Ive got a great idea on how to determine matchups!

[Sneak Sneaks]

This is the idea! Lets make 2 tourneys of one character only (depending of the character subforum), one for 3ds and one for Wii U, and the players that ended on the first 7 places of the Wii U tourney with the players that ended on the first 3 places of the 3ds tourney will challenge another character's 10 best team (each against the according place that they ended in the same character tourney) and according how much winners over losers we have on the challenge, will be the matchup's percentage! What do you think?

 

[nodle]

Well, we would effectively end up with only 10 data points per matchup. That puts a huge amount of uncertainty into our results. Our (95%) margin of error would be something like +-30%. If we scaled up your proposal to 100 man teams, then we achieve a (95%) margin of error of about +-10%. This would mean 100 games per match-up with 55 choose 2 matchups. We would need to manage and record the results of 148,500 matches. Not impossible, but pretty in-feasible. Even if we were content with the 30% margins on 10 person teams, we would still need to manage and record the results of 14,850 matches. While I think it would be great to build teir list based on actual character results, the logistics of it seem impossible. Its very easy to underestimate the amount of data we would really need to make good claims.

 

[Sneak Sneaks]

But it would be more accurate than just taking tournament results... why? Because not all characters have the correct representation on tourneys.

 

[nodle]

It would be more accurate. But we would lose all precision. With 10 person teams, half our matchup ratings would be off by more than 10%

10 man teams would also still be 14,850 matches to organize and record the results for.

 

[Sneak Sneaks]

It would be more accurate. But we would lose all precision. With 10 person teams, half our matchup ratings would be off by more than 10%

10 man teams would also still be 14,850 matches to organize and record the results for.

2 subforums being on the challenge can work together for both of their matchups

 

[Xermo]

But it would be more accurate than just taking tournament results... why? Because not all characters have the correct representation on tourneys.

Not everyone is on the same skill level. Taking a bunch of randoms from character subforums and throwing them into a tourney doesn't say "X character has a 60:40 matchup against Y character because more players of X won over more players of Y." There's a number of reasons why Y could have lost.

Character representation in tournaments aren't the sole factor in what decides a character matchup. It's all theory and data study based on the most optimal a character can perform against another at top level play. The Rosalina/Ness MU for example isn't decided because good Rosalinas place more in tourneys than Ness players, it's due to Rosa literally possessing tools that shut down Ness hard.

Also, do pray tell how would the top from 3ds and Wii converge with one another in a 10-person team if they have separate games.

 

[RayNoire]

Some characters don't even have 10 competitive players (yo), while I'm pretty sure there are at least 10 Sheiks in the top 50 players alone.

 

[Sneak Sneaks]

Not everyone is on the same skill level. Taking a bunch of randoms from character subforums and throwing them into a tourney doesn't say "X character has a 60:40 matchup against Y character because more players of X won over more players of Y." There's a number of reasons why Y could have lost.

Character representation in tournaments aren't the sole factor in what decides a character matchup. It's all theory and data study based on the most optimal a character can perform against another at top level play. The Rosalina/Ness MU for example isn't decided because good Rosalinas place more in tourneys than Ness players, it's due to Rosa literally possessing tools that shut down Ness hard.

Also, do pray tell how would the top from 3ds and Wii converge with one another in a 10-person team if they have separate games.

Random? I literally explained that there must be a ditto tournament first to determine which arethe 10 best

 

[Xermo]

Random? I literally explained that there must be a ditto tournament first to determine which arethe 10 best

"Randoms", as in random players from a character board, not random characters. I understood what you meant.
Even then, character dittos don't determine who's the best of the character.

 

[Wintropy]

Xermo said it best. In addition to that, there's the fact that it doesn't consider if the player just isn't playing very well that day. You get players that, fundamentally, don't match up to other players, just that the other players happened to SD at an inopportune moment and forfeit the match.

I mean, I play Pit. I think I'm pretty good with Pit. I can get reasonable results with Pit. Yet if I played Nairo, a player who doesn't even main Pit? I'd get bopped. That's okay, it's a good reflection of our respective skillsets. Consider, on the other hand, if I somehow beat Nairo in Grand Finals of the Pit tournament: the winner of the tournament is therefore an inferior player. The inferior skillset is reflected in the matchup charts, which makes for weird results.

 

[nodle]

In an ideal world, we would have access to the ForGlory data. We could then subset the data and look only at games in which both players have winning records. We could then break up that subset into character matchups and determine, for each matchup, what the bias was like. (i.e. Study the ForGlory data to see how likely it is for a good DK to beat a good Ness.)

 

[Sneak Sneaks]

Ok I get it -_- even though... why dont we implenent a league or something? And this wouldnt determine the matchup completely, just give some info and many replays for future reference. It would be nice having subforums wars.

 

[NotAnAdmin]

I like the idea of have a subforum war, but it wouldn't accurately tell what the match-up is.
There are many pieces of "RNG" that go into a human vs human match of Smash bros.

Skill level is probably one of the biggest factors.
Then you just have the simple fact that you're just not playing up to par that day, or something is on your mind. It would help us learn match-ups we don't see much, but it won't set anything in stone.

 

[Sneak Sneaks]

I like the idea of have a subforum war, but it wouldn't accurately tell what the match-up is.
There are many pieces of "RNG" that go into a human vs human match of Smash bros.

Skill level is probably one of the biggest factors.
Then you just have the simple fact that you're just not playing up to par that day, or something is on your mind. It would help us learn match-ups we don't see much, but it won't set anything in stone.

Ok, for the skill issue: dependending on how you placed on the tourney you wll fight the opponent that place the same place in the opposite tourney

 

[nodle]

Even if the mirror matchup tournament system worked and did correctly identify who the most skilled players were. and got us the 10 best players for each character, and the matchup system took player skill out of the system, and ...


The system would still be at the mercy of basic statistics. 10 data points just isn't enough to make good claims. Let's say we follow your procedure and find that the kirby-DK match-up (arbitrary example) favors DK 8-2. That is to say that DK won in 8 of the matches and lost in 2 of them. Because 10 data points leads to such a high margin of error, the DK-Kirby match-up could actually be as good as 5-5. In fact, we expect that over half of our macthup ratings would be off by more than 10 percent.

What would be better is to record the results of every game and have each pairing play 5 games. This increases our number of data points to 50 per matchup, reducing our (95%) margin of error to 14%. Half our matchup ratings would (in theory) be accurate to within 3.4%. This is of course assuming that our system wasn't introducing any form of bias to the results (which it probably is.)

EDIT: its worth stating that this would end up giving us information that isn't really equivalent to matchup ratings. I acutally think knowing the probability of winning individual games is more important though. It lets you generalize the data easily to different matchup lengths (3 games vs 5 vs 7) while also being more informative for counter picks.You would acutally be able to compute your probability of winning a matchup while taking into account counter picking strategies. (example : Maining Charizard gives me >=50-50 matchups for all characters except 4. My worst matchup is 30-70 against ZZS. ZZS's worst matchup is against villager with 60-40. If I Main charizard with the plan to counterpick villager against ZZS, then my probablity of winning is given by 0.3*0.6+0.3*0.4*0.6+0.7*0.6*0.6=0.504, making my matchup with counterpicking strategy 50-50. All numbers matchup numbers used in this example are hypothetical. It is merely meant to show that knowing the odds of winning individual games is more useful than knowing specific matchups.)

 

[Sneak Sneaks]

Even if the mirror matchup tournament system worked and did correctly identify who the most skilled players were. and got us the 10 best players for each character, and the matchup system took player skill out of the system, and ...


The system would still be at the mercy of basic statistics. 10 data points just isn't enough to make good claims. Let's say we follow your procedure and find that the kirby-DK match-up (arbitrary example) favors DK 8-2. That is to say that DK won in 8 of the matches and lost in 2 of them. Because 10 data points leads to such a high margin of error, the DK-Kirby match-up could actually be as good as 5-5. In fact, we expect that over half of our macthup ratings would be off by more than 10 percent.

What would be better is to record the results of every game and have each pairing play 5 games. This increases our number of data points to 50 per matchup, reducing our (95%) margin of error to 14%. Half our matchup ratings would (in theory) be accurate to within 3.4%. This is of course assuming that our system wasn't introducing any form of bias to the results (which it probably is.)

Anything that gets us in subforums wars!

 

[V4n1sh]

Ok, for the skill issue: dependending on how you placed on the tourney you wll fight the opponent that place the same place in the opposite tourney

Well that wouldn't really work well. Let's take Paragon 2015 Smash 4 Singles as a example. In that tournament, ESAM and False got 13th. ESAM is without a doubt the better player in results (Top 8 EVO and CEO 2015), but they both got 13th. ESAM might've been playing off his game, out of practice, etc.

 

[Sneak Sneaks]

So.. what do you think about the subforum wars?

 

[V4n1sh]

So.. what do you think about the subforum wars?

It wouldn't work well because the same reason NotAnAdmin said, but instead I put a tournament example, it would seem nice though

 

[LightLV]

This is the idea! Lets make 2 tourneys of one character only (depending of the character subforum), one for 3ds and one for Wii U, and the players that ended on the first 7 places of the Wii U tourney with the players that ended on the first 3 places of the 3ds tourney will challenge another character's 10 best team (each against the according place that they ended in the same character tourney) and according how much winners over losers we have on the challenge, will be the matchup's percentage! What do you think?

too many variables to consider and unless it happened consistently, too small a study group.

I don't see the harm in doing it anyway, but attempting to get an objective result out of this would be very, very difficult, if not impossible when you consider things like individual player skill, each player's range of character knowledge, ect.

Results could get skewed very easily and would be misleading without alot of repeats.

 

[Sneak Sneaks]

You know what, I think I will just make a new thread about subforum wars and nothing about matchups, just for fun