Official Metaknight Discussion

[HeroMystic]

Serious Question: Why was 90 points used as the par?

That's my only problem with the calculation.

 

[Crow!]

@ Omni's recent numbers:

I think Omni missed my post explaining the traps to avoid when analyzing the data in order to avoid invalidating it, because he proceeded to blunder into the biggest problem anyway. Here's a link to it: http://www.smashboards.com/showpost.php?p=9476865&postcount=2159

Omni's method here is to decide upon some cutoff to classify players as "mattering" and the divide the total points - a pretty natural sounding thing to do. Unfortunately, this does several things to ruin otherwise usable data:

First, the cutoff he chose leaves too few data points to take the results from anyone other than MK seriously. N=2 for Snake, ICs, and Marth, and N=3 on Diddy? I know Omni's not into statistics, so rather than insult Omni for leaving things that way I'm just going to say: you can't do statistics on numbers that low; even if the "set a cutoff value for importance and average from there" method is accepted those numbers are just too small to analyze Snake individually, or ICs individually, etc.

Second, and worse, it averages out the data points that it does leave rather than showing the distribution - when you're trying to compare 13 MKs to 2 Snakes, placing as much importance in the 12th best MK as on the 1st place Snake and 2nd place MK is pretty ridiculous. Omni's results are in no way surprising given this: MK is the most represented character. MK therefore is very likely to have the largest number of mains who are close to Omni's arbitrarily selected 90 point cutoff, which is most likely driving his average down in comparison to characters whose only mains to reach the cutoff selected were those who are doing especially well.

One of the most important (and hardest) objectives in research is to make sure your methods don't influence the results. Unfortunately as it is right now, the cutoff method makes any report based on it tell you more about the cutoff selected than about the data being analyzed.

This CAN be repaired; maybe the data could well have something important to show when it's not being obscured by the methods used. I'm sending Omni a PM; I have an idea to display the same data in a way which actually tells what's going on; he's sooo close to having it. Right now I have a guess as to what the data is, but until it's presented properly we can't really know.


Oh, by the way, the cutoff method isn't a total wash. It avoided cherrypicking, so some stuff can be pulled out of it safely. While the cutoff screws up interpretation of scoring data, the cutoff ITSELF can be studied, so you see how many mains of each character managed to contribute at least 90 points to Ankoku's rankings list. Thanks to Omni, we now have that data tallied up! I'll just post it here in pretty, pie form, and leave it for other people to interpret.

N=22 for the above graph, which seems fairly reasonable to use in this case.

 

[Omni]

The 90 point cut-off was recommended by Ankoku.

You do realize the numbers get worse for MK as we go down the list, right? The amount of MK's increase rapidly and his average grows lower and lower.

Simply put, I took a specific sample size that reflects the Top 30 (29 in this case) players. There was no other reason besides that.

Edit: Crow, did you even see the Top 30 chart that Ankoku posted? There was a very LARGE amount of MK players way above the 200 mark INCLUDING M2K who was at 800! That more than makes up for any MK that was near the 90 mark.

 

[etecoon]

You haven't proved me wrong... at all. I explicitly said 4 months...

after the fact, that's why I said you were backpedaling, you modify your statement after you discover that you were wrong.

Maybe you're just trolling me? That's the conclusion I'm reaching.

you're much further under the bridge than I am, don't be pointing at the kettle

 

[Overswarm]

There are a lot more of MK players, do I really need to prove that or should I google them(or equivalent)? Other than this, what assumptions have I made?

Yes. You'd need to compare the numbers of MK to the numbers of Diddys, Snakes, etc., etc.

It'd also be prudent to see how many there used to be compared to now, but that's near impossible.

I'd also like to note that your assumption that "picking out of 1,000 or 100" leads to better results from the 1,000 is incorrect. If you were picking random numbers out of a bag, yes, your odds of picking a big number would increase... but that isn't the case. This is a competition, so when you pick 1,000 characters and then 100 of another and compare tournament placings, you're only getting the cream of the crop from the 100 but you're getting a lot of crappy ones from the 1k. Once you divide by 1,000 from the good character, you've mangled the number in such an obvious way that it doesn't represent anything.

You don't fight against the best MKs in the world and say "Sure, HE'S good, but if I divide his skill level by the number of MKs on average and compare it to the number of my character mains at the tournaments and my skill level, I technically have an advantage!"

 

[Overswarm]

The 90 point cut-off was recommended by Ankoku.

You do realize the numbers get worse for MK as we go down the list, right? The amount of MK's increase rapidly and his average grows lower and lower.

Simply put, I took a specific sample size that reflects the Top 30 (29 in this case) players. There was no other reason besides that.

You didn't just get ninjaed.

You got ninja WRECKED.

 

[Flayl]

after the fact, that's why I said you were backpedaling, you modify your statement after you discover that you were wrong.



you're much further under the bridge than I am, don't be pointing at the kettle

Maybe if if I go step by step with you:

I asked why no Falco besides DEHF has taken a set off of a top MK in the past 4 months (you know, the period of time I used to collect data)

You mention SK92 beating DSF

I say DSF hasn't lost to SK92 in the past 4 months (you know, the period of time I used to collect data)

You accuse me of backpedalling somehow.

Glad you saw you were wrong about me backpedalling about Razer though.

 

[Kuraudo]

This is why we can't have good things, pro-ban. You ask for stats; Omni gives you stats. Overswarm sees stats. Overswarm says stats are bad. \

I really don't understand why no other pro-ban players are giving their opinion on what was said. You guys ask for a decent argument, you get one, then it's either completely shrugged off and/or ignored.

For shame.

We could say the same thing about anti-ban.

We give you what you want, and you **** on it anyway. This is going back and forth. Now it's just an endless cycle.

**** this. I might as well either main Meta Knight, or just give up in general. The community's killing me.

I don't even care for if he's banned or not anymore. Put me back on the fence. Picking a side blows.

 

[Omni]

Also, that chart is not the information I was attempting to deliver, Crow.

The most important information were the average character ranking points. We can simply look at the Character Rankings list and count that information without needing a pie chart or graph.

@Kuraudo: Read the last part of my latest post. It explains how you're feeling in a nutshell.

@Overswarm: Uhuh. Let me know when you're no longer mad.

 

[etecoon]

because you said that afterwards.

I deleted/edited my previous comment to SFP about this but seriously, SWF needs an ignore feature. it would make topics like these much more tolerable if you could sort out the noise

 

[RDK]

Maybe, just maybe, I'm basing it off tournament outcomes. Of what happens in real life. That would be a good guess.

So tournament placements are what we're basing tiers off of now?

If CF takes the top 10 spots of the next national tourney can we expect to see him soar in the tier list?

Or maybe - just maybe - there's something more to tiers than tournament representation?

Like, I don't know, this is a shot in the dark now - the fact that M2K is a freak of nature who just so happens to be using the best character in the game?

 

[Overswarm]

This is why we can't have good things, pro-ban. You ask for stats; Omni gives you stats. Overswarm sees stats. Overswarm says stats are bad. \

I really don't understand why no other pro-ban players are giving their opinion on what was said. You guys ask for a decent argument, you get one, then it's either completely shrugged off and/or ignored.

For shame.

The stats were bad >_>

 

[Spelt]

MK debates will never end, and MK will never get banned.
if there was any slight hope of him getting banned something like a temp ban would've been implemented by now.
because that would just give us actual data to use and stop all this random theory crafting.
after arguing for a few years everyone will be on the fence.
anti-bans devious plan.

 

[swordgard]

Yes. You'd need to compare the numbers of MK to the numbers of Diddys, Snakes, etc., etc.

It'd also be prudent to see how many there used to be compared to now, but that's near impossible.

I'd also like to note that your assumption that "picking out of 1,000 or 100" leads to better results from the 1,000 is incorrect. If you were picking random numbers out of a bag, yes, your odds of picking a big number would increase... but that isn't the case. This is a competition, so when you pick 1,000 characters and then 100 of another and compare tournament placings, you're only getting the cream of the crop from the 100 but you're getting a lot of crappy ones from the 1k. Once you divide by 1,000 from the good character, you've mangled the number in such an obvious way that it doesn't represent anything.

You don't fight against the best MKs in the world and say "Sure, HE'S good, but if I divide his skill level by the number of MKs on average and compare it to the number of my character mains at the tournaments and my skill level, I technically have an advantage!"

Please, let me assume there are tons of MKs, one of the main arguments of proban is that MK is overcentralizing. I can tell you that there is probably at least a 1 to 5 ratio between number of MKs and snake.


Now, your argument falls appart when we take into account the bell curve. Il give you this distribution.

Please follow my reasonning:
Pretend there are 10 snake users vs 50 MK users. Out of these, 40% of them will fall approximately at the mean. 20% will fall just below the mean, and 20% just above. Finally you get a 10 % that will be REALLY good and another 10% that will be really bad.

This gives
1 really good snake, 2 good snakes, 4 average snakes and 2 bad snakes and 1 awefull one.


For MK, its 5 really good, 10 good, 20 average, 10 bad, 5 very bad.


Your method catters to taking stats only about the best top 3, that means:
3 really good MK
1 really good snake, 2 good snakes.

Method favors MK, not due to how good he is, but due to how popular he is. Now even with this, MK has a higher mean skill level due to the amount of pros that switched to him(and not to snake, pros switching to a better chars usually switch to the best, which is perfectly normal). DESPITE ALL OF THIS, MK and snake top 3 results are very close, MK didnt win a national in 8 months and doesn't even have any MK that gets close to ADHD, Ally or M2K in terms of wins/consistency. Maybe MK ain't as good as you make him out to be.

 

[Flayl]

because you said that afterwards.

I deleted/edited my previous comment to SFP about this but seriously, SWF needs an ignore feature. it would make topics like these much more tolerable if you could sort out the noise

1 - SWF has an ignore feature. Go to a user's public profile -> user list -> add to ignore list. Let me know if you're gonna ignore me though.

2 - http://www.smashboards.com/showpost.php?p=9480149&postcount=2342
Read the quote. I didn't change anything after the fact.

So tournament placements are what we're basing tiers off of now?

It's a very important part, we have to take centralization into account too.

If CF takes the top 10 spots of the next national tourney can we expect to see him soar in the tier list?

Not this **** again. People play to win. Cut it off with the hypothetical bulll****.

Or maybe - just maybe - there's something more to tiers than tournament representation?

Of course there is. I don't see how any of this is going to make Falco VS MK an even matchup, which is what I'm disputing in case you've forgottten.

Like, I don't know, this is a shot in the dark now - the fact that M2K is a freak of nature who just so happens to be using the best character in the game?

Yes, we all recognize he is a league above the other MKs. Still don't see how this is going to prove Falco VS MK is even.

 

[RDK]

Flayl, I never even disputed the idea that Falco vs. MK is even. What I'm calling you out on is your dependence on supposed matchup ratios when everyone here has seen you say that you think matchup ratios are basically meaningless.

And no, tournament placements have nothing to do with how good a character is.

 

[Flayl]

Flayl, I never even disputed the idea that Falco vs. MK is even. What I'm calling you out on is your dependence on supposed matchup ratios when everyone here has seen you say that you think matchup ratios are basically meaningless.

Jesus **** can you not read? I'm saying the ratios CURRENTLY ON THE BOARD are DOWNPLAYED. If they were correct, I wouldn't be disputing anything.

And no, tournament placements have nothing to do with how good a character is.

Nothing to do? Are you sure you want to stand by that affirmation?

 

[Overswarm]

Please, let me assume there are tons of MKs, one of the main arguments of proban is that MK is overcentralizing. I can tell you that there is probably at least a 1 to 5 ratio between number of MKs and snake.


Now, your argument falls appart when we take into account the bell curve. Il give you this distribution.

Please follow my reasonning:
Pretend there are 10 snake users vs 50 MK users. Out of these, 40% of them will fall approximately at the mean. 20% will fall just below the mean, and 20% just above. Finally you get a 10 % that will be REALLY good and another 10% that will be really bad.

This gives
1 really good snake, 2 good snakes, 4 average snakes and 2 bad snakes and 1 awefull one.


For MK, its 5 really good, 10 good, 20 average, 10 bad, 5 very bad.


Your method catters to taking stats only about the best top 3, that means:
3 really good MK
1 really good snake, 2 good snakes.

Method favors MK, not due to how good he is, but due to how popular he is. Now even with this, MK has a higher mean skill level due to the amount of pros that switched to him(and not to snake, pros switching to a better chars usually switch to the best, which is perfectly normal). DESPITE ALL OF THIS, MK and snake top 3 results are very close, MK didnt win a national in 8 months and doesn't even have any MK that gets close to ADHD, Ally or M2K in terms of wins/consistency. Maybe MK ain't as good as you make him out to be.

You just added another assumption: That we'd be using a bell curve when looking at skill level. This is obviously not the case, and you can actually make your own graph using Ankoku's data.

 

[swordgard]

You just added another assumption: That we'd be using a bell curve when looking at skill level. This is obviously not the case, and you can actually make your own graph using Ankoku's data.

Ankokus data only takes into consideration top 8 doesn't it? How could I do that >.>

 

[RDK]

Nothing to do? Are you sure you want to stand by that affirmation?

Uh, yes. Tournament placements are used when forming tier lists, but the placements themselves have no direct causative relationship with how good the characters are.

We can take into account which characters seem to win more and place better when doing matchup ratios and determining tiers. But that's about it.

 

[swordgard]

You just added another assumption: That we'd be using a bell curve when looking at skill level. This is obviously not the case, and you can actually make your own graph using Ankoku's data.

You know what, just for the hell of it, just assume it is. Do you find any logical problem with my argument? I don't have time to do a graph atm, might do one after this weekend. Just tell me if you find anything else wrong. If you dont and the graph shows what I said, your ****ed.

 

[HeroMystic]

Here's the problem I'm seeing as far as the calculation goes, which is what I'm seeing from Overswarm, Crow, and a few others.

It takes into account of popularity well, which is what we've been trying to kill off. What it doesn't take into account however is skill and attendance.

The 90 point cut-off is flawed in a way that it favors attendance to tournaments more than anything. Skill is out in the open because there's no way of gauging other players (which is where OS' argument comes in).

Otherwise it's a pretty legit calculation that just needs a few more variables pulled out.

 

[Overswarm]

I don't agree with any argument that starts off with "assume this this and this"

 

[MarKO X]

It just looks like everyone is manipulating numbers now.

That must be what's hot in the streets.

 

[Flayl]

I actually have a semantics question. Should I say the ratio of MK:Falco is downplaying MK's advantage, or can I simply say the ratio is downplayed. My bad if the only correct way to say is the former, although I'm pretty sure most people understood me either way.

 

[swordgard]

I don't agree with any argument that starts off with "assume this this and this"

Actually, the placings in themselves imply that there is a skill curve and that MK has a curve which does not favor top placings compared to diddy and snake.

 

[ADHD]

What if everyone just ****ing played Ice climbers instead of mk?

 

[swordgard]

What if everyone just ****ing played Ice climbers instead of mk?

It would be hilarious as ****.

 

[Eddie G]

What if everyone just ****ing played Ice climbers instead of mk?

Then I guaran****tee you there would be more Peaches, Lucas', and Snakes out there. Hell maybe even more Marths. That's one small example of MK's absence allowing those who were previously on the borderline of being viable to be well...viable (not really including Lucas overall though).

*cough* it isn't a bad thing either.

 

[ADHD]

It would be hilarious as ****.

Well, I don't mind. But the next thing noobs are to whine about without mk are the infinites. Just wait.

 

[Spelt]

What if everyone just ****ing played Ice climbers instead of mk?

it would be hilarious because nobody would be able to chain grab anybody.

 

[Tien2500]

Well, I don't mind. But the next thing noobs are to whine about without mk are the infinites. Just wait.

People have been arguing about IC's infinites but it isn't taken very seriously because its not a slippery slope. The community is in general smart enough to see the difference.

And if everyone played ICs Snake would start dominating tournaments and Falco would drop big time.

 

[MarKO X]

What if everyone just ****ing played Ice climbers instead of mk?

every match would be a doubles match.

 

[Delvro]

I love the fact that people are pulling out graphs, statistics, and bell curves. It makes me happy ^_^

However, it is absolutely vital to understand the assumptions you are making before you measure a random variable. While you are right in saying that with a larger sample (1000 as compared to 100), you are more likely to get an observation or two that is farther from the mean, you are assuming that player skill is normally distributed with the mean equal to the mode. I do not believe that this is the case.

To have a normal distribution, each sample from the population must be independant of every other variable. This is MOST DEFINITELY not true in brawl. Everyone should know that a region in the U.S. that has one great player tends to have a LOT of great players. The reason for this is obvious: Other players play the great player and get better practice, thus learning faster and getting better faster than other players in other regions.

So what you have is various (not quite random) distributions that vary greatly from region to region, but have surprisingly little inter-region variance.

Also, I bring up the point below, again.

Differences in population, while it would be helpful to know, turns out to be misleading because of two reasons. One is the fact that the subset of players that choose to play metaknight is difficult to measure. How many are there? Is the average MK player an accurate sample of the SSBB community at large? I do not think that it is. There is an unknown variable that determines whether a player switches to MK or not and we currently have no clue and no way to measure whether or not this variable has an affect on player skill.

Second, and more importantly, is that OS uses data where all but the top 8 placings in 100+ tournaments are thrown out. At this point, differences in popularity become less important. Let's say you add 20 (skilled) Link mains and 20 (skilled) mario mains to the Pound 4 tournament. Would the top 8 results have changed at all? I very much doubt it.
Add 20 Pit mains? It may change the results very slightly. Perhaps a highly lucky! (and highly skilled, of course, but skill is a variable which should be held constant for these arguments) player breaks top 8.
Add 20 MK mains? This one should be obvious (taking skill out of the equation)

The result is that while popularity DOES have an influence upon character placements, the importance is dampened by the fact that the character is better, plain and simple, and that only a constant sampling from each tournament (8) is drawn from these large samples with varying popularity for each character.

 

[CRASHiC]

There are a lot more of MK players

Erm. . . .

I'm sorry, but I'm not very sure how well this holds up when applied to competative games.

In Melee, there is only 1 bowser who can hang at the top.
But in Brawl, we only see 1 Snake reaching the top despite the large numbers of Snake, and even more, that Snake player has picked up Metaknight in order to deal with DDD. Now, despite the huge number of Snake's, there is a disperportionate number of Metaknights at the top. The last numbers I saw had Snake as 2nd used and about half as much as Metaknight, yet, only Ally seems to be able to make it into the top. And what of Lucario? There are far less Lucario's than Snakes, but never the less, Lucario did better than Snake at Pound.

You honestly can't say that Metaknight has a porportionate distrubution of top spots with the rest of the cast.

 

[adumbrodeus]

50% is doable, but I think there should be other evidence that suggests overcentralization, because as far as I'm seeing Metaknight doesn't shut down 50% (though he has the advantage), but the fact that other characters shut down those potentially viable characters vs Metaknight, only to be murdered by MK himself raises a red flag for me.

Not necessarily, first required piece bit that we NEED is to redo how we understand MUs.

So if I understand correctly, you want a certain number of best players (2-3 players)'s tournament performance added up and averaged out?

Actually finding a certain amount of players as close as we can get to the top of the metagame, randomly sampling them, and averaging them out.




I'm sorru, I forgot who actually said it, but I've always suggested that most MU ratios are BS. Many assume that players won't "play gay", many don't take into account things like "mindgames potential" (yes, totally measurable, I made a thread about calculating that for various characters in various situations).


The core problem is that MUs need to be mathmatically defined as efficiently as possible, and that requires a concerted and quite frankly is very difficult.

Good current MU ratios are estimates, bad ones are pure shots in the dark.

 

[popsofctown]

Delvro is smartt

 

[Master Raven]

lol what is this math class?

 

[Chuee]

Then I guaran****tee you there would be more Peaches, Lucas', and Snakes out there. Hell maybe even more Marths. That's one small example of MK's absence allowing those who were previously on the borderline of being viable to be well...viable (not really including Lucas overall though).

*cough* it isn't a bad thing either.

Getting rid of mk won't have anything to do with Lucas. The MU is 60-40 and even without mk Lucas still gets ***** by snake.

 

[popsofctown]

People who don't understand the value of academia work for people who do.