SEC OOC schedule

[mad2mc]

Our schedule is just about the same as when we were in the Big 12.

1 Super Soft Cupcake
1 Semi competent Cupcake
1 Low Power Conference Team
1 Decent Game 2 hours away (BYU now....used to be Illinois is STL)

IIRC, we got our soft scheduling model from Kansas St.

I haven't looked at your schedule, but have you changed your scheduling towards the end of the season? More in the sense that the next to last game, or the before, is a bye or against a soft cupcake?

 

[4down20]

Not so sure why you continually respond to me since I insulted you, according to your forum posting etiquette, but wouldn't the information and stats be available somewhere in cyberspace? To continue playing the devil's advocate, if you had a system that was 75% successful, why couldn't you use something to account for injuries and teams with stock going up/down? Is it possible that you were throwing out some of the wrong things that 'matter'?

Well, I'd need to have a complete list of injuries every week to start with. But even then it's near impossible to do accurately. How do you rate the backup? How do you know if the backup is any good?

Look at Ohio St this past year. Win the national championship on their 3rd string QB, who really looked better than the previous 2 IMO. How do you put that into math form? Do we say the backup is x % the strength of the previous? Do we expect X amount of dropoff?

Everything with my rankings was 100% based off stats, no human opinion. I didn't rate teams manually or anything of such, it was all based on stats on the field. Which also brings up another issue - limited data. With only 12 games, the data is limited. So if you plug in for the injury, your data becomes even more limited, thus increasing the likelihood of bad results.

So while it's not perfect because you can't take that into account, it's actually the best way to do things overall.

As I've explained before, I could care less who disagrees or believes me. I don't see much value in lying on these boards nor believe in little 'victories' as you truly feel is a must in your world. When people do not believe you, you get all bent out of shape and start the insults.

I'm familiar with trolls from other websites that most of us on this board have been. If you believe I'm a troll, I'm fine with that. Not sure why I would troll my own site, but whatever. Maybe if you approached this group in a different manner, somebody may have found your hobby interesting. I wouldn't, but somebody may.

A troll is someone who only posts for the purpose of trying to piss people off.

 

[WVUDAD]

Well, I'd need to have a complete list of injuries every week to start with. But even then it's near impossible to do accurately. How do you rate the backup? How do you know if the backup is any good?

Look at Ohio St this past year. Win the national championship on their 3rd string QB, who really looked better than the previous 2 IMO. How do you put that into math form? Do we say the backup is x % the strength of the previous? Do we expect X amount of dropoff?

Everything with my rankings was 100% based off stats, no human opinion. I didn't rate teams manually or anything of such, it was all based on stats on the field. Which also brings up another issue - limited data. With only 12 games, the data is limited. So if you plug in for the injury, your data becomes even more limited, thus increasing the likelihood of bad results.

So while it's not perfect because you can't take that into account, it's actually the best way to do things overall.



A troll is someone who only posts for the purpose of trying to piss people off.

Ala the quartet.......

 

[bbwvfan]

Well, I'd need to have a complete list of injuries every week to start with. But even then it's near impossible to do accurately. How do you rate the backup? How do you know if the backup is any good?

Look at Ohio St this past year. Win the national championship on their 3rd string QB, who really looked better than the previous 2 IMO. How do you put that into math form? Do we say the backup is x % the strength of the previous? Do we expect X amount of dropoff?

Everything with my rankings was 100% based off stats, no human opinion. I didn't rate teams manually or anything of such, it was all based on stats on the field. Which also brings up another issue - limited data. With only 12 games, the data is limited. So if you plug in for the injury, your data becomes even more limited, thus increasing the likelihood of bad results.

So while it's not perfect because you can't take that into account, it's actually the best way to do things overall.

I'm interested in this stat based ranking system. Did you account for stats generated against AQ vs nonAQ opponents? Did you account for stats generated against FCS opponent?

If the stats were simply input from the results on the field without differentiating between opponent, the ranking would be biased.

 

[mad2mc]

Ala the quartet.......

:rollseyes:

 

[4down20]

I'm interested in this stat based ranking system. Did you account for stats generated against AQ vs nonAQ opponents? Did you account for stats generated against FCS opponent?

If the stats were simply input from the results on the field without differentiating between opponent, the ranking would be biased.

It didn't use labels like "AQ" and "nonAQ" as such a label doesn't mean anything.

The very first thing that is done with stats is to get what I call "quality stats". Because rushing for 200 yards against New Mexico St(or the worst rush defense in college football) is hardly the same as rushing for 200 yards against the best rush defense. So all the stats for games were adjusted and normalized in this manner to get the real strength of the performances.

And I did not use FCS game stats at all because for the most part there is nothing of value to be gained. I penalized teams for playing FCS teams in the rankings as well.

 

[bbwvfan]

So how do you determine a quality stat? Good Rush defenses could simply be due to poor rushing teams in a conference. SEC may have good pass defense stats, but recently they have had crap at QB in the conference.

You normalize, but don't differentiate between AQ and nonAQ. While statistical college football sites will show nonAQ teams are less competitive against AQ teams, you say such a label doesn't mean anything.

But, it does mean something...

 

[4down20]

So how do you determine a quality stat? Good Rush defenses could simply be due to poor rushing teams in a conference. SEC may have good pass defense stats, but recently they have had crap at QB in the conference.

You normalize, but don't differentiate between AQ and nonAQ. While statistical college football sites will show nonAQ teams are less competitive against AQ teams, you say such a label doesn't mean anything.

But, it does mean something...

Trends form over a season, the effect of such things you mention are pretty minimal usually.

The label doesn't mean anything, but the difference shows up big time in the normalized stats. That's the purpose for normalizing them. Lots of other sites do the same things, they call them "advanced stats". There is no trouble at all finding who are the better teams.

As for "crap at QB" in the SEC, maybe you should check the NFL for that.

Which conference produces most NFL starting QBs? « Big Ten Network

The Pac12 has the most starters with 6, the SEC has 5 tied with the Big10 and the Big12 has 4.

 

[bbwvfan]

Trends form over a season, the effect of such things you mention are pretty minimal usually.

The label doesn't mean anything, but the difference shows up big time in the normalized stats. That's the purpose for normalizing them. Lots of other sites do the same things, they call them "advanced stats". There is no trouble at all finding who are the better teams.

As for "crap at QB" in the SEC, maybe you should check the NFL for that.

Which conference produces most NFL starting QBs? « Big Ten Network

The Pac12 has the most starters with 6, the SEC has 5 tied with the Big10 and the Big12 has 4.

Not sure what starters in the NFL has to do with the current state of the SEC. For the past two years, QB play in the SEC has been among the worst. To add to this argument…. great college QB's and production in college mean little to what the NFL is looking for… or what may work in the NFL compared to the college game. Bringing up the number of starting QB's in the NFL is a common retort by SEC homers…. gets a chuckle from most football fans every time.

It is a common fallacy to believe stats normalize over the course of a season. All we have to do is look at the 2013 Bama team for example. Bama ended the year ranked #11 in pass defense… averaging 180.3 yds/gm. OU which ended the season ranked #90 in with 199.1 yd/gm average passed for 348 yds in the Sugar Bowl.
Had Bama played in the Big 12 in '13, there is no way its defense would have averaged giving up only 180 yds passing for the year.

 

[4down20]

Not sure what starters in the NFL has to do with the current state of the SEC. For the past two years, QB play in the SEC has been among the worst. To add to this argument…. great college QB's and production in college mean little to what the NFL is looking for… or what may work in the NFL compared to the college game. Bringing up the number of starting QB's in the NFL is a common retort by SEC homers…. gets a chuckle from most football fans every time.

It is a common fallacy to believe stats normalize over the course of a season. All we have to do is look at the 2013 Bama team for example. Bama ended the year ranked #11 in pass defense… averaging 180.3 yds/gm. OU which ended the season ranked #90 in with 199.1 yd/gm average passed for 348 yds in the Sugar Bowl.
Had Bama played in the Big 12 in '13, there is no way its defense would have averaged giving up only 180 yds passing for the year.

You site raw stats as proof of something that happens. In normalized stats, the average yards/gm that Alabama did means nothing specific. It's all adjusted. So if the SEC isn't among the top passing offenses, then the SEC defenses aren't getting credit the same way they would be if they are playing top passing offenses.

Obviously the concept of normalization is beyond you, and yet you call it a fallacy. Because you site a problem which only happens in raw stats and is what the normalization is specifically used to prevent.

 

[bbwvfan]

You site raw stats as proof of something that happens. In normalized stats, the average yards/gm that Alabama did means nothing specific. It's all adjusted. So if the SEC isn't among the top passing offenses, then the SEC defenses aren't getting credit the same way they would be if they are playing top passing offenses.

Obviously the concept of normalization is beyond you, and yet you call it a fallacy. Because you site a problem which only happens in raw stats and is what the normalization is specifically used to prevent.

I have to admit, I didn't sleep at a Holiday Inn Express last night… and it has been a long time since I took statistics. I hated that damn class anyways.

I did not say the concept of normalization is fallacy. I said it is common fallacy to believe stats normalize over the course of a season. There are too many variables to attempt normalization of the data.

I cited but one example…

 

[4down20]

I have to admit, I didn't sleep at a Holiday Inn Express last night… and it has been a long time since I took statistics. I hated that damn class anyways.

I did not say the concept of normalization is fallacy. I said it is common fallacy to believe stats normalize over the course of a season. There are too many variables to attempt normalization of the data.

I cited but one example…

Stats don't normalize over the season, the data becomes more reliable over the season as trends present themselves and you get cross conference games which provide links between the conferences. The data is normalized by formula.

I didn't just pick winners for games, I predicted scores as well. Hitting them was not very common actually, but it was common for my spread to be real close or exact to Vegas. I'm pretty sure Vegas uses formulas a good bit for initial lines.

 

[bbwvfan]

Stats don't normalize over the season, the data becomes more reliable over the season as trends present themselves and you get cross conference games which provide links between the conferences. The data is normalized by formula.

I didn't just pick winners for games, I predicted scores as well. Hitting them was not very common actually, but it was common for my spread to be real close or exact to Vegas. I'm pretty sure Vegas uses formulas a good bit for initial lines.

Ohh, if I remember correctly...most mathematicians criticized some computer models because they did not account for SOS, what teams were beat or lost to. Why many critiqued the top 20 ranking of Kent St one year by one system.

As I said, there is a fallacy with normalization....

 

[bbwvfan]

Back in the CBS days, kingbd and sleepy used to have a game among posters to PTW and pick ATS. I had over 80% winning rate in the PTW...many weeks in the 90-95% range.

I know you have to be impressed...

 

[4down20]

Ohh, if I remember correctly...most mathematicians criticized some computer models because they did not account for SOS, what teams were beat or lost to. Why many critiqued the top 20 ranking of Kent St one year by one system.

As I said, there is a fallacy with normalization....

Not all formulas are the same. I don't know how you can do anykind of accurate rankings without the data behind the SoS rankings(the rankings are just the order of the data).

 

[4down20]

Back in the CBS days, kingbd and sleepy used to have a game among posters to PTW and pick ATS. I had over 80% winning rate in the PTW...many weeks in the 90-95% range.

I know you have to be impressed...

Better than 75% is good anywhere.

 

[bbwvfan]

Not all formulas are the same. I don't know how you can do anykind of accurate rankings without the data behind the SoS rankings(the rankings are just the order of the data).

That is what I remember being the big beef about the computer ranking which was criticized. I can't recall, but I think some of them refused to provide their algorithm.

I did not understand why you did not differentiate between AQ and nonAQ as this distinction affects the SOS.

 

[4down20]

That is what I remember being the big beef about the computer ranking which was criticized. I can't recall, but I think some of them refused to provide their algorithm.

I did not understand why you did not differentiate between AQ and nonAQ as this distinction affects the SOS.

Boise St is a nonAQ, does that mean they can't be as good as a P5 team or provide the same type of schedule strength as a P5 team? All teams are treated equally, there is no special rules for them outside FCS or FBS.

Maybe you are confused with SoS and what it is. I do not use anyone elses SOS and I would never use win% or anything like that, which is the worst form of SoS and is how you get Kent St ranked high. Even the ones that use opponents of opponents win% are flawed badly. Here's a basic rundown of the process.

Ok, so it starts with getting quality data out of it all. This is where the separations of team strength is really done and is where done. I explained it earlier, but who you rush for yards against and so forth is adjusted based on the opponents base stats(obviously have to start somewhere).

These newly adjusted stats are what gives a team it's general power that is used in the future for other functions. As in this state, the stats can be totaled and ordered to give you a raw power score for the teams. I use that data in order to predict games, and it has no trouble at all predicting Alabama is going to beat Kent St by 50+ points.

You can use that power score and make a SoS from it, where say SoS = average power score of opponents as a simple form. I tried something similiar to that for awhile. But ultimately what I ended up doing was using the predictor function and I would play each team against every other FBS team and get their predicted record. I would then use the % they would be predicted to win in as my power score. I would treat all FCS scores as 0% and I would not include their stats with the other games. Get the average power score for all teams and you have SoS rankings.

That score is important when it comes to making rankings. I took the power score(not the rankings, they aren't useful for formulas) and if you beat the team, you gained their power score. If you lost to them, then you lost points, 100 - power score. So you lose to a bad team, you lose a bunch of points, you lose to a good team, you don't lose so much. Like wise, if you win against a good team you gain points, and if you win against a good team you don't gain much. Winning a FCS game = 0 pts, losing = -100 pts. Add that score up, average it out and sort them and you have good rankings.

If someone asked me for the algorithm I'd laugh at them. That's like asking Coca Cola for their recipes. Don't mind explaining the process, but if I wanted to put it in the public domain I'd just do it.

 

[WVUDAD]

That is what I remember being the big beef about the computer ranking which was criticized. I can't recall, but I think some of them refused to provide their algorithm.

I did not understand why you did not differentiate between AQ and nonAQ as this distinction affects the SOS.

So, are you trying to say playing Boise is easier than playing indiana?
That Kansas is better win than colorado state? Grading teams by the league they play in is ridiculous, Marshall would have has five wins at least in the B11 this past year.

 

[bbwvfan]

I get confused easily. Part of the problem of being inbred.

I like your BSU reference. Boise St. plays in a weaker conference. Back when you were doing your system, they might play 1 or 2 AQ schools per year. Those games occurred at the start of the year. Then, the rest of it they played weaker teams.

If Boise St. played in an AQ conference, I do not believe they'd fare very well. While playing in an nonAQ conference, they can rack up the wins. So a loss to them by UGA or VT might not work against them as much.

So, I do not agree that they provide the same schedule strength.

It is why you cannot use mathematical models...