Conference Realignment

[UgaBlows]

It seems to me that this game is going to come down to the Michigan defensive line vs Washington offensive line. If Michigan's d-line gets quick penetration and does not allow Penix time to survey the field for the deep throws that he likes, I think Michigan will win.

He is one of the best i’ve seen in awhile at moving around in the pocket to avoid pass rushers but still surveying the field for receivers, i’m excited to see how this game plays out

 

[CEB]

He is one of the best i’ve seen in awhile at moving around in the pocket to avoid pass rushers but still surveying the field for receivers, i’m excited to see how this game plays out

No doubt… and its effortless looking when he does it.

I think Michigan could control the tempo, run the football and wear the Huskies out. I also think that Michigan might overthink it and not pound the run the way they should.

Penix is going to move the football and score points. If Wash D can get a couple of early stops, Michigan might find themselves playing from behind and further abandon the run.

I think it goes down the wire, scores are in the upper 30s, and Penix / Washington makes one more play than UM.

 

[slugboy]

I get what you’re saying but in the case of Oregon and Washington, there is not a single impressive Oregon win that Washington doesn’t also have or better.
Washington also has a win over Arizona (Oregon didnt play them), and two wins over Oregon. I don’t see three wins on Oregon’s schedule that are as good as those three.
Means the metrics must be emphasizing some style points to get there. It’s odd, to say the least.

If I looked at the link correctly, they have Texas, ND and Bama above Washington too?

The predictive models will put emphasis on point differential before garbage time, EPA, offensive and defensive efficiency, etc. it’s not just who you beat but how thoroughly you dominated them and how good they were.
In boxing terms, if two boxers fought the same opponents, but one kept knocking their opponents out in early rounds, you’d favor the boxer with the knockouts.

For the playoffs, I like weighting resumes more—put in the teams that earned it, like FSU and Washington. Predictive rankings favor UGA and Oregon more.

There are so few games in a college football season, that people keep pointing to a game or two. In baseball, you have 162 games. The Braves can lose the season series to the Marlins, but be 15 games ahead of them, and no one bats an eye. No one reasonable says “the Marlins beat the Braves, so they’re a better team”. In college football, people get wrapped up in one game because there are so few of them.

Twelve games are so few that bad luck, an injury, or a bad 1:1 matchup can sway the outcomes a lot.

And, on the “avoiding the TCU blowout” question—TCU won their semifinal. They belonged in the CFP. UGA would’ve been a bad matchup for almost anyone.

 

[RonJohn]

The predictive models will put emphasis on point differential before garbage time, EPA, offensive and defensive efficiency, etc. it’s not just who you beat but how thoroughly you dominated them and how good they were.
In boxing terms, if two boxers fought the same opponents, but one kept knocking their opponents out in early rounds, you’d favor the boxer with the knockouts.

Nothing negative against you, but I don't like models and statistical analysis for football for anything other than pure entertainment. Even something as old and tested as Sagarin's ratings is not very good at predicting. You can use some stats, like passing yards per game gained and allowed to predict that a prolific passing team is going to have a field day against a troubled passing defensive team. However, trying to predict who will actually win games using those models is ridiculous in my opinion. There isn't enough data available in any season to get an accurate model established.

I think the biggest problem is that some people see "statistical model" and believe that they produce accurate results. One of the most glaring to me is the ESPN win probability number. I don't think it does anything at all except for perk up people's interest in a game. (Which is the only reason that ESPN has it in the first place)

 

[Jim Prather]

 

[Jim Prather]

Sorry about the size

 

[slugboy]

Nothing negative against you, but I don't like models and statistical analysis for football for anything other than pure entertainment. Even something as old and tested as Sagarin's ratings is not very good at predicting. You can use some stats, like passing yards per game gained and allowed to predict that a prolific passing team is going to have a field day against a troubled passing defensive team. However, trying to predict who will actually win games using those models is ridiculous in my opinion. There isn't enough data available in any season to get an accurate model established.

I think the biggest problem is that some people see "statistical model" and believe that they produce accurate results. One of the most glaring to me is the ESPN win probability number. I don't think it does anything at all except for perk up people's interest in a game. (Which is the only reason that ESPN has it in the first place)

I’m not saying use a computer model to pick teams. I’m saying use objective criteria and objective rules. Define in advance what it means to deserve your shot, and stick with it.

But, as far as accuracy,!I can show you how well they do in predicting games and how they do against the point spreads. They’ll do better than a committee vote.

A lot of stats get abused. The “should I go for it on 4th and 2” stat you see used doesn’t have a “the defense has Suh and Campbell at DT” component to it. The percentage win prediction is probably not all that far off—the odds of someone not taking a knee and fumbling and the scoring in 30 seconds from deep in their own side are pretty low. It’s probably off some—it needs “fatter tails”.

I’ve used models to talk about who to pick, but one reason why is that they’re clear about the rules. Before you make a big decision, set up your decision criteria—make it clear and objective. If you’re buying a car, decide what a good deal is in advance—if the dealer keeps making changes, walk away.

What the CFP committee did was make a decision in the same way people end up buying too much car or a house they can’t afford. It’s also the same way a CEO pushes a merger when it’s not a good value, or how an athletic department hires a bad head coach. They went unprepared into a decision, and they kept rationalizing what they were doing.

MLB wouldn’t let a committee put the Yankees in the playoffs if they didn’t meet a pre-defined standard. Smart people set up rules so they don’t panic. The CFP committee panicked.

 

[stinger 1957]

So far the TV folk got what they wanted, two outstanding games. Just insures they continue to drive the bus as all this goes forward IMO.

 

[Northeast Stinger]

So far the TV folk got what they wanted, two outstanding games. Just insures they continue to drive the bus as all this goes forward IMO.

But, it’s still an arbitrary distinction. There have been 9 college football championship games in this format and 6 of them have been blowouts.

I hope this one is a good, tight game but blowouts have been common.

So what’s my point? I have no point. Just chewing the fat.

 

[CEB]

The predictive models will put emphasis on point differential before garbage time, EPA, offensive and defensive efficiency, etc. it’s not just who you beat but how thoroughly you dominated them and how good they were.
In boxing terms, if two boxers fought the same opponents, but one kept knocking their opponents out in early rounds, you’d favor the boxer with the knockouts.

For the playoffs, I like weighting resumes more—put in the teams that earned it, like FSU and Washington. Predictive rankings favor UGA and Oregon more.

There are so few games in a college football season, that people keep pointing to a game or two. In baseball, you have 162 games. The Braves can lose the season series to the Marlins, but be 15 games ahead of them, and no one bats an eye. No one reasonable says “the Marlins beat the Braves, so they’re a better team”. In college football, people get wrapped up in one game because there are so few of them.

Twelve games are so few that bad luck, an injury, or a bad 1:1 matchup can sway the outcomes a lot.

And, on the “avoiding the TCU blowout” question—TCU won their semifinal. They belonged in the CFP. UGA would’ve been a bad matchup for almost anyone.

I completely understand what you’re saying and I think the methodology may be sound when it comes to teams who haven’t faced each other and have few if any common opponents. I also certainly understand the issue of a football season producing relatively few data points (games) relative to other sports.
I guess what I am saying is that as sound as their logic might be for what they’re doing, their data is missing something significant to effectively dismiss quality wins and even head to head results twice. If Washington had lost a game or if they had split with Oregon in one of their two meetings, I could see it. As it stands, there appears to be a pretty big omission in their data / formula.

 

[forensicbuzz]

Not sure where to put this so….

What is everyone’s thinking about the Michigan vs Washington game?

Who do you want to win?

Who do you think will win?

Isn’t this fun for a change, having two interesting non-SEC teams in a game that you can just enjoy watching without it being some propaganda piece for ESPN?

Michigan advantages: Most physical team in the country, strong defense and balanced offense. Disadvantages: Special teams are capable of giving a game away.

Washington advantages: Perhaps the most dynamic quarterback in the nation. Speed to burn. Offense with a lot of complications for any defense to figure out. Disadvantages: Can get sloppy at times in any given phase of the game. Not quite as physical as most top teams. Defense, especially secondary, occasionally gives up the big play.

Who do you like, what do you think?

You put it in Other Professional and College Sports, where it belongs.

 

[slugboy]

I completely understand what you’re saying and I think the methodology may be sound when it comes to teams who haven’t faced each other and have few if any common opponents. I also certainly understand the issue of a football season producing relatively few data points (games) relative to other sports.
I guess what I am saying is that as sound as their logic might be for what they’re doing, their data is missing something significant to effectively dismiss quality wins and even head to head results twice. If Washington had lost a game or if they had split with Oregon in one of their two meetings, I could see it. As it stands, there appears to be a pretty big omission in their data / formula.

Lots of people are getting hung up on this. It’s simple

1. If your standard is “pick the four teams that have the best odds of winning it all on December 1st”, Bama probably wouldn’t make it. Maybe you get UGA, OSU, MICH, ORE. Maybe you swap Washington for Oregon. You could take the top 25 teams and go to Vegas and get the odds. You could use a computer model. They’d be about the same.
2. If your standard is “pick the four teams that earned their way here as of the selection date”, you get Washington, FSU, Michigan, and Texas. That’s the easiest to figure out
3. If your standard is “give me the three most compelling games”, you’re going to pick four evenly matched teams—not necessarily 1-4. Heck, we might be one.

The process here was Sankey saying “you gotta pick an SEC team”. He might have even told ESPN that he’d take the SEC games to Fox or CBS for all I know. This was the CFP committee being scared to leave the SEC out.

I can find a model that liked Washington better than Oregon. The predictive models are going to be the summed value of all the plays (minus garbage time) weighted vs your opponents. The resume models are going to be more focused on wins and losses.

The models that are the best predictors of final scores aren’t the same as the ones that measure how much you accomplished at the end of the season, and the rankings come out significantly different. The latter are the ones that focus on who beat who.

 

[CEB]

Lots of people are getting hung up on this. It’s simple

1. If your standard is “pick the four teams that have the best odds of winning it all on December 1st”, Bama probably wouldn’t make it. Maybe you get UGA, OSU, MICH, ORE. Maybe you swap Washington for Oregon. You could take the top 25 teams and go to Vegas and get the odds. You could use a computer model. They’d be about the same.
2. If your standard is “pick the four teams that earned their way here as of the selection date”, you get Washington, FSU, Michigan, and Texas. That’s the easiest to figure out
3. If your standard is “give me the three most compelling games”, you’re going to pick four evenly matched teams—not necessarily 1-4. Heck, we might be one.

The process here was Sankey saying “you gotta pick an SEC team”. He might have even told ESPN that he’d take the SEC games to Fox or CBS for all I know. This was the CFP committee being scared to leave the SEC out.

I can find a model that liked Washington better than Oregon. The predictive models are going to be the summed value of all the plays (minus garbage time) weighted vs your opponents. The resume models are going to be more focused on wins and losses.

The models that are the best predictors of final scores aren’t the same as the ones that measure how much you accomplished at the end of the season, and the rankings come out significantly different. The latter are the ones that focus on who beat who.

I understand… and I’m not even talking about how we got the four picked this postseason. That’s more political and media driven than anything, even though they are trying to sell it to us as resumes, data and analytics.

Just saying at some point your process may be flawed if your “better” teams have achieved less and continue to lose to “lesser” teams. Which is why the “best” vs “most deserving” garbage is so disingenuous… it starts throwing out actual results.

But I’m not wanting to dive into the selection committee decisions… just saying that a “predictive” model may not be synonymous with selecting “best” teams especially if you don’t adjust for actual results.

 

[stinger78]

Nothing negative against you, but I don't like models and statistical analysis for football for anything other than pure entertainment. Even something as old and tested as Sagarin's ratings is not very good at predicting. You can use some stats, like passing yards per game gained and allowed to predict that a prolific passing team is going to have a field day against a troubled passing defensive team. However, trying to predict who will actually win games using those models is ridiculous in my opinion. There isn't enough data available in any season to get an accurate model established.

I think the biggest problem is that some people see "statistical model" and believe that they produce accurate results. One of the most glaring to me is the ESPN win probability number. I don't think it does anything at all except for perk up people's interest in a game. (Which is the only reason that ESPN has it in the first place)

The small sample sizes also hinder predictive power. There are ways to deal with small samples, but it’s not easy with the error inherent in sports events.

 

[leatherneckjacket]

Let's say for the sake of argument a team starts the season with their starting QB injured and lose three times in the first few weeks of the year. Then, after their QB comes back, they proceed to win the rest of their games by a large margin, including against a #1 team. They win their conference championship, but have a worse record than four other teams who are undefeated from other four power conferences. By the eye test, they are the clearly the best team in the country, but they have much worse record as well as a loss to the undefeated conference champion from the ACC. Further, the other four teams have had significant injuries in their conference championship games. It seems to be that the logic of some of the fonts is that the three loss team should get in over one of the undefeated conference champion because they are the best team and the other teams are not as good as their record states since they have current injuries. Is this correct? I guess I did not know it was solely based on subjective who one thinks is the best team on selection day criteria.

 

[forensicbuzz]

I understand… and I’m not even talking about how we got the four picked this postseason. That’s more political and media driven than anything, even though they are trying to sell it to us as resumes, data and analytics.

Just saying at some point your process may be flawed if your “better” teams have achieved less and continue to lose to “lesser” teams. Which is why the “best” vs “most deserving” garbage is so disingenuous… it starts throwing out actual results.

But I’m not wanting to dive into the selection committee decisions… just saying that a “predictive” model may not be synonymous with selecting “best” teams especially if you don’t adjust for actual results.

Isn't that why you play the games? I know it's not apples-to-apples, but an NCAA 15-2 upset is a head-to-head game that determines who moves on. Now, we all know that if they play 20 times, the 15 will upset the 2 once, maybe twice. But, the 15 team won on the field and deserves to get the chance. That's why the cfp (lower cased on purpose, just like uga) decision was bogus. Alabama and Texas didn't earn it on the field. FULL STOP.

Edit: I'm not disagreeing with what you said. Using your statement and a point.

 

[ThatGuy]

Not sure where to put this so….

What is everyone’s thinking about the Michigan vs Washington game?

Who do you want to win?

Who do you think will win?

Isn’t this fun for a change, having two interesting non-SEC teams in a game that you can just enjoy watching without it being some propaganda piece for ESPN?

Michigan advantages: Most physical team in the country, strong defense and balanced offense. Disadvantages: Special teams are capable of giving a game away.

Washington advantages: Perhaps the most dynamic quarterback in the nation. Speed to burn. Offense with a lot of complications for any defense to figure out. Disadvantages: Can get sloppy at times in any given phase of the game. Not quite as physical as most top teams. Defense, especially secondary, occasionally gives up the big play.

Who do you like, what do you think?

This one is a tough one for me. I'm definitely glad to have two non-SEC teams in there, but...

On the one hand, you have the cheaters (UofM).
On the other hand, you have the liars (Washington - who bailed on the Pac12, then tried to sue to get out of their agreement and take all the money from the two teams they left behind).

I don't like either of them because of their behavior.

On a play side, I liked watching Washington this week. I thought they were fun to watch, and Penix is for sure a baller. Amazing that he transferred in from Indiana - kinda wish Indiana had been able to do something with him, but that's another post for another time).

But then I come back to my original take - cheaters vs. liars.

So I'm going to put my Team Chaos hat on, and hope that Michigan wins it all. And then the NCAA comes in and strips them of some of their wins due to recruiting violations or some other violation, and the whole year goes up in flames (as is fitting, after what the cfp did with the selection process this year).

 

[Root4GT]

Hmmmm. If the metrics keep saying Oregon is better but Washington keeps beating Oregon, I think I would start questioning the metrics....but that's just me.

The metrics in college football really are not very good or useful. The data points they use are not very comparable as with 130 FBS teams the actual comparison of teams vs teams is very low. Then of course these same metrics when they do have data form two teams where one team beat the other two times and the winner of both games never lost the metrics say the loser is better, Pretty much sums up how vaid the metrics are. Just additional information, no better or worse than a group of people watching multiple games of the 10 or so teams with the best records.

 

[Root4GT]

I’m not saying use a computer model to pick teams. I’m saying use objective criteria and objective rules. Define in advance what it means to deserve your shot, and stick with it.

But, as far as accuracy,!I can show you how well they do in predicting games and how they do against the point spreads. They’ll do better than a committee vote.

A lot of stats get abused. The “should I go for it on 4th and 2” stat you see used doesn’t have a “the defense has Suh and Campbell at DT” component to it. The percentage win prediction is probably not all that far off—the odds of someone not taking a knee and fumbling and the scoring in 30 seconds from deep in their own side are pretty low. It’s probably off some—it needs “fatter tails”.

I’ve used models to talk about who to pick, but one reason why is that they’re clear about the rules. Before you make a big decision, set up your decision criteria—make it clear and objective. If you’re buying a car, decide what a good deal is in advance—if the dealer keeps making changes, walk away.

What the CFP committee did was make a decision in the same way people end up buying too much car or a house they can’t afford. It’s also the same way a CEO pushes a merger when it’s not a good value, or how an athletic department hires a bad head coach. They went unprepared into a decision, and they kept rationalizing what they were doing.

MLB wouldn’t let a committee put the Yankees in the playoffs if they didn’t meet a pre-defined standard. Smart people set up rules so they don’t panic. The CFP committee panicked.

The CFP used the last line in their guidance on leaving FSU out. Their guidelines gave them plenty of leeway. The CFP Committee is just as good as any model for picking the teams. There needs to be some leeway for the committee as there was this year. Had they picked FSU that would have been fine as well. The games turned out as good as possible and next year the 12 team CFP makes this debate pointless at this time.

 

[CEB]

So I'm going to put my Team Chaos hat on, and hope that Michigan wins it all. And then the NCAA comes in and strips them of some of their wins due to recruiting violations or some other violation, and the whole year goes up in flames (as is fitting, after what the cfp did with the selection process this year).

How about if Washington wins it all, and the courts rule that any and all proceeds belong to the “PAC 12,” to be split between Oregon St and Washington St?

I’d be OK with Michigan having to vacate wins, except for the fact that Bama would probably claim another bogus national title.