All-Star Team Ranking System

[BlueCat]

For the math geeks out there, I am trying to come up with an objective-ish way to rank the senior level 5 teams over the last 10 years or so using primarily their results at Worlds. This is BY TEAM, not necessarily by program.

My main goal is to come up with a group of about 10-20 or so teams in the first "tier", then perhaps a group of 10-20 or so that would make up the next "tier". I would like to weight more recent results a bit more than ones in the past, but not over-do that. Those groupings are more important to me than the actual ranking within those groups, if that makes sense.

Thoughts?

Welcome to Google Docs

 

[Ren]

Sounds fun, I'm just about to get on the tube but I'll have a look. What about teams which have changed divisions? Treat them as a new team once they change division? What if they go back again? New team again or pick up from their previous results in that division x amount of years ago? Just a couple of things off the top of my head.

 

[BlueCat]

Teams switching divisions is tough. Should priority go to team names or divisions?

It gets even tougher when the same gym wins the same division multiple times with different team names. In CA's case, we have won Small Senior with both Jags and Panthers, International Coed with both Pumas and Wildcats, Large Coed with both Wildcats and Cheetahs, and Large Senior with Supercats and Panthers.

 

[12stepCheermom]

Maybe if you stick with divisions that will work itself out. Even if a team won but was only in that division once, over a team that consistently places top 5 over multiple years the math would rank them properly?

I think....


The Fierce Board App! || iPhone || Android || Upgrade Your Account!

 

[King]

For the math geeks out there, I am trying to come up with an objective-ish way to rank the senior level 5 teams over the last 10 years or so using primarily their results at Worlds. This is BY TEAM, not necessarily by program.

My main goal is to come up with a group of about 10-20 or so teams in the first "tier", then perhaps a group of 10-20 or so that would make up the next "tier". I would like to weight more recent results a bit more than ones in the past, but not over-do that. Those groupings are more important to me than the actual ranking within those groups, if that makes sense.

Thoughts?

Welcome to Google Docs

What you want is exponential smoothing with a weighted moving average. That isn't exactly easy unless someone programs it to take it into account.

You also want to figure out how to classify a team. The best way would be to do it by difficulty and deduction scores. I think winning doesn't really matter for what you are looking for and would skew the data.

 

[mstealtoyou]

Smart people thread alert.........lol.......but I can't wait to see what y'all come up with.

 

[BlueCat]

What you want is exponential smoothing with a weighted moving average. That isn't exactly easy unless someone programs it to take it into account.

You also want to figure out how to classify a team. The best way would be to do it by difficulty and deduction scores. I think winning doesn't really matter for what you are looking for and would skew the data.

Agree, except that we are dealing with limited information. There were years when USASF wouldn't even release the final scores, if you can imagine.

 

[King]

Agree, except that we are dealing with limited information. There were years when USASF wouldn't even release the final scores, if you can imagine.

What if we said 'screw it' to all past results and created a system for moving forward? I think if we throw of the baggage of trying to quantify now it might be a more liberating experience and come up with better answers.

 

[BlueCat]

Also, even the USASF scoring we do know doesn't translate well to mathematical analysis for several reasons. It is designed to produce a ranking, but not much else. ELO-chess or similar might work, but that is beyond my programming pay grade.

 

[BlueCat]

My main goal would be to simply identify different echelons of teams, with groups of about 15-20. That may be simpler than accurately ranking every team individually.

Example: 15 "Ultra-Elite" teams, 15 "Elite" teams, etc.

 

[xomckenzieee]

I think using Worlds alone won't produce the best results. When businesses are ranked and grouped based on what they produce, it's over a period of time and based on different scales. I think instead of going by Worlds alone, we add in large nationals as well. And it shouldn't be done by who won but more by a comparison to the other teams since the Worlds scoresheet is based on comparison (although people will argue otherwise). It should also be divisions, not team names, aka CA Small Senior is ranked this no matter what team what year. There's different statistical formulas we can use, we would need to determine how we wanted to compare and with what information. Getting all that info will probably be the most difficult part.


The Fierce Board App! || iPhone || Android || Upgrade Your Account!

 

[CEAlover08]

If it is just using worlds, it is going to be a basic statistics formula. If you add in nationals, it might make it a little bit more difficult but not by much. As long as you are only doing it by rank, it would probably be a pretty easy task after you get all the results. If you want help, I can try to figure out the formula and stuff.

 

[cheercurl]

I am pretty sure this is why I did not take Statistics in college...

 

[King]

@ashley is pretty good. I can design a database and how it should flow, but code monkeying is not my strong suite.

 

[Ashley]

@ashley is pretty good. I can design a database and how it should flow, but code monkeying is not my strong suite.

This isn't really spinning up a quick web application which is more my forté. This is something that could be done in excel if someone could come up with a formula for it.

In either case, the code monkeying I could handle. The math part of it, not so much.