Need in help in developing an algorithm for the 2014 college football season

[WNY_FOOTBALL_DUDE]

I have decided to come up with an algorithm/objective formula/RPI rating for the 2014 College football season, and I would like to know your suggestions on how I can spice it up. I want to do it before the season starts, so there's no tweeting of reality to fit an agenda.

Here's what I have come up with so far:

(Winning% + Opponents Winning% + Opponents of Opponents Winning%)/3
FCS opponents receive a flat .167 rating (2/12)

Is there anything more I should add to the formula?

 

[occupant]

Goat milk.

Babies will develop strong resistance to allergies with goat milk supplemented by acidophilus and cod liver oil.

When you're making a formula, YOU get to make the rules. Personally I would add HOME/AWAY variable.

 

[Codaxx]

the real question is how complicated you want to make the formula. What values you want to assign for home and away wins. Basketball is at .60 and 1.4 and baseball at .70 and 1.30. If it is just a simple RPI rating than that is pretty much it. Beyond that you can use winning margin, actual vs expected results, etc .

 

[WNY_FOOTBALL_DUDE]

the real question is how complicated you want to make the formula.

I generally think it is best to keep it simple, so people can understand it and there isn't any "unfair" variables.

MOV is mostly a deceptive stat, because it doesn't factor in how good or how bad the opponent is, and discriminates against teams which play grind it out football or consistently play tough opponents every week.

I am considering weighing home and away games differently.

 

[HuskerinBig10]

Goat milk.


When you're making a formula, YOU get to make the rules. Personally I would add HOME/AWAY variable.

I like that idea of home/away.

If you play nine or more home games = -0.25

If you play eleven or more games in your state(including bowl game but not conference championship) = -0.25

I

 

[Jack_John_Mark]

I'm making one as well.

The problem is (as I've found in the past) once you get started you begin to realize there are hundreds of key variables needed.

 

[ckhokie]

You're weighting the winning % of the team in question the same as the winning % of their opponents opponents?

 

[occupant]

I'm making one as well.

The problem is (as I've found in the past) once you get started you begin to realize there are hundreds of key variables needed.

Yep.

Don't do anything in Access, stick with Excel...unless you're a SQL wizard of course. Trust me.

 

[Codaxx]

I generally think it is best to keep it simple, so people can understand it and there isn't any "unfair" variables.

MOV is mostly a deceptive stat, because it doesn't factor in how good or how bad the opponent is, and discriminates against teams which play grind it out football or consistently play tough opponents every week.

I am considering weighing home and away games differently.

Home and away is different in all the currently used RPI ratings. I like the .60 and the 1.4. Commonly accepted that home field is worth about a FG, and that equates pretty closely to 60 (I think it 57, but 60 is close enough).

MOV can all be adjusted that is why I said it gets very complicated.

 

[Jack_John_Mark]

Yep.

Don't do anything in Access, stick with Excel...unless you're a SQL wizard of course. Trust me.

Yea I used SQL last time. I'll do that again this time as I just have way too much data to process and it takes quite a bit of code to do it.

 

[Red_Alert]

I have decided to come up with an algorithm/objective formula/RPI rating for the 2014 College football season, and I would like to know your suggestions on how I can spice it up. I want to do it before the season starts, so there's no tweeting of reality to fit an agenda.

Here's what I have come up with so far:

(Winning% + Opponents Winning% + Opponents of Opponents Winning% + SEC (+1) + non SEC (-1) /3
FCS opponents receive a flat .167 rating (2/12)

Is there anything more I should add to the formula?

fify

 

[HammerDown]

Interesting, I love this sort of stuff but you're telling me that no one has already done this?

 

[Codaxx]

Interesting, I love this sort of stuff but you're telling me that no one has already done this?

I believe their was a similar formula in the original BCS formula. BCS has steadily decreased the influence of quantitative measurements over the years.

 

[Jack_John_Mark]

Interesting, I love this sort of stuff but you're telling me that no one has already done this?

Oh there are hundreds of us idiots out there trying it.

Not a lot of people have the programming smarts to make a program that can handle enough useful data though. Takes thousands of lines of coding.

 

[Codaxx]

Oh there are hundreds of us idiots out there trying it.

Not a lot of people have the programming smarts to make a program that can handle enough useful data though. Takes thousands of lines of coding.

It is getting more and more interesting. Unfortunately college football is one of the hardest to use quantitative results for. Lack of data and similar opponents. Football Study Hall/Football Outsiders do a lot of good work on this stuff, if you want to understand it but are too lazy to try to put it all together.

 

[WNY_FOOTBALL_DUDE]

You're weighting the winning % of the team in question the same as the winning % of their opponents opponents?

Yes. 1/3 winning%, 1/3 opponents winning%, and 1/3 opponents of opponents winning%.

 

[RegentDenali]

Chris Petersen = Wins * Infinity

 

[4down20]

Anything remotely based on win% is horrible.

 

[WizardHawk]

Anything remotely based on win% is horrible.

Ok, so he asked for suggestions. Surely you can do better than just saying what he has so far is shit right?

 

[4down20]

Ok, so he asked for suggestions. Surely you can do better than just saying what he has so far is shit right?

Not really. If I knew some great simple end all formula like that, I'd already be using it. And so would all these people: College Football Ranking Composite

There is no real formula like he wants that works because there isn't enough games in a football season for one to work properly. It works in basketball because there are so many games to even things out.

I've mentioned the problems with win% in the past. It sees a 10-2 mac team as being equal to a 10-2 SEC/Pac12 team. Cumulative win% helps that to some degree, but not really.

Not being a dick, but if he wants to create his own formula then this is what it's about - understanding the flaws in these things. I think my first idea was similar to this.