The North Carolina Brawl Power Rankings! (Update 11/08!)

[DJRome]

I have a question for the panel. I'm currently working on an independent mathematical power ranking model.

1. How should absences be weighted compared to placing low? Which is more detrimental to the ranking?

2. Is an attendance requirement adequate to account for attendance in the ranking?

3. Are multiple sets in one tournament equal to value to the same number of sets on separate days?

4. Do the marginal value of sets versus one person stay equal or do they diminish? Or even increase?

5. What is a good value to account for past rankings? 10%? 20%? 40%?

6. Does the weight of a tournament depend only on the top half of attendance or the total number of attendees?

7. How do out of state attendees figure into the equation?

Thanks for your time.

 

[Dr Peepee]

I think all 3 panelists would answer these differently, sometimes even oppositely.

Just do what you feel is best.


PS: If you want a better answer, you should PM the 3 of us(I'd be cool with being AIM'd.

 

[DJRome]

whose the third panelist? u karn &?

 

[Dr Peepee]

Billbillbill

 

[DJRome]

aight, i guess i'll aim u next time ur on

 

[lord karn]

1. Absences shouldn't matter. The program should look at all the set counts we have available between players in NC within the ranking period. If people are absent, then they won't have much data and won't be able to be ranked high.

2. See above. Obviously, one tournament is necessary, but you'll probably have to be at more than one to get enough data to be placed accurately.

3. I don't really know about this. I would say that sets on the same day should be weighted less. I can see arguments for the other side, though, so I don't really know.

4. Hmmmm. I think they should decrease, if I'm understanding the question right. If one player beats the player who is probably going to be ranked #3 4 times but loses to a lot of other people, the data only shows that said player beats player #3. Basically I'm saying that someone can't go beyond a certain point just because they beat one person over and over again. It's most important to see how well people do against as many other people as possible, and more set counts only make the results between players more accurate.

5. I don't think we should look at them at all if we have enough data.

6. Tournies shouldn't be weighted. A bigger tourney will just simply yield more data because more players will play more players. Again, I can see the other side of the argument, though. Higher stakes, etc. Hmmmmmm.

7. The only way to do this without using the same system to rank other states would be to assign a value to out of state players based off of how good we know they are. Or not count them. But I'd go with the former.

 

[DJRome]

Actually, what I mean by #4 is that how much should a second set of 1 person vs another be. Say is the second set as valuable, more valuable, or less valuable than the first set result (you can substitute in indicative for valuable if you wish).

 

[THO]

Yo DJRome, I was working on a statistical ranking system a few months ago but quit when I realized I don't give a ****. But yeah man, if your doing it what is your game plan.
I'm all for this **** and I'll help you out however I can.

When I was doing it I was trying to base it off of the NCAA Mens basketball ranking system.
They make your match ups worth 25% of your score, your opponents match ups worth 50% of your score, and your opponent's opponents match ups worth the remaining 25% of your score. This way you get a solid representation of the scene, with heavy weight on your strength of competition.

They also include various data such as home or away games and how close the matches were, the later would be really good to include but would require even more data from each tournament (like if each match went to last stock or something).

I was super exited because we could pretty much rank every player in the state once we get a program running and that would be tight. People would really begin to appreciate how important each match really is, especially for those who aren't in the top 10.

But yeah, you can do it however you want, I just wanted to let you know I'm interested.

 

[null55]

lol soon we're going to get into stock count.

 

[KillerSOS]

Sounds hard THO.

my eyes would bleed numbers

 

[NC-Echo]

Sorry bros, I'll try to get us all together to talk about this when I get off of work tonight.

 

[DJRome]

yeah karn tole me about ur attempts. i'm gonna run the numbers through matlab. i think it's the best way to jam through the number crunching.

i don't think i'll place much value beyond set counts though. and i'd like to place a slight emphasis past rankings.

also, i think tournaments should be weighted because i believe in looking at placements as well as set counts, and placements would be kind of meaningless without said weightings.

another points i'm looking at is the importance of absence. i'm kind of struggling to place it in anything. from the previous data of the circuits other states have run, attendance is important, so i feel like it should be too.

 

[Duo55]

yeah karn tole me about ur attempts. i'm gonna run the numbers through matlab. i think it's the best way to jam through the number crunching.

i don't think i'll place much value beyond set counts though. and i'd like to place a slight emphasis past rankings.

also, i think tournaments should be weighted because i believe in looking at placements as well as set counts, and placements would be kind of meaningless without said weightings.

another points i'm looking at is the importance of absence. i'm kind of struggling to place it in anything. from the previous data of the circuits other states have run, attendance is important, so i feel like it should be too.

I agree, you can't keep your spot if your not there to defend it.

 

[Dr Peepee]

You people make this all too hard for a simple process.

Explain to me how 25/50/25% is the best you should assign vs 26/48/26%? How do you know? Arbitrary values, combined with several UNNECESSARY variables you want to include(which we don't have to worry about right now because A. we update frequently(ex: age of tournies within ranking period) or B. it just isn't important unless there's a tie that needs to be broken(ex: old PRs).

Why are you so dead-set on this anyway, DJ? In light of what I just said, the best thing your mathematical chart will do will be to serve as a TOOL for AIDING the decision process. It should NOT replace the process generally carried out by panelists though. I'm basically wondering if the staggering work you would put into this design is worthwhile.

 

[NC-Echo]

You people make this all too hard for a simple process.

Explain to me how 25/50/25% is the best you should assign vs 26/48/26%? How do you know? Arbitrary values, combined with several UNNECESSARY variables you want to include(which we don't have to worry about right now because A. we update frequently(ex: age of tournies within ranking period) or B. it just isn't important unless there's a tie that needs to be broken(ex: old PRs).

Why are you so dead-set on this anyway, DJ? In light of what I just said, the best thing your mathematical chart will do will be to serve as a TOOL for AIDING the decision process. It should NOT replace the process generally carried out by panelists though. I'm basically wondering if the staggering work you would put into this design is worthwhile.

Although I approve of the attempt and I hope it works out to a certain extent pp is right.

 

[zCasanova]

PR lists based off of number crunches suck.

 

[THO]

YO PP,
The 25/50/25 split is simply what the NCAA uses for college ranking. The numbers seem solid because they have been time tested to quite accurately predict the outcomes of the NCAAs year after year. Obviously they were initially arbitrary and over time the system became fine tuned into what it is today.

Still though, i'm not sure if using 26/49/25 or whatever would make much of a difference. The point is simply that this is the system that the professionals use and i don't see any reason in changing what they do, rather I see great reason to emulate them (and if there are problems, we can fix it).

Also what you said about the TOOL FOR AIDING aspect of these numbers. Yes, this data would be used to aid the PR. Right now you guys are going off of arbitrary mental perspectives of who is actually good, there is no backbone to the PR.

There are many factors that come into making the PR and this should be one of them, in fact I don't think matchups should be looked at in any other way (except for the strange case when, say, for example, Omni just so happens to play Plur first round in every Brawl tourney and then also Rag in losers, and for four tourneys in a row he just gets ****ed by random bad luck (he can't beat those two people and only plays them for some odd reason)).

The only times when the panel is important is when outcomes are really close, data isn't so bountiful, major johns are being called into play (like my mom died right before that match so maybe don't count it) some **** like that.

Also, the reason for the 25/50/25 split is so that even if you don't play every body, maybe those people you did play, played every body, so by taking into account not only your matches, but your opponents matches, and even further your opponents' opponents matches you can actually get a pretty good perspective on the overall result of a single person.

I don't really know what data you guys are using now, or even how you are determining rankings but I believe that the community should have a solid understanding of where this all is coming from and not just be like oh look i'm number 7, woop.

There needs to be a little scientific method to all this, thats all i'm sayin,
but wait...I don't even play brawl, **** it

 

[Dr Peepee]

1. NCAA is not equal to smash. It may work for their teams, but there are more teams variables(which makes things even less applicable to our SINGLES RANKGINS) and it's just a different sport entirely. Also, how are the numbers STILL not arbitrary? do you even know how they decided upon them?

2. OFSKMJVJOJVOAJOIVSAV

FOR THE LOVE OF EVERYTHING OBJECTIVE, THE PANEL IS NOT NOT NOT NOT

NOT NOT NOT NOT NOT

NOT NOT NOT

NOT NOOOOOOOT

Making up perspectives about who is good. Arbitrarily defined numbers vs head to head performances/tournament placings? I don't see how you think numbers are better(imo numbers might actually have a case to beat out set counts now that I think about it, but they will NEVER beat tournament placings).

3. That is an extreme case as well as highly unlikely. not sure why you used it.

4. apparently the panel IS needed.

5. Tournament placings do the same thing(not sure if there's any even slightly objective way to do it for set counts.....this would probably be fixed by just having longer ranking periods though)

6. The community is apparently going to think what they want to about how these should be done regardless of what I, or any other panelists say. People need to accept their placing and if they don't like it they shouldn't criticize the system or the panelists.

7. there is scientific method to this

 

[null55]

lol dude i love PP.

 

[THO]

Maybe I didn't explain this so good, the numbers 25/50/25 are used because the people who designed the system believe that how good you do against your opponents and how good your opponents do against their opponents should be equally weighted, and that those two combined should be equally weighted with your opponents skill level, hence the 50%.

Furthermore, the idea that the three numbers could be 10/90/0 is completely possible, also 5/2/93 as well. The reason 25/50/25 has stuck is because these proportions give good results, aka, seeding based off of pre tourney results often leads to nearly the same result as the system predicted, thus the 25/50/25 seems to make sense.

Applying a ranking system used by Basketball to smash obviously has differences, i'm not saying that, i'm just saying their strategy makes a lot of sense and how they weight things as well.

But PP, woah woah, I didn't mean to sound so harsh, but looking back it seems as if i was. I wasn't trying to say you guys just make **** up, you obviously use your minds to come up with this stuff. I shouldn't have said that the way I did.

All I mean was that if I were being ranked I would like to know why I was in a certain spot. I was looking at the florida brawl PR and they have pretty legit explanations for each placing in top 10. This was very helpful and gave me great insight into how they did the rankings. I looked on our first page and didn't really see that. Of course I could be wrong and you could have posted that later in the thread and I didn't have time to look through 100 odd posts, so if you did explain i'm sorry.

Now I gottsa talk about 5, I'm not sure what you mean by tournament placings. I was under the impression that placings shouldn't matter at all, rather only match ups should matter.
This can clearly be seen in the following example:
If randomly you placed some scrub as top seed at a tourney where say 4 other PR people showed up, then that person wouldn't play any of those people until top four. Ensuring at least fourth for some one who didn't beat a single person who had any true skill. Also, coming from the other side of the bracket, one of the other PR people was eliminated before they even got to that round and so this scrub places higher than them. It just seems a bit silly that being high on the PR can be so self perpetuating as that. I just don't think placing matters at all.

But, yes I agree people shouldn't criticize the PR once they are made, that is just silly, rather they should recommend ways to improve them so next time they are as accurate as possible.

Also, I'm sure there is scientific method to this all, I was just being silly, sorry about that.

 

[lord karn]

David you should come over and play melee.

 

[Foxy]

NC is terrible at Brawl, we're making our rankings seem like a bigger deal than they are.

Really.

They're going to be pretty random and change a lot, we don't have that many consistently good players.

 

[Dr Peepee]

That post is much better. =)

You would still have to determine the best numbers for smash, and, until you find these "magical numbers" (which I would still believe to be subject to change, especially over the course of a ranking period, that it is literally impossible to even be close), then it would just be another frustrating arbitrary system.

I wouldn't mind explaining it, but every time I have explained a placement, there have been ignorant responses(from most people that voice their mind) that make me really against this idea. Otherwise I would be up for doing explanations.

Well, this is because you have talked to Karn about PRs only, I guess. As usual, a nearly impossible scenario is used to back this up. Maybe there are skilled players that aren't PR'd that attend tournaments? This has been true at EVERY Brawl tournament NC has had, so the "scrub placing well" theory is dead as far as this state goes, I think.

I don't think many suggestions are necessary, as the system is about as good as it gets(until we extend the ranking period for a slightly longer amount of time), but I am willing to admit that a mathematical model would, at the very least, be something interesting to see.




@Foxy's last point: This is why I want to extend the ranking period. Maybe results will become more consistent over time.

 

[lord karn]

I don't see how the scrub placing well thing doesn't work. . . There are lots of examples that show faults in the system.

The most obvious is the scrub one where an unknown player comes to a tourney and is seeded really poorly because no one knows who he is. He beats like 3-4 good players and ends up getting like 4th or 5th, but only loses to the top two players. However, the player who ends up placing 3rd is a previously ranked players and so gets a good seed. He only beats, for example, two good players, because he has a high seed. For the sake of the argument (although it still works otherwise), let's say that the unknown player beats all of the same people that the previously ranked player beat, but also beat like one other person. Now we have an unknown who places worse than a ranked player, even though he beat the same good players as well as others. Yet if we do power rankings based off of tournament placings, the guy who was previously ranked will continue to be ranked, because he continues to get good seeds. The unknown player will always have to fight more good people, and in fact might even beat more good people, but place worse. How can you justify ranking the unknown lower than the previously ranked player even though he beats more people?

Now situations like that happen all the time to some extent, so saying it will never happen does not throw this situation out the window. Set counts look at the same data that tournament placings do, just without the random or biased data that seeding puts into the equation.

Now, there are still things that we would have to decide on as a panel to weight. Even when we have the data from set counts, which is simply superior to placings data, we still have to decide if we want weight more recent tournaments more, if we want to penalize players for attendance, plus a few other things. However, this is not a question about data, which is pretty clear cut, but a question of what we are looking for in the data we have. I've always analyzed that data with the assumption that we are making a list of how our players do vs one another, essentially how well each of them would be expected to place in a tournament with no seeding/random bias (IE a complete round robin). This may or may not be what we are going for, or maybe I sometimes make mistakes trying to analyze the data based off of my goal, but I really can't comprehend why you would ever want to use placings over set counts. Unless, of course, you want the rankings to be random and biased.

 

[Dr Peepee]

I don't see the point of arguing this with you.

set counts, which is simply superior

No, clearly placings is superior.

 

[stingers]

I like to simplify things

Just look at who beats who
organize it
and

When there's something weird going on (say Shady beat Keith who beat Duo who beat Shady) then just look for consistency (Did Keith lose to a random who placed 33rd?) as a secondary factor.

Which is why NC needs to ****ing upload brackets instead of just posting placings x.x. Placings don't matter as much as the bracket does.

imo

 

[lord karn]

I don't really expect you to. :/

 

[Dr Peepee]

I don't really expect you to. :/

Then why post what you did?

 

[lord karn]

Then why post what you did?

To provide the light for all who want to see. -_-

 

[Dr Peepee]

Looooooooool

 

[-ACE-]

Good post Karn.

I agree with stingers for the most part.

 

[OmniOstrich]

Are having high seeds that important to everyone? Alex always has a high seed and still loses round 2-3 and has to suck it up through loser's bracket just like he didnt have a good seed at all.

If you dont care about your seed, I dont see why the PRs need to be as super accurate as possible, they don't do anything else other than stroke your ego and I suppose give OOS a way to find out who our best is but really...NC brawl is not anywhere near comparable to NC melee...most OOS do not care about our brawl players =X

 

[Dr Peepee]

How about the competitive drive it gives our Brawl players which will increase our chances of being noticeable one day?

 

[ZIO]

I was gonna post "loooooooooooool" but then I realized that could give me infractions. But being that this post has no point yet, it could still get it.

Hmm . . .

PRs will always be terribly flawed to the individual. It's why a panel is chosen (in which manner, I am unsure) and have them sort things out. And for points, I like the idea.

 

[*CT*]

I say the panel will create the best motivations to make us play more.

Seeding I care not for.

 

[ph00tbag]

I don't see how the 25/50/25 numbers shouldn't work. The only variable they consider is outcomes. They don't consider any venue-specific variables. In any competitive venue that pits one unit against another, outcome is a universal variable, usually boolean. If I gave you just two numbered list of units with their performance against every other unit within the respective lists, and asked you to provide a statistical analysis, you'd analyze both lists the same way. Why would you analyze those lists differently if I told you the two lists came from different venues?

After all, that's what this tool would be used for: ranking units based entirely on data without consideration for mitigating factors. That's what the NCAA uses it for, then they consider variables that are specific to the game and the environment. 25/50/25 needn't change from venue to venue because it is inherently valueless. The values considerations come from the Rankings Panel, the math just gives them something to apply their values to.

Also, match-ups matter more than placings.

I don't even care, really, it's just a neat philosophical question.

 

[Pyrostormer]

Lol, NC PRs.

We r doin it rong obv

 

[TheLastCacely]

NC is terrible at Brawl, we're making our rankings seem like a bigger deal than they are.

Really.

They're going to be pretty random and change a lot, we don't have that many consistently good players.

speak for yourself mk user

 

[TheLastCacely]

How about the competitive drive it gives our Brawl players which will increase our chances of being noticeable one day?

this man makes sense, omni is dumb, we got people like rag( who is in dk backroom, and is probably better than cable) duo, who is probably the best fox user... unless anyone knows of another, Keith, who is the sexiest black man, pp, who could probably go toe to toe with neo, shady, who is a ****in terrorist, mits, who's wolf is probably better than jj's, and so on, and so on.

edit: stingerz, who beat dabuz and took a game off of adhd

 

[stingers]

jumpman is the best wolf @.@