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

[WNY_FOOTBALL_DUDE]

I gave you helpful feedback.

But hey, what do I know about the topic right?

Perhaps it's too early in the morning, but you posted a link to college football statistics and made the erroneous claim that a 10-2 MAC team and 10-2 SEC/Pac-12 are viewed equally. They are not. The current formula I am using, is 2/3 SOS.

I am not going to base my model on vegas odds logic.

I still haven't heard any suggestion to spice up my algorithm.

 

[Codaxx]

It has drive data.

in the play.csv, which lists and gives the stats for every play:

"Game Code","Play Number","Period Number","Clock","Offense Team Code","Defense Team Code","Offense Points","Defense Points","Down","Distance","Spot","Play Type","Drive Number","Drive Play"

I will have to check that out.. Thanks.

 

[Codaxx]

Perhaps it's too early in the morning, but you posted a link to college football statistics and made the erroneous claim that a 10-2 MAC team and 10-2 SEC/Pac-12 are viewed equally. They are not. The current formula I am using, is 2/3 SOS.

I am not going to base my model on vegas odds logic.

I still haven't heard any suggestion to spice up my algorithm.

Here is the Colley Matrix explained. It is one of the few that publish their formulas. Probably a bit more robust than you were thinking, but perhaps worth a read to give you some ideas.

 

[4down20]

Perhaps it's too early in the morning, but you posted a link to college football statistics and made the erroneous claim that a 10-2 MAC team and 10-2 SEC/Pac-12 are viewed equally. They are not. The current formula I am using, is 2/3 SOS.

I am not going to base my model on vegas odds logic.

I still haven't heard any suggestion to spice up my algorithm.

Your SoS metric is based on win% and is generally known as the worst way to judge SoS for the reasons I mentioned.

If you want to spice it up, then you need to recognize and understand where the current formula fails and then change it in order to address it. But you aren't even willing to hear criticism when I am trying to show you a basic part of this stuff and a very common mistake you are making.

 

[occupant]

You'll need a 'weed smokers' column in there somewhere...and maybe a 'baby daddy' column too?

 

[WNY_FOOTBALL_DUDE]

Your SoS metric is based on win% and is generally known as the worst way to judge SoS for the reasons I mentioned.

Remind me again here. With SOS, the harder the schedule, the more tolerate the final result of multiple losses. The lower the SOS, the less tolerant of losses. A 10-2 non-AQ would never be seen on the plane as a 10-2 AQ school. Never.

If you want to spice it up, then you need to recognize and understand where the current formula fails and then change it in order to address it.

There's no CURRENT formula. College football is mostly subjective. For 2014 and beyond, the 7 different computer formulas are gone, and replaced with a 13 person selection committee.

I don't see how any of the past computer formula "fail," nixing the non-transparent nature.

If we simply go by MOV and offense and defensive stats, then we're discriminating against teams playing tough schedules, conference rivals, and teams playing "grind it out football."

Wins and losses, SOS, common opponents/head-to-head match-ups, and conference championship/division status are the traditional pillars for determining playoff teams in sports.

I am all ears, and I am leaning toward weighing home and away games differently.

 

[Codaxx]

Here is the Colley Matrix explained. It is one of the few that publish their formulas. Probably a bit more robust than you were thinking, but perhaps worth a read to give you some ideas.

forgot to add the link.. .http://www.colleyrankings.com/matrate.pdf

 

[4down20]

Remind me again here. With SOS, the harder the schedule, the more tolerate the final result of multiple losses. The lower the SOS, the less tolerant of losses. A 10-2 non-AQ would never be seen on the plane as a 10-2 AQ school. Never.

Your formula is what would measure them as pretty much even. Cumulative helps, but not enough.

There's no CURRENT formula. College football is mostly subjective. For 2014 and beyond, the 7 different computer formulas are gone, and replaced with a 13 person selection committee.

I don't see how any of the past computer formula "fail," nixing the non-transparent nature.

If we simply go by MOV and offense and defensive stats, then we're discriminating against teams playing tough schedules, conference rivals, and teams playing "grind it out football."

Wins and losses, SOS, common opponents/head-to-head match-ups, and conference championship/division status are the traditional pillars for determining playoff teams in sports.

I am all ears, and I am leaning toward weighing home and away games differently.

There are tons of formulas out there.

 

[WNY_FOOTBALL_DUDE]

Your formula is what would measure them as pretty much even. Cumulative helps, but not enough.

There are tons of formulas out there.

No, lets look at an example from Colley Matrix:

South Carolina in 2013 went 10-2, and was ranked 8th overall. Ball went 10-2 as well, and they were ranked 38th.

I will go on with my examples:

Auburn (12-1) - Ranked #1
Ohio State (12-1) - Ranked #4
Alabama (11-1) - Ranked #6
Michigan State (12-1) - Ranked #7
Baylor (11-1) - Ranked #10

Now lets turn to the non-AQ schools and I will include AAC:

UCF (11-1) - #15
NIU (12-1) - #17
Louisville (11-1) - #20
Fresno State (11-1) - #21

Same winning%, but the SOS factor meant that 2-loss Stanford (#3), 2-loss Missouri (#5), 2-loss SC (#8), and 3-loss Arizona State (#9) got ranked higher.

 

[WNY_FOOTBALL_DUDE]

forgot to add the link.. .http://www.colleyrankings.com/matrate.pdf

Thank you Codaxx. I'll be done with my formula within the next couple of weeks. If you have any more suggestions I would love to know.

 

[Codaxx]

Thank you Codaxx. I'll be done with my formula within the next couple of weeks. If you have any more suggestions I would love to know.

-
I have a feeling you will finish your RPI formula shortly and then by next year you will be going to linear regression models.

 

[4down20]

No, lets look at an example from Colley Matrix:

South Carolina in 2013 went 10-2, and was ranked 8th overall. Ball went 10-2 as well, and they were ranked 38th.

I will go on with my examples:

Auburn (12-1) - Ranked #1
Ohio State (12-1) - Ranked #4
Alabama (11-1) - Ranked #6
Michigan State (12-1) - Ranked #7
Baylor (11-1) - Ranked #10

Now lets turn to the non-AQ schools and I will include AAC:

UCF (11-1) - #15
NIU (12-1) - #17
Louisville (11-1) - #20
Fresno State (11-1) - #21

Same winning%, but the SOS factor meant that 2-loss Stanford (#3), 2-loss Missouri (#5), 2-loss SC (#8), and 3-loss Arizona State (#9) got ranked higher.

 

[Jack_John_Mark]

You'll know you have a good formula if your system only gets more accurate as the year goes on.

If it doesn't, then you aren't using enough data and are involving too much opinion.

A computer program shouldn't have much of a clue what's going on in the first few weeks but should begin to form identities for each team as data is collected.

 

[WNY_FOOTBALL_DUDE]

You'll know you have a good formula if your system only gets more accurate as the year goes on.

If it doesn't, then you aren't using enough data and are involving too much opinion.

A computer program shouldn't have much of a clue what's going on in the first few weeks but should begin to form identities for each team as data is collected.

I will only look at the results for the last two weeks of the season, and use my results as a base for my final four picks.

 

[KansasSooner]

-
I have a feeling you will finish your RPI formula shortly and then by next year you will be going to linear regression models.

Sagarin uses a Bayesian method...

 

[WNY_FOOTBALL_DUDE]

Sagarin uses a Bayesian method...

I am not sure if I want to dip into "probability land." Making assumptions about team's strength got us into trouble in the first place.

It would be easier to use probability, if college football didn't lack common opponents.

Just last season, 3/5 BCS Bowls went to unfavored teams.

Baylor was expected to beat UCF.
Alabama was expected to Oklahoma.
Ohio State was expected to beat Clemson

Auburn and Florida, and Stanford and MSU were the only match-ups where the favorites won.

 

[KansasSooner]

I'm not saying you should use a specific method. I was just pointing out what method Sagarin uses. And his poll is considered on the better polls. Honestly, I'm not a fan of it though.

 

[4down20]

I am not sure if I want to dip into "probability land." Making assumptions about team's strength got us into trouble in the first place.

It would be easier to use probability, if college football didn't lack common opponents.

Just last season, 3/5 BCS Bowls went to unfavored teams.

Baylor was expected to beat UCF.
Alabama was expected to Oklahoma.
Ohio State was expected to beat Clemson

Auburn and Florida, and Stanford and MSU were the only match-ups where the favorites won.

You think probability is an assumption?

 

[7NationalChampionships]

last time I checked an algorithm never won or lost a game...

 

[WNY_FOOTBALL_DUDE]

You think probability is an assumption?

My point is look at history.

Alabama was suppose to rip apart Utah and Oklahoma. They didn't.
Florida was suppose to be no match for Louisville.
Clemson was a big favorite over West Virginia.

Why must rankings be base on vegas odds? It should be about RESUME, not on subjective assumptions or probabilities.