Points/Drive vs. Power 5 (long post)

AE 87

Helluva Engineer
Messages
13,030
Okay, quick review, my points/drive (ppd) stat is artificial, calculating drives by number of rushing and passing TDs, FG atts, lost fumbles and interceptions, failed 4th dns, and punts and calculating points by crediting 7pts for every rushing or passing TD and 3pts for every made FG.

This calculation has the advantages of (1) not crediting offenses for defensive and special teams TDs and (2) normalizing for the difference between GT (averaged 9.9 drives/game) and Clemson (averaged 13.5 drives/game). The 128 FBS teams averaged 12.5 drives/game for the season. Since GT averaged 3.56 ppd, their 2.6 fewer drives/game per the average, results in a difference of 9pts/game.

This calculation has the weakness of potentially miscounting drives if a lost fumble occurs on defensive or special teams return or a drive ends in a safety. Both of these seem to me to be statistically insignificant over 120-200 drive season. It also shares with other (relatively) raw stats that it is not adjusted for opponents.

However, I did calculate the stats based only on games against Power 5 opponents. Of the 129 FBS teams, 125 had at least one game against a power 5 opponent. I then subtracted the P5 Dppd from the P5 Oppd for each of these 125 teams. It produced some interesting results.

A difference between conferences?
A built in bias exists since most P5 conference opponents will be from the same conference. Conference games cancel each other out. We're really comparing the 1-3 ooc P5 games a P5 team plays with the 1-3 played by a non P5 team against P5 opposition.

That being said, here are the average P5 ppd differentials for the Power 5 and non Power 5 conferences:
ACC, -0.03; Big 10, -0.10; Big 12, -0.06; Pac 12, 0.03; SEC 0.0
Am, -1.22; CUSA -1.67; MAC, -1.59; MEAC, -1.6; Sun, -1.89
The FBS Ind: 0.17

Over a 12 drive game, the differential for a P5 conference is basically 1 pt or less. By contrast, the differential for the non-P5 conferences would be 15 to 23 pts, over two TDs per game. It seems to me that this data suggests a pretty significant parity between the P5 conferences (though the search for parity in bowl-pairings and the low sample size might skew the data some), and a pretty significant difference between the P5 and non-P5 conferences. Marshall played no P5 teams last year fwiw.

Rankings
I also ranked the teams based on their games against P5 oppositions. Here, of course, the conference games don't cancel out since each team is being ranked individually. While there are some obvious outliers, I thought these results were pretty interesting.

upload_2015-6-19_14-52-10.png


Any thoughts?
 

Attachments

  • upload_2015-6-19_14-54-43.png
    upload_2015-6-19_14-54-43.png
    26 KB · Views: 13

Legal Jacket

Ramblin' Wreck
Messages
561
Not that long of a post. It's not really clear to me what Oppd - Dppd represents. Re Tech, is that our offensive ppd minus our defensive ppd? Or is that our offensive ppd minus the other teams average defensive ppd?
 

AE 87

Helluva Engineer
Messages
13,030
Not that long of a post. It's not really clear to me what Oppd - Dppd represents. Re Tech, is that our offensive ppd minus our defensive ppd? Or is that our offensive ppd minus the other teams average defensive ppd?

Yeah, it's each team's Offensive ppd minus their Defensive ppd. It's like a margin/drive stat I guess. It's how many more points/drive they earn than they give up.

I don't do anything that accounts for strength of schedule. Doing that for 128 teams seems to require more time/effort than I'm willing to spend.
 
Last edited:

Boomergump

Helluva Engineer
Featured Member
Messages
3,284
I like the breakdown. When you consider that our offense was historically good, in terms of efficiency, it shows what a little development in the defense would mean. Still though, out playing the P5 opposition by a point per drive is pretty impressive.
 

zhavenor

Jolly Good Fellow
Messages
468
Does espn box score still have the drives that happened in a game? If so "volunteers" could do a couple of college teams and add up the number of drives. I would do a couple. We would only need 64 volunteers and we would have the numbers for all D1 college football teams. If each volunteer would do 3... you get the picture. Let me know if you need some help.
 

AE 87

Helluva Engineer
Messages
13,030
Here's the Offense and Defense PPD top 25. There's a few more non-Power5 teams, especially in Defense. The smaller sample size probably matters more here.

upload_2015-6-19_17-12-26.png
 

GTNavyNuke

Helluva Engineer
Featured Member
Messages
10,075
Location
Williamsburg Virginia
Not to me. The Power Ranking does account for strength of schedule which should make it more accurate but could also introduce bias.

I think the way you did this was a good way to evaluate the strength of teams against their opponents. As you (I'm pretty sure) have pointed out there is no perfect way to reduce team capability down to a factor or a couple of factors without losing a lot.

I don't want to go all zen, but in the end numbers doesn't matter. (And I'm the one who just did 111 years of them.) We had a great team last year that did us all proud by the way they constantly got better. This year is a whole new rodeo. We have a stronger base and higher expectations. If our D gets a good bit better, we could well be at the top of your chart by the end of the year.
 

AE 87

Helluva Engineer
Messages
13,030
Does espn box score still have the drives that happened in a game? If so "volunteers" could do a couple of college teams and add up the number of drives. I would do a couple. We would only need 64 volunteers and we would have the numbers for all D1 college football teams. If each volunteer would do 3... you get the picture. Let me know if you need some help.

Good Idea, but the ESPN drive data can be wonky more often than you'd guess. I'm satisfied with my drive numbers for my purposes.

I guess it could affect the ranking of top Ds more than top Os because ppd is more sensitive to # of drives when there are fewer points, but fumbled returns and scrimmage safeties are so rare I'm not bothered. I also don't count end of half/game drives.
 

AE 87

Helluva Engineer
Messages
13,030
While ppd doesn't account for opposition, this data reveals another interesting datum. GT earned its Off ppd v P5 stat while playing against 4 of the top 19 P5 Ds by this measure, 3 of the top 10.
 

SecretAgentBuzz

Ramblin' Wreck
Messages
817
Location
ends of the earth
Two things jump out at me:

1) Man, our offense was good. I already knew that, but this is just more fuel for the fire. 3.52 points per drive? That is insanely good. If you were to balance this for quality of opponent, there is no doubt in my mind we had the #1 offense in college football last year.

2) In the original chart (Oppd-Dppd), there is a big drop-off after the first 9 teams. Those were pretty clearly the top 9 teams in the country, if you ask me. Even an average Dppd (2 maybe?) would have put us first overall.

I'm ready for 2015. Let's do this.
 

AE 87

Helluva Engineer
Messages
13,030
Two things jump out at me:

1) Man, our offense was good. I already knew that, but this is just more fuel for the fire. 3.52 points per drive? That is insanely good. If you were to balance this for quality of opponent, there is no doubt in my mind we had the #1 offense in college football last year.

2) In the original chart (Oppd-Dppd), there is a big drop-off after the first 9 teams. Those were pretty clearly the top 9 teams in the country, if you ask me. Even an average Dppd (2 maybe?) would have put us first overall.

I'm ready for 2015. Let's do this.

I think these are both right.

In 2014, 2ppd D vs P5 would've been #37 D, but that would still include a lot of non P5 teams capitalizing on 1 or 2 good games. Still, I think that's a reasonable goal. Duke allowed 1.98 ppd v P5 last year.

If our D had been as good as Duke's, we would've beaten UNC and FSU and maybe have made the playoffs. Our P5 ppd differential would have been 3.54, right there with O and tOSU.
 

AE 87

Helluva Engineer
Messages
13,030
Okay, I don't know when I'll have a chance to do defense and team differential. However, I thought I'd share some new data without starting a brand new thread.

Since 2008, 100 teams have averaged more than 2 BCS AQ or Pwr5 opponents per year. The other current 28 FBS teams obviously averaged fewer.

Of those 100 teams over the last 7 years, GT's offense has come in as #4 in raw Offensive efficiency, averaging 2.69 ppd (vs BCS AQ or P5).

Of the 92 teams who have played more than 8 BCS AQ or P5 teams over the last 4 years, GT's offense is #3, averaging 2.93 ppd.

Just to be clear, these stats reflect only games against BCS AQ (2008-2013) or Pwr 5 (2014) opposition.

The teams ahead of us over the last 4 years are Oregon 3.26 ppd and Baylor 3.22 ppd. Alabama came in right behind us at 2.90 ppd.

Obviously, again, these stats like any raw stats don't account for strength of schedule.
 

AE 87

Helluva Engineer
Messages
13,030
So, I used the same method described above to calculate the Differential Points Per Drive against BCS AQ teams, how many points they earn per drive minus how many points they allowed per drive. Its the teams average PPD margin against BCS AQ (2011-2013) and Pwr 5 (2014) opponents. I then averaged the PPD margins for the last four years and ranked the teams based on these averages

Assuming that I didn't make any mistakes in getting the data into my spread sheet, here are the results:
upload_2015-7-2_17-43-19.png


In 2013, we ended up outside of the top 25 at #26.
 

AE 87

Helluva Engineer
Messages
13,030
Here are the Offenses and Defense for the last 4 years (I let the ties at 25 overflow but didn't disambiguate ties within the tables).

upload_2015-7-2_18-46-9.png


upload_2015-7-2_18-46-33.png


Our Defense in PPD vs Pw5 (BCS AQ) was:
Def 2011: 2.42, #52
Def 2012: 2.34, #51
Def 2013: 2.20, #49
Def 2014: 2.46, #54
4yr Ave: 2.33, #51
 
Top