Metagame Statistical Analysis Project

[DeLux]

Alright, this is mostly for John12346, Tuen, Guest, and Myself, but if anyone wants to join in on this venture, feel free to volunteer. I would stress though that Yikarur and the frame data project are the primary objective of the organization until completion.

Brief: I am going to spearhead a project where we begin to interpret the metagame from a results standpoint.

Mission: Establish an infrastructure of information to dissect the metagame in a way that has not been done before. By the time the project is up and running, we want to be able to produce a purely results based matchup chart and a results based power ranking of characters with statistical insight provided on both.

How: Matchslips from TO's to provide raw data mining. From there developing the matchup chart is very easy. Developing the formula for ranking characters can be determined on what we deem fit. I'm debating between RPI or one of the BCS computer system ranking.


Progress:
- Petitioned URC for voluntary matchslip recording program

To Do:
- If URC is unwilling, decide if we offer BBR a chance to step in as data mining team or open to general public
- Determine how to separate Ruleset Specific Data, Regional Data, Chronological Order Data
- Input data into usable terms
- Maintain chart for all of eternity

Estimated Launch Date of Data Collection: January 9 (post MK Ban)
Estimated Launch Date of Match-Up Chart / PR Tier list

I'll guide discussion later when I get the chance to update it and PM the players

 

[Dr. Tuen]

Oh man, this is rad.

If you haven't already, check this out. It's a player ranking system I researched and worked a while back. It's real old now and it never had the last two ideas worked into it (mostly for player drift reduction), but I'm posting it to kind of pitch my knowledge with ranking systems.

That effort worked with wins and losses only (no set count). If we get detailed match slip data, the possibilities are endless!

As for statistical analysis, that's pretty easy. I'm pretty well versed with parametric and non-parametric analysis methods. If I can be involved in determining what data is mined, you guys won't have to hear from me again until you have the data mined and summarized. The first meeting is important though. If the wrong data is collected, or something is missing, there's no going back.

Anyways. This sounds like a great final project to be on. Well, final for me that is. I'm happy to lend my assistance!

 

[Sieguest]

Cool beans, now I'll have something to fill for the latter half of my Winter Break to keep me busy once the novelty of being free fades lol.

 

[John12346]

So... what? Do I just keep doing what I'm doing?

>___>;

 

[DeLux]

Yeah, the money and usage is going to factor in somehow for the tier list/PR aspect of the project imo

I would almost be tempted to tell you that you could stop counting usage, but then I realized that would only work if we got matchslips for every tournament. I'm going to assume that we're going to get matchslips on a minority of tournaments.


What normalized Data did we want to collect in matchslips?
It can't be anything too extensive because TO's might have to input the data

I was thinking something like:
Date
Event/Tournament
Player 1 Name
Player 2 Name
Game 1 Stage, P1 Character, P2 Character, Game 1 Winner
Stage Bans
Game 2 Stage, P1 Character, P2 Character, Game 2 Winner
Game 3 Stage, P1 Character, P2 Character, Game 3 Winner
Game 4 Stage, P1 Character, P2 Character, Game 4 Winner
Game 5 Stage, P1 Character, P2 Character, Game 5 Winner
Any games went to time?

 

[rPSIvysaur]

include time left on clock maybe

 

[Dr. Tuen]

I was thinking something like:

Player 1 Name
Player 2 Name
Game 1 Stage, P1 Character, P2 Character, Game 1 Winner
Stage Bans
Game 2 Stage, P1 Character, P2 Character, Game 2 Winner
Game 3 Stage, P1 Character, P2 Character, Game 3 Winner
Game 4 Stage, P1 Character, P2 Character, Game 4 Winner
Game 5 Stage, P1 Character, P2 Character, Game 5 Winner
Any games went to time?

For each event, the results and money awarded would be awesome as well.

 

[DeLux]

Since the matchslips would be coming for each tournament, I would think that we could just cross reference info from the results threads?

 

[Dr. Tuen]

Since the matchslips would be coming for each tournament, I would think that we could just cross reference info from the results threads?

We can, just be sure to have it posted. I've sifted through those before, and you don't always get pay breakdowns or a placement order past top 4. It'd be nice to have that pay breakdown and get top 8.

 

[DeLux]

Update:

So far, the URC has been pretty non-responsive if they want to do partake in this or not.

If people are unopposed, I'd like to add Marc on to the project. In doing so, we could partner the project with the BBR and see if that group is willing to help us mine data. Any objections?

I'll also be making a public announcement for non URC/BBR/Smash Lab related TO's for help.

I'll also draft a normalized matchslip

 

[Stealth Raptor]

i am not unopposed to adding marc. would be interesting to see what results this turns up

 

[DeLux]

i am not unopposed to adding marc. would be interesting to see what results this turns up

What opposition (yeah I get the double negative) do you have with Marc being involved?

 

[Stealth Raptor]

ummmmmmmmmmmmmmmmmmmmmmm













im not opposed LOL sorry, i guess i did double negative it