF AND AiB. So now I'm not sure when this will be done. It's a lot of data mining, to be honest. I will, however keep you in mind when publishing the results. Though it couldn't hurt to friend me here or on AiB (I'm also "Tuen" there) to make sure I remember :-p.

Thanks for your interest in my work!

These are interesting statistics but what are they for?

 

[Grim Tuesday]

Perhaps you should actually play other fighting games before assuming your favorite to be the most difficult.

For starters, I said one of the most technical, it is an important distinction to make.

I main one of the hardest characters timing wise in both Blaz Blue and Melee (Carl and ICs) and I consider the two games to be of pretty similar technicality, though Melee certainly feels faster...

Of course, I could be wrong, lets look at two other fighting games I play. Melee is more technical than SF4 and 3S.

Oh, also MvC3, oh hey, Melee is more technical.

You're spouting theory.

As I said before, people can't be one hundred percent sure that this is going to happen.

Well... duh. He wasn't trying to imply that he could predict the future, lol, of course it was just a theory.

 

[Dr. Tuen]

Can you put that into non-statistical terms, Tuen?

What exactly does the p-value of 0 mean?

The p-value helps determine if something is statistically "different" (in this case, is MK "different" from the rest of the cast). If you have a value close to 1, it means the two things you are looking at are indistinguishable from one another. If the value is close to 0, the two are "statistically significant" or "really really different". And once things become significant, it's usually easy to tell how they are significant (we know MK differing significantly would be him being over the rest of the cast, not under).

Hope that help!

 

[Dr. Tuen]

If you're using chi-squared, the variables have to be independent correct? I'm pretty sure that not only are characters not evenly distributed (it would be more like a normal distributive curve with MK at 50% and tiers descending outward), I'm pretty sure character picks aren't independent of each other (they factor in matchups against other characters, low tiers MUs against high tiers, etc.), but I'm pretty sure even tournament results aren't completely independent (people may see previous results and base their picks on that).

I was about to ask for your alpha, but then I realized it doesn't really matter with a p of 0 XD

But yeah, assuming independence and even distribution, I could see where MK winning so much could get a p of 0. But I'm not terribly good with statistics, so I'm not sure exactly how to go about testing this. I've never tried a 1000 test 36 variable significance test before lawl.

It's still independent in that the person picking the character is not effected by another other peoples choices, I guess (though you could have counterpicks for those that main more than one character, but their choice to do so was fairly independent). The tier list effecting decisions is the reason why a linear (or even exponential) distribution would be good to test against, as opposed to an even distribution.

Just so we're clear here, this isn't an "official" result. I'm prepping another, more robust analysis for this purpose.