The Weekly Six

Submitted by The Mathlete on September 5th, 2013 at 11:11 AM

Last month I put together a new way of looking at down and distance and some new metrics. Like the four factors that have become prevalent in basketball, here is my shot at looking six factors for evaluating a football team. There are two key areas that aren’t included. Turnovers, which are critical to explaining past outcomes but poor at predicting future outcomes. The second is special teams. As Brian noted in the previews, special teams are funny because a lot of the value is derived by the presence or absence of big plays. Like turnovers, these are obviously key plays, but they make predicting future performance a challenge because they show up in very inconsistent ways.

The Six Factors


You could call him a factor (Fuller)

  Field Pos Early Conv Bonus Yds Avg 3rd Dist Adj 3rd Conv Red Zone
Offense 25.4 (19) 49.2% (39) 107 (51) 7.2 (59) +28% (4) 7.0 (1)
Defense 19.2 (55) 44.1% (30) 40 (10) 11.2 (2) -6% (37) 3.0 (9)

The first week is going to have some big outliers, with not a lot of competitive games, so the rankings should smooth out over the next month or so.

The defense was really outstanding against Central Michigan. There were almost no big plays, they put the Chips in awful third down situations and limited them both times the offense set them up in good position to enter the red zone.

The offense wasn’t really great in early downs but was exceptional in high leverage situations. Bonus yards (although possessions were limited) and first and second down plays were below NCAA average for the first week but third down and red zone performance was outstanding.


Field Position: A team’s expected points based on where a team started its drives

Early Conversion: The percentage of first downs' that are converted prior to a third down play

Bonus Yards: All yards gained after the first down marker

Average 3rd Down Distance: Average yards to go on third down

Adjusted 3rd Down Conversion: Rate of conversion for a team on third down, adjusted for the standard conversion rate based on yards to go, 0% is average

Red Zone: Points per red zone trip (TD’s counted as 7 regardless of PAT)

All categories except field position are based solely on plays in competitive situations (all first half plays and any second half plays where the drive begins or ends within two scores). Only games against FBS opponents are included, but after last week maybe I need to reconsider that.

Individual Game Scores

It’s just week one so we won’t kick in opponent adjustments  for another month. I also included all garbage time plays in the totals since there were so many new players getting touches in the second half.

Devin Gardner: +16.4 (12.2 in the first half)

Fitzgerald Toussaint: +3.1

Derrick Green: +4.4

Shane Morris: +2.2

Deveon Smith: –0.8

*All numbers are PAN, Point Above Normal, a representation of how many points a player adds or subtracts from the team’s final score as compared to an average player/team

If you’re on twitter or look at stats at all, you’ve probably experienced the NCAA stats overhaul that happened this year. In general, it’s awful, but now has auto-play video ads. From the play by play they have stripped out first names of players (so this week we won’t be able to differentiate between Cam and Thomas), no longer list the starters for a game, removed tackler information unless it is a sack, provide no description of a penalty, removed targeted receivers for incomplete passes and removed all yardage detail from punts and kickoffs. It’s pretty awful.

The one slight benefit is they did add directional information to plays. So we can look at how Michigan performed over each side of the line. When running here is how Michigan performed (scrambles and NORFLEEEEET removed):

Left: 5.9 YPC, +2.7, 14 plays

Middle: 3.2 YPC, +5.3, 18 plays

Right: 3.0 YPC, +1.4, 4 plays

Running left was Michigan’s most successful direction but all the benefit was in big play generation. Michigan yielded 50 yards on its two big runs to the left, but was under 3 yards per carry on all other plays to the left.

Ron Zook Dumb Punt of the Week

Dumb punts typically fall into two different types,

1. Punting on short yardage deep in opponent territory for “field position” reasons, even if the values to field position are highly debatable.

2. Punting while trailing in the fourth quarter when future possessions are highly limited

Group 1

There were 13 punts last week (excluding mercy punts with large leads, not that Hoke would care). 4 of the 13 came from current or future B1G members. Mark Dantonio and Michigan State are the only ones to do it twice. Dantonio wanted to start the year off right so before midway through the first quarter he had twice punted on short yardage from Western Michigan territory. The first time on 4th and 1 from the 41 and the second time on 4th and 3 from the 48. #B1G

Group 2

27 times teams punted in the fourth quarter within two scores of the lead. Georgia, New Mexico, California and Fresno State all punted from opponent territory with no more than 7 yards to go for a first down. Northern Illinois and Iowa combined for four 4th quarter punts while trailing or tied in their matchup.

And the winner is…Mark Richt and Georgia. The Bulldogs punted twice while trailing in the fourth quarter. The first was on 4th and 7 from the Clemson 40 with about 12 minutes left, trailing by 3. Clemson would take the ball 87 yards for a touchdown to go up by 10. On the next possession, Georgia was facing a tough 4th and 15 on their own 43 but now there were only 6 minutes left and they were down 2 scores, and Clemson had scored 38 points on them. Richt still chose to punt away. Georgia got the ball back down 10 with less than 2 and half minutes to play. They did get a quick score but failed to recover the onside kick.

Notes from around the NCAA

  • You may not have heard but Michigan State’s offense was kind of bad last week. With defensive touchdowns and other big turnovers, an average team would have scored 41 points given State’s field position. They were one of the worst offenses in the country in both early conversions (38%) and adjusted third down conversion rate (-15%).
  • Alabama won easily over Virginia Tech but the offense did not look strong. The Tide generated only 24 bonus yards, only four teams were worse.
  • Three games saw both teams generate over 200 yards past the first down line. Georgia-Clemson, Northwestern-Cal and Vandy-Ole Miss were some of the most exciting games of the week with big plays on both sides. Oklahoma versus Louisiana-Monroe was the  worst, with only 75 yards combined from both teams.
  • All the Texas A&M talk was around The Manziel “controversy” train but the A&M offense was amazing on early downs. It’s hard to be good at both converting early downs and limiting third down distance. The Aggies did both, tops in the country with 1.7 average yards to go on third down to go with a 59% early conversion rate.

Notre Dame 6 Factors

  Field Pos Early Conv Bonus Yds Avg 3rd Dist Adj 3rd Conv Red Zone
Offense 13.8 (68) 49.2% (39) 278 (2) 5.4 (26) +1% (37) 7.0 (1)
Defense 10.9 (11) 52.5% (60) 91 (32) 8.1 (19) +14% (75) 3.5 (18)

If you watched any of the Irish game on Saturday, it’s not too hard to see where things went right for their offense. 278 bonus yards was second only to Georgia for the week, and Notre Dame did it in only 7 drives. Notre Dame looked like they were playing NCAA Football on the peewee difficulty setting with the way they broke out the big plays.

With so many big plays and with the opponent being an overmatched Temple squad, I don’t know that we know a whole lot else about Notre Dame’s offense from Saturday’s results. On defense the Irish mostly held serve. The Owls were faced with limited field position, expected to score only 11 points based on field position for the game. Notre Dame allowed over 50% early conversions and was awful (75th out of 88 teams) on third down.

With both Michigan and Notre Dame putting up easy wins against overmatched opponents in week one, I’ll have to revert to preseason rankings for a prediction. Going into the season I had Notre Dame at #15 and Michigan ranked 17th, essentially tied. With the game at Michigan I think they will have a slight edge. If they can keep the turnovers and big plays even, I think it’s a clear advantage Michigan.

Michigan 24 Notre Dame 21



September 5th, 2013 at 2:13 PM ^

One to the plus side - they did run a successful fake punt in their own territory.  I think that was earlier in the game and believe they proceeded to march down the field and score a TD after the fake punt.


September 5th, 2013 at 3:48 PM ^

To be fair, if I had Dantonio's offense and his defense, I would have punted both of those as well.  The odds of converting were lower than low and even putting WMU back at the 20 was a better play then trying to convert a fourth down with an offensive line that can't block and receivers that can't run routes or catch the ball.


September 5th, 2013 at 8:52 PM ^

The math behind Romer theory necesarily simplifies the problem to the generic case by taking a shit-ton of data and averaging the hell out of it. But, real decisions are always specific. The concept presented in the Romer paper is fine but to really get down to dishing out awards, you need to factor in the strenghts and weaknesses of each unit of each team in the situation. More often than not (because math says so) the generic case is close enought to correct. Here though MSU has units on opposite end of their respective spectrums: awesome defense, inept offense. The correct Romer decision may very well have been to punt.

Judging by the final yardage totals, CMU's D was dominating MSU's O and vice versa. Defensive game...feels like a good punt to me.

In short, real decisions are hard.

Also, I have finally embraced my dorkdom. It's been a strange multi-decade journey but I'm happy to report that I have arrived.

The Mathlete

September 5th, 2013 at 9:02 PM ^

The flip side is that from my work, teams with great defenses should go for it more, not less. They are probably going to stop the other team either way, so take advantage of a potential short field for the offense and go for it. When you are on the opponent's side of the field, trading punts is almost always a net negative for the team in opp. territory.


September 5th, 2013 at 10:41 PM ^

I guess I could see that with a dominant dominant defense. MSU certainly fits that bill. In my first comment I overlooked that the fact that the punt decisions both occured in the first half and I was thinking of a late-game-with-the-lead situation. Maximizing expected value makes sense when you're behind or in point stocking mode but eventually focus needs to switch over to maximizing winning. So, yeah, MSU was or should have been in Acquire Points mode when they punted.

However, Chaos is a mofo badder than AIRBHG. If a coach thinks the defense's job is to minimize the likelihood that the opposing offense scores, then pushing them as far back as you can does make sense from a certain angle. Punting is a defensive concept. Dantonio is a defensive coach. Crazy stuff happens which is exactly what a WMU would need to score on a defense that has them outgunned by that much. Like how CMU scored on Michigan last week. A team with an offense as bad as MSU's is particularly susceptible to chaos. State's defense cannot do anything to help the offense and that must be a scary thought for them. So, given the data, the theory, and the realities of my team which no one really knows but me (i.e. Mark Dantonio of Clan Defense)...I wonder if decisions optimized in a mathematical vacuum would still make sense.

Check this out. Western only had two drives where they started beyond their own 30-yard line (WMU 40 and WMU 44); they scored TDs on both of those drives.The two punts in question occured on the WMU 41 and the WMU 48. Uh... MSU's badass defense went ofer from there because defense is hard even when you're great. Two more chances from primo field position translates to what, 8-ish points (7 - 10 in the real world)? Should Dantonio be counting on 14 points on the come from his defense at any point in a game? Um... those punts ended up being really crucial in retrospect.

Of course, the database is extensive enough that chaos is baked in at least to some degree or another but thats over the long term. Chaos rules in the short term. Saying  "at least I got the math right" won't console you when you loose, at least not enough.  And it certainly wont turn down the temperature on the hot seat.

Not trying to get in your face, just trying to walk the bridge from theory back to reality. Keep it coming, I'm a big fan.

steve sharik

September 5th, 2013 at 11:14 PM ^

"Turnovers, which are critical to explaining past outcomes but poor at predicting future outcomes. The second is special teams. As Brian noted in the previews, special teams are funny because a lot of the value is derived by the presence or absence of big plays. Like turnovers, these are obviously key plays, but they make predicting future performance a challenge because they show up in very inconsistent ways."

Doesn't this mean that football games between relatively evenly matched teams are extremely hard to predict, and predicting the point spread even harder?


September 6th, 2013 at 3:38 PM ^

Greetings; excellent post. I wanted to crosspost the Ron Zook Dumb Punt of the Week to, but the rest of the article wouldn't have as wide-ranging interest. Is there any way to get anchors (a name="") embedded in the posts so that we can link directly to particularly excellent sections?