Stats models and rankings

[slugboy]

The top 10 teams in time of possession this year:

RK	Team	G	Total	Avg
1	Air Force	6	208	34
2	Penn St.	6	207	34
3	Liberty	7	240	34
4	Utah	6	205	34
5	Florida	7	238	34
6	Georgia	7	235	33
7	West Virginia	6	201	33
8	Ohio	7	234	33
9	Stanford	6	196	32
10	Cincinnati	6	196	32

The top 10 in total offense:

RANK	TEAM	G	PLAYS	YDS	YDS/PLAY	YPG
1	Oregon	6	429	3320	7.74	553.3
2	LSU	7	487	3853	7.91	550.4
3	Washington	6	384	3262	8.49	543.7
4	UCF	6	414	3100	7.49	516.7
5	Georgia	7	500	3566	7.13	509.4
6	Oklahoma	6	452	3036	6.72	506.0
7	Miami (FL)	6	411	3011	7.33	501.8
8	North Carolina	6	465	3008	6.47	501.3
9	Southern California	7	444	3442	7.75	491.7
10	Texas St.	7	505	3427	6.79

And the top 10 in scoring offense:

RANK	TEAM	G	TDS	FG	PTS	PPG
1	Oregon	6	38	7	291.00	48.5
2	Southern California	7	45	5	331.00	47.3
3	LSU	7	41	9	317.00	45.3
4	Oklahoma	6	35	8	271.00	45.2
5	Penn St.	6	35	7	266.00	44.3
-	Washington	6	36	4	266.00	44.3
7	Florida St.	6	33	7	253.00	42.2
8	Ole Miss	6	31	11	250.00	41.7
9	Georgia	7	35	12	281.00	40.1
10	Michigan	7	37		276.00	39.4

As you can see, there is very little to zero correlation between time of possession and offensive potency. Only 1 team in the top 10 of TOP is also in the top 10 of either total offense or scoring offense: Georgia. The available yards stat is essentially an offensive (or defensive) efficiency metric that shows you how well your offense moves the ball on each drive, or how poor of a job your defense does at preventing the other team from moving the ball. It’s pretty simple really.

Of the top 10 TOP teams, here’s where they rank in the offensive available yards stat:
1. Air Force -> #7 OAY
2. Penn State -> #26 OAY
3. Liberty -> #9 OAY
4. Utah -> #105 OAY
5. Florida -> #36 OAY
6. Georgia -> #5 OAY
7. West Virginia -> #67 OAY
8. Ohio -> #55 OAY
9. Stanford -> #50 OAY
10. Cincinnati -> #49 OAY

In today’s era of college football, a high TOP is usually a signifier of an inefficient offense not capable of producing big plays.

Air Force is an outlier—efficient, high TOP, runs the 3O. They score, but they own the clock, too. They don’t put up Oregon or Southern Cal points, though, because of pace

Rk​	Team	Rec​	FBS​	OFEI​
13​	Air Force​	6-0​	5-0​	.88​

 

[Augusta_Jacket]

In today’s era of college football, a high TOP is usually a signifier of an inefficient offense not capable of producing big plays.

In the modern game, excluding the 3O teams, the only time TOP is really important to control is in the 4th quarter when protecting a lead.

 

[ibeattetris]

Or, say you got the ball at the 50 and drove down to the one, where you left it on fourth down. You gobbled up 98% of your available yardage, and true you left them deep in their own territory, but you put zilch on the scoreboard, where it counts. My point is, that extra 2% you would have gotten had you scored barely moves the needle on available yardage but makes all the difference as to who wins the game. So it's 98% on the stat, but nothing on the board. Where it counts.

If you start from your own one, drive 49 yards and leave it on the 50, you will also have "zilch", will have gained the same total yards, but will have left the ball at the 50 for the opposing team. Yes, I will choose the 98% drive over the 49% drive, even with the same number of yards and points scored.

You obviously want drives with 100% because it means you are scoring touchdowns. Consistently scoring touchdowns is how teams win football games. Therefore, you want a team who, on average, gets closer to scoring touch downs (including actually scoring) than other teams.

 

[ibeattetris]

Is there a method similar to Colley that starts everyone on equal level and only uses the final score?

Massey's rating used to be this way back in the BCS era. He had one where score was not used (due to BCS rules) and another where it was.

Massey Ratings FAQ

First, to avoid confusion, be aware that I publish two different sets of rankings:

• the "Massey Ratings", which utilize actual game scores and margins in a diminishing returns fashion
• the BCS compliant version, which do not use the actual score

I will summarize the latter, since that is the one relevant to college football fans.
Massey's BCS ratings are the equilibrium point for a probability model applied to the binary (win or loss) outcome of each game. All teams begin the season rated the same.

The main version does consider scoring margin, but its effect is diminished as the game becomes a blowout. The score of each game is translated into a number between 0 and 1. For example 30-29 might give 0.5270, while 45-21 gives 0.9433 and 56-3 gives around 0.9998

The maximum is topped at 1, so the curve flattens out for blowout scores. In addition, I do a Bayesian correction to reward each winner, regardless of the game's score.

The net effect is that there is no incentive to run up the score. However, a "comfortable" margin (say 10 points) is preferred to a narrow margin (say 3 points).

In summary, winning games against quality competition overshadows blowout scores against inferior opponents. Each week, the results from the entire season are re-evaluated based on the latest results. Consistent winners are rewarded, and a blowout score has only marginal effect on a team's rating.

 

[ibeattetris]

I can see both sides of the argument over the utility of this stat. (After I finally overcame my semantic confusion over y'all's use of the term "available yards", which is just a function of starting field position x # of drives, when you actually mean the ratio of actual yards to available yards (DAY) as a defensive metric, or OAY as an offensive one.)

One can think of numerous examples which either support the utility of this metric or undermine it, simply by moving the starting field position. Case in point, in @ibeattetris example of a team that started from their 50, gained 20 yards, kicked a field goal. What if they actually started at their opponent's 25 (due to a blocked punt, for ex.) and did the same? Most would laud the other team's D as "winning" this drive by holding to a field goal, despite the fact that the team on offense gained 80% of available yards and scored 3 points.

It seems to me that a successful "bend but don't break" defensive scheme would normally have a relatively high DAY, which undermines the concept of using DAY to evaluate defensive quality. On the other hand, it seems like it would be somewhat useful in comparing teams with elite defenses that force a lot of 3-and-outs, or elite offenses that score a lot. But wouldn't ppd be better for that?

I think the problem may be that many advanced stats need to be used in the proper context, rather than in isolation. That's probably why the rating systems that use them generally combine a number of advanced stats into a formula to yield a single number for ranking purposes.

I agree with a lot of what you said. I don't think there is a single "perfect" stat, and it is why a combination of these stats are used to determine the "best" team in the algorithms.

Most would laud the other team's D as "winning" this drive by holding to a field goal, despite the fact that the team on offense gained 80% of available yards and scored 3 points.

In a vacuum the drive would be a failure, but I understand the fact that the drive doesn't exist in a vacuum. If a team drives from the 25-> opposing 30 they had to have made at least one first down along the way. At that point the defense did have to defend a drive in plus territory and succeeded. The entire drive was not a success, but once you are in a situation where a drive has gone into plus territory, you may as well celebrate giving up 3 vs 7.

It seems to me that a successful "bend but don't break" defensive scheme would normally have a relatively high DAY, which undermines the concept of using DAY to evaluate defensive quality. On the other hand, it seems like it would be somewhat useful in comparing teams with elite defenses that force a lot of 3-and-outs, or elite offenses that score a lot. But wouldn't ppd be better for that?

PPD is my favorite go-to as it captures the gist of what teams are trying to do. Even in my examples above though with the 2014 GT team, I put three pretty different games with different OAY, but all of them were right around 4.0 PPD. The Pitt game despite having better OAY and best PPD was probably the least impressive of the three games because the average starting field position was almost 20 yards better than the other two games. The ODE stat does its best to capture the scoring efficiency of the team with starting field position in the mix.

Regarding "It seems to me that a successful "bend but don't break" defensive scheme would normally have a relatively high DAY", I would posit there is no successful bend but don't break defense. Of the top 5 PPD defenses, only Ohio State is not n the top 5 of DAY (13th). Of the top 10 PPD, Oklahoma is the outlier with a 45th rank DAY. Their defensive turnover rate is at 20% and third rank. Oklahoma will either continue to live by the turnover, or we will see their defense begin to worsen is my guess.

 

[slugboy]

I recently found a site that tracked yards per point.

On the face of it, it looks like a silly stat. It’s not perfect, but it measures how much offensive effort is required to get points. If you have one yard per point, you’re getting into the end zone quickly (when you get in). If it’s 100 yards per point, you waste a lot of energy trying to score.

These aren’t perfect stats—there aren’t perfect stats—but there are stats that explain

 

[bobongo]

As you can see, there is very little to zero correlation between time of possession and offensive potency.

Just to be clear, I never said there was. I said there was a correlation between time of possession and % gained of "available yards".

 

[ibeattetris]

Just to be clear, I never said there was. I said there was a correlation between time of possession and % gained of "available yards".

He legitimately just showed you that time of possessions and OAY% have no correlation. Did you read the post? Only three of the top 10 in TOP are in the top 10 of OAY and the rest aren’t even close. The two stats don’t have much in common.

 

[ibeattetris]

I recently found a site that tracked yards per point.

On the face of it, it looks like a silly stat. It’s not perfect, but it measures how much offensive effort is required to get points. If you have one yard per point, you’re getting into the end zone quickly (when you get in). If it’s 100 yards per point, you waste a lot of energy trying to score.

These aren’t perfect stats—there aren’t perfect stats—but there are stats that explain

College Football Stats - College FB Team Yards per Point | TeamRankings.com

College Football yards per point, by team.

www.teamrankings.com

Yeah, it’s an interesting one. I haven’t been able to figure out how I would look at it in a vacuum, since starting field position affects it so much. I do wish there was a segmented version that did fg and td separately. It just seems incongruous to me that an offense could technically have a starting field position in plus territory, gain no yards, and kick fgs and blast away this stat. I realize this won’t happen, but it is why I would value yards per touchdown differently than raw yards per point (if all that makes sense).

 

[bobongo]

He legitimately just showed you that time of possessions and OAY% have no correlation. Did you read the post? Only three of the top 10 in TOP are in the top 10 of OAY and the rest aren’t even close. The two stats don’t have much in common.

I'm sorry, I didn't see OAY referenced in his post. I saw a ranking of TOP, YPG, and PPG.

 

[slugboy]

College Football Stats - College FB Team Yards per Point | TeamRankings.com

College Football yards per point, by team.

www.teamrankings.com

Yeah, it’s an interesting one. I haven’t been able to figure out how I would look at it in a vacuum, since starting field position affects it so much. I do wish there was a segmented version that did fg and td separately. It just seems incongruous to me that an offense could technically have a starting field position in plus territory, gain no yards, and kick fgs and blast away this stat. I realize this won’t happen, but it is why I would value yards per touchdown differently than raw yards per point (if all that makes sense).

If Miami hadn’t paired a boatload of yards to a handful of points, I never would have looked for it

 

[roadkill]

I agree with a lot of what you said. I don't think there is a single "perfect" stat, and it is why a combination of these stats are used to determine the "best" team in the algorithms.

In a vacuum the drive would be a failure, but I understand the fact that the drive doesn't exist in a vacuum. If a team drives from the 25-> opposing 30 they had to have made at least one first down along the way. At that point the defense did have to defend a drive in plus territory and succeeded. The entire drive was not a success, but once you are in a situation where a drive has gone into plus territory, you may as well celebrate giving up 3 vs 7.


PPD is my favorite go-to as it captures the gist of what teams are trying to do. Even in my examples above though with the 2014 GT team, I put three pretty different games with different OAY, but all of them were right around 4.0 PPD. The Pitt game despite having better OAY and best PPD was probably the least impressive of the three games because the average starting field position was almost 20 yards better than the other two games. The ODE stat does its best to capture the scoring efficiency of the team with starting field position in the mix.

Regarding "It seems to me that a successful "bend but don't break" defensive scheme would normally have a relatively high DAY", I would posit there is no successful bend but don't break defense. Of the top 5 PPD defenses, only Ohio State is not n the top 5 of DAY (13th). Of the top 10 PPD, Oklahoma is the outlier with a 45th rank DAY. Their defensive turnover rate is at 20% and third rank. Oklahoma will either continue to live by the turnover, or we will see their defense begin to worsen is my guess.

Thanks for your comprehensive response.

Your statement that there is no such thing as a successful bend-don’t-break defense is interesting and certainly seems correct for most of the top defenses. Like Oklahoma, we are living by turnovers for our wins. Is bend-don’t-break really a concept, or is it just an after-the-fact description of what happens when a relatively bad defense gets lucky with turnovers? I don’t know the answer, but if I were a DC and had a relatively slow, poor tackling set of players to work with, I would spend a lot of practice time focusing on getting turnovers.

 

[ibeattetris]

Thanks for your comprehensive response.

Your statement that there is no such thing as a successful bend-don’t-break defense is interesting and certainly seems correct for most of the top defenses. Like Oklahoma, we are living by turnovers for our wins. Is bend-don’t-break really a concept, or is it just an after-the-fact description of what happens when a relatively bad defense gets lucky with turnovers? I don’t know the answer, but if I were a DC and had a relatively slow, poor tackling set of players to work with, I would spend a lot of practice time focusing on getting turnovers.

Yeah, that was me inserting my opinion with regards to bend but don’t break. I think it has use. If you truly have abysmal DB’s there is really no reason to purposefully get torched. Just let the teams gain yards, and maybe someone makes a bad decision somewhere along the way (whether that be turnover or big penalty). I think if given the choice, I’d prefer to not need to rely on those things.

 

[takethepoints]

It has literally nothing to do with time of possession, and I have no idea how you even came to that conclusion. Case in point: we held the ball for 40 minutes in the Ole Miss game, yet they out gained us by 80 yards.

And BGU held the ball twice as long as Tech did and handed us our head in the process.

It is true that TOP isn't the be-all-end-all of football. However, as a general rule, you want to hold the ball longer then the other side. The best way to win is always to keep the other side's O off the field. Given the way the rules today favor O (well … they do), the best way to do that is to sit on the ball like a slug. Expecting the D to deliver on every drive doesn't compute.

 

[slugboy]

Yeah, that was me inserting my opinion with regards to bend but don’t break. I think it has use. If you truly have abysmal DB’s there is really no reason to purposefully get torched. Just let the teams gain yards, and maybe someone makes a bad decision somewhere along the way (whether that be turnover or big penalty). I think if given the choice, I’d prefer to not need to rely on those things.

G5 teams and B12 teams in the past decade, in conferences with wide open offenses, had defenses that seemed bent on takeaways. Try to keep the ball in front of you, try to pick the ball, try to give your offense another shot to score, and win 55-48. That’s a different kind of defense, but a valid strategy

 

[CEB]

PPD is my favorite go-to as it captures the gist of what teams are trying to do.

This

Call me a caveman simpleton, but the rest is just fodder for folks who want to make up stats to talk about. They are stats looking for an application.
I brazenly say even though I’ve read the entire thread…

Drives ending in TD > Drives ending in FG > drives ending in punt / turnover, no matter how long they take or how far you have to go.
Inverse is true on defense.
It’s hard to find a stat that correlates more closely with success.

 

[kittysniper101]

And BGU held the ball twice as long as Tech did and handed us our head in the process.

It is true that TOP isn't the be-all-end-all of football. However, as a general rule, you want to hold the ball longer then the other side. The best way to win is always to keep the other side's O off the field. Given the way the rules today favor O (well … they do), the best way to do that is to sit on the ball like a slug. Expecting the D to deliver on every drive doesn't compute.

While resting the D is a benefit, I wouldn't say the best way to win is to keep their O off the field. If anything, shortening the game and reducing possessions introduces more variation in the result. A single turnover becomes more influential. If you feel you have an advantage on the field you should press that advantage and maximize scoring efficiency. Explosiveness is even more strongly correlated with winning than success rate (efficiency). Even an efficient offense risks more variability if you prolong drives and introduce chance to be off field due to a failed 3rd or 4th down. https://www.footballstudyhall.com/2014/1/24/5337968/college-football-five-factors

 

[takethepoints]

While resting the D is a benefit, I wouldn't say the best way to win is to keep their O off the field. If anything, shortening the game and reducing possessions introduces more variation in the result. A single turnover becomes more influential. If you feel you have an advantage on the field you should press that advantage and maximize scoring efficiency. Explosiveness is even more strongly correlated with winning than success rate (efficiency). Even an efficient offense risks more variability if you prolong drives and introduce chance to be off field due to a failed 3rd or 4th down. https://www.footballstudyhall.com/2014/1/24/5337968/college-football-five-factors

I'll buy most of this. However, if your advantage derives from longer possessions, then you should indeed shorten the game, especially if your D is doubtful. A team I can't remember did this for over a decade. Rats! It's right on the top of my tongue! Give me a minute!

 

[slugboy]

While resting the D is a benefit, I wouldn't say the best way to win is to keep their O off the field. If anything, shortening the game and reducing possessions introduces more variation in the result. A single turnover becomes more influential. If you feel you have an advantage on the field you should press that advantage and maximize scoring efficiency. Explosiveness is even more strongly correlated with winning than success rate (efficiency). Even an efficient offense risks more variability if you prolong drives and introduce chance to be off field due to a failed 3rd or 4th down. https://www.footballstudyhall.com/2014/1/24/5337968/college-football-five-factors

One of the worst things ESPN has done over the past decade (aside from what they’ve done to the conferences) is to put that guy’s articles behind a paywall.

I think most football fans are better off having read that.

I also wish Cal had a Sonny Dykes now. There’s too much conformity in FBS today.

 

[ibeattetris]

This

Call me a caveman simpleton, but the rest is just fodder for folks who want to make up stats to talk about. They are stats looking for an application.
I brazenly say even though I’ve read the entire thread…

Drives ending in TD > Drives ending in FG > drives ending in punt / turnover, no matter how long they take or how far you have to go.
Inverse is true on defense.
It’s hard to find a stat that correlates more closely with success.

The one caveat for me is, if you were predicting a game and you had two teams playing each other and both have 4 OPPD and 1 DPPD, how do you predict a winner? At that point you need *something* else, whether than be some SOS based metric, some yardage based metric, a coin flip.

I only brought up our defenses DAY in my original post because in my eyes, it has been the main contributing factor to why our DPPD has been bad (rank 102). At the end of the day, I don’t actually care what our stats are if we win. I just proposed the stat that, in my opinion, would help in that endeavor. Obviously, if our offense starts scoring 7ppd, it doesn’t really matter what our defense does.