at least it's not just us?
[Ed.: Bump. This makes sense to me: Michigan should mostly dump special teams once it gets across midfield.]
As Brian highlighted in the UMass round-up, maybe forgoing the punt altogether might not be such a bad decision. He noted my earlier look at the the topic and I wanted to pull it back and revisit and refine some of the work.
I looked at the years 2004-2009 and only looked at the top 20 rated offenses for each year. This study assumes that Michigan’s offense this year will be at a top 20 caliber and provides a broad enough definition of greatness that there is a good sample size. I did not distinguish what type of offense (Texas Tech Air Raid vs Georgia Tech triple option vs spread and shred) was used to get into the top 20. I will detail more assumptions as they are applicable along the way. In place of fourth down conversion percentages I used third down conversion percentage since the data pool is much larger and covers a wider variety of opponent levels. Since the thought process on a third down and fourth downs are roughly the same in most all (for now, anyway) situations, it seems reasonable to use the third down numbers.
Time for a you know what…
Assumptions: Top 20 offense, average defense, average punt game, average field goal kicker.
Based on these assumptions, except for long yardage, the punter should grab a seat once the offense crosses midfield. On your own side of the field the decision still makes sense starting around the 30 for shorter yardage situations and becomes more viable for longer yardage as you cross further down the field. Field goals become practical with 4+ yards to gain and only from about the 5-25 yard lines.
There are two big advantages a potent offense has that make 4th down tries more logical. The first is that they have more to gain by success. With a limited number of drives in a given game, why give them away for free? The second is that they are more likely to make them. Good offenses are more likely to be in better position on fourth down and more likely to make it. Here is a chart of great offenses fourth down conversions compared with all offenses. The right hand column was the one used for the above chart.
|To Go||All Teams||Great Off|
It’s not a huge advantage on any one given down, but Top 20 offenses convert the same opportunities about 2-3 percentage points more often than the average offense. Note: the rate of conversion for great offenses was much higher in the original analysis and is part of the reason the chart isn’t quite as go for it as the original.
But we don’t have an average <blank>
<blank> = Kicker
Let’s start with the kicking game, which is currently 5 points below average on the season and rated third worst in the country after the first three weeks.
Assumptions: Top 20 offense, average defense, average punt game, below average field goal kicker (FG make odds are reduced by 25% everywhere on the field).
The decisions near midfield obviously aren’t changed but now attempting a field goal on 4th and 5-9 from inside the 25 is no longer the most valuable option.
<blank> = Punter
I know it hasn’t been the most Zoltanic of starts for Will Hagerup, but at this point if he can hold onto the snap, there is no point in adjusting him to below average, even if he isn’t an advantage at this point.
<blank> = Defense
This is the one that seems a bit counterintuitive and Brian and I disagree on. I say that the strength or weakness of your defense is irrelevant to your offensive decision on whether or not try a fourth down conversion. My belief that it is irrelevant is based on this chart.
Great defense obviously give up fewer points than bad defenses but the key point is that the difference between a great defense and a bad defense is consistent up and down the field. Giving the opponent a first down at midfield isn’t a guarantee of a touchdown even with a bad defense and isn’t a guarantee that pinning an opponent deep against a great defense will keep the other team off the board. In fact, the gap between the two is about .25 points per first and 10 all the way from the 1 to the 90. If this is true, then the ability of the defense is irrelevant to the offense’s decision to go for it. For that to be the case, there would have to be evidence that the difference between a good defense and a bad defense changes at different points on the field.
So what does all this mean
If Michigan can maintain their feverish offensive pace this year and fail to find an adequate kicker, I think their decision set in all but late game score specific situations should look something like this:
As I noted previously, if you buy into this mentality, it opens up another opportunity, changing your early down play calling. If your four down strategy has changed, so should your down by down playcalling. It may become more viable to risk a wasted down with deep ball knowing that you have an extra, or it might just make sense to keep the ball short in the air and on the ground knowing that over four plays instead of three the likelihood of getting the yardages greatly increases so play to have the shortest possible fourth down attempt if you don’t convert before that.
Right now the database appears to be overvaluing dominant wins against bad teams. This will change in the next two weeks as there is enough interplay to start doing better adjustments for team strength based on this season’s games.
There is obviously a lot of noise in here still but I want to keep the numbers clear of human intervention to see how they straighten out as the season goes on. As a reminder, the In Season rating is made up of one quarter of the rating from each game against a FBS opponent (or loss against a FCS opponent) and the remainder is pre season rating. A team with three FBS opponents is 75% in season, 25% pre season. A team like Indiana who has only played one FBS opponent to date, is 25% in season and 75% pre season. Michigan is 50/50 right now. Game success is adjusted for strength of opponent based on the pre-season PAN number.
|Rank||Team||Conf||Preseason PAN||In Season PAN|
|2||Oklahoma St||Big XII||5.0||17.9|
|5||Ohio St||Big Ten||11.8||16.0|
|14||Texas Tech||Big XII||7.0||11.3|
|15||W Virginia||Big East||8.2||10.9|
|17||Air Force||Mtn West||0.4||10.7|
|23||Texas A&M||Big XII||(0.6)||7.7|
- Michigan holds on at #25. There change is a reflection of the movement of other teams. The UMass game had no factor in the final calculation although it would have if we had lost.
- Oklahoma St’s high powered offense and multitude of weak opponents has them at a way too high #2 ranking.
- A lot of conference hodge podge right now. No ACC teams, only W Virginia from the Big East. Michigan and Ohio St only two from the Big 10 after Iowa’s poor showing in Arizona. A lot of Big 12, SEC and PAC 10 in the ratings right now along with 5 teams from the future/past Mountain West conference.
- Michigan Projection
- Since Michigan scraped by UMass for the win, their is no knock on Michigan for their performance. The only adjustments are for changes to their remaining opponents. The projection is largely unchanged but still hovers around 8.5 wins.
Bowling Green – rank 84, 97% chance of Michigan win
@ Indiana –67th, 67%
Michigan State –40th, 66%
Iowa –39th, 65%
@ Penn State –31st, 43%
Illinois –65th, 82%
@ Purdue –43rd, 52%
Wisconsin –26th, 59%
@ Ohio State –5th, 16%
Projected Big 10 finish
The distinctions between Penn St and Purdue are very slight as the six teams between are projected between 4.2 and 5.1 wins. Right now it looks like a solid #1 (Ohio St), four bad teams (Illinois, NW, Indiana and Minnesota) with everyone else lumped in the middle. Michigan would look much better if not for their patsy free Big 10 schedule. Michigan only plays 2 of the bottom four while everyone else in the middle group plays at least 3 and Penn St and Purdue get all four.
Through the first two weeks of the season, here is what my ballot looks like.
|Rank||Team||Conf||Preseason PAN||In Season PAN|
|4||Ohio St||Big Ten||11.8||13.6|
|9||Oklahoma St||Big XII||5.0||10.9|
|13||Texas Tech||Big XII||7.0||9.8|
|18||W Virginia||Big East||8.2||8.4|
Other Big 10 Teams
30 Notre Dame
34 Michigan St
36 Penn St
This is a purely data driven poll, there is no human intervention or manipulation. PAN stands for Points Above Normal. It equates directly to points scored Oregon at +20 PAN is 20 points better than the average team and about 12.5 points better than Michigan (+7.5 PAN) is projected at this time. The normal that it used is the average of the 120 FBS/1-A schools. For BCS conferences like the Big Ten, about 3/4 of the conference is usually above Normal.
Games against FCS/1-AA schools are not included in the calculation unless the FCS team wins, in which case a game score of –30 PAN is applied to the losing FBS team.
The pre-season metric is calculated based on team PAN’s for previous years. The In-season metric is calculated with each game played so far this year worth 25% and the remainder made up of the preseason number. Michigan is 50% actual and 50% preseason since they have played two qualifying games. Purdue is 75% preseason and 25% actual since they have only played one FBS opponent.
I will use this methodology for the first month of so of the season until all teams have played at least two FBS games to do a proper in season accounting of strength of schedule. Until then opponent strength will be accounted for based solely on the preseason rating.
UMass: 100% win likelihood
Bowling Green: 100%
At Indiana: 80%
Michigan St: 66%
At Penn St: 49%
At Purdue: 59%
At Ohio St: 23%
Putting the rest of the season together, yields an average win total of 8.7 wins which of course is impossible. Running a Monte Carlo simulation on these odds yields the following spread:
Nine wins (5-3 in conference) is the most likely scenario with about a 75% likelihood of 8-10 wins. This picture will change over the course of the next 2-3 weeks as we get more interplay between FBS teams and get a better gauge of team strength, but based on performance to date Michigan is obviously grading out very well.
A lot of interesting notes from a game I am still replaying in my head.
Another ground game worth 12 PAN, just like last week. My database goes back to the 2003 season and during that time there have been a total of 107 games where a player has recorded a PAN of 12 or higher. Of those 107 times, there are 10 players who have done it at least twice (4 have done it three times). The only players to have put up a dozen on the ground twice in one season versus BCS teams, Denard and two others, Jerome Harrison at Washington State vs Stanford and UCLA in 2005 and Chris Barclay at Wake Forest vs Clemson and Maryland in 2003.
The passing was obviously hit or miss for Denard on Saturday. The opening quarter started very strong with a +5 PAN to start things off. From the start of the second quarter until the final drive, the rating plummeted to a –8 during that stretch only to be brought back with a +4 through the air on the final drive.
The Special Teams
Will Hagerup did not have a beautiful looking day as many of his punts looked like shanks or line drives but his day not as ineffective as it appeared. He had five punts under 40 yards on the day, but only one of them was truly damaging. Two were fair caught inside the 15. Two of the others went out at the 20 and 24 which, although painful to watch, were about the effect as his 49 yard kick through the end zone. On top of that, he only allowed 9 yards of return on the day, all on his first kick which still netted 40 yards. All told, his punting performance had a PAN of –0.2, which came back to zero when adding in Denard’s quick kick. That said, his PAZ (Points Above Zoltan) was about –100.
Kicker: the sample size is small but the start could not be much worse for Michigan. So far this year Gibbons is –5 PAN, the second worst of college football team. Only Toledo’s –7 is worse.
Who knows what’s up with this group. It’s hard to say right now. Smith and Shaw saw their qualifying carries cut from 29 against UConn to 12 against Notre Dame. Against UConn both went +2 PAN while against ND Smith went –1 and Shaw –3. This weekend saw only four carries go for positive value (not positive yards) and one of those was the sole Hopkins carry.
This game is a very difficult game judge the defense on. There were a lot of really great moments (3 interceptions!) and a lot of really bad moments. Based on Notre Dame’s starting field position, they would be expected to score 25 points on the game, 27 if you count the two end of half drives. In that regard Michigan’s defense was slightly above average and if you add in the three interceptions there were some things to hang their hat on.
The two big pass plays were obviously killers, costing Michigan 12 points, as did the QB switch. ND was –9 PAN without Crist in the game and +16 when he was in, with most of the value coming from the two big pass plays.
The three interceptions helped Michigan dominate the starting field position. Michigan’s starting position put them in line for 32 points on the game, 7 more than ND. This was the 11th largest spread of field position for the weekend (OSU was first with +15 against Miami (YTM)).
There are two main reasons Denard and the offense were able to put up such large PAN numbers but not outscore their expected points based on field position. The first is the two missed field goals which deflated the total points scored despite the offense getting them into position. The second is penalties, offensively Michigan’s five penalties cost the team a full touchdown and in each of the five occurrences helped prevent any further first downs on the drive.
On defense Michigan’s penalties helped make things a little too dangerous at the end of the half, but Michigan still managed to break even on the day on the defensive side of the ball.
It took absolutely zero numbers to know that Denard was very very good against UConn Saturday. We could easily just leave the conversation there, but then I wouldn’t get to use my handy database this week, and really, what fun would that be.
First games are very hard to put into perspective right away. We really have no idea if we played against the future Big East Champions or a mediocre team from a mediocre conference. To solve this, I took two independent approaches. The first was to just strip out the strength of opponent factor and compare it to other performances regardless of opponent. The second way is to add the games from this weekend into the database as if they were a part of last season and use the 2009 adjustment factor.
Not to spoil all the fun but if you don’t want to read on, I’ll make this real easy. Those of you who follow me on twitter already saw that Saturday was the best performance by Michigan quarterback that I have on record, which is almost every game since 2003.
On The Ground: +12 PAN
Without adjusting for competition for anyone, Robinson’s day on the ground was the best rushing game by a Michigan player in my database. It was .01, basically tied with, ahead of Mike Hart’s 2004 game against Illinois when we went for 231 yards on 39 carries.
It was the third best rushing performance of anyone under Rodriguez as a head coach. Only Pat White’s +14 against Pitt in 2005 and Kay-Jay Harris’s +16 against East Carolina in 2004 were better.
It was the third best ground performance by a Big Ten player, the 27th best of any quarterback and the 90th best overall, including many performances against teams much weaker than UConn.
Through The Air: +9 PAN
+9 through the air is a very good day by itself, even without the spectacular +12 on the ground. In fact, the unadjusted +9 is 15th among Michigan passers. Ahead of him are 10 Henne games, 2 Navarre games and Forcier’s first two games of last year.
In his three years under Rodriguez, Pat White had four +9 performances, as did Rasheed Marshall.
Putting it All Together: +21 PAN
Looking at the total performance, Dernard cracks the top 200 performances list that is dominated by heavy passers. When accounting for competition, Denard still comes up 7/11 PAN (7 passing, 11 rushing). Only three BCS conference players had ever done that before Denard on Saturday. Brad Smith and Jake Locker each did it once while Vince Young managed the feat three times.
Getting First Downs
Maybe the most impressive thing about Denard’s day was his ability to get first downs, especially on third down. Excluding the final run out the clock drive, Robinson was good for 12 first downs rushing. Only two QB’s since 2003 have ever topped that and 12 is good for top 50 including running backs.
On third down, Robinson’s seven rushing conversions were the fourth most since 2003. His 11 overall conversions were top 20.
A look at Michigan’s opening opponent through the eyes of PAN*.
When Michigan Rushes
Let’s kick the season off with a nice chart, Michigan Rush Offense PAN vs. UConn
Last year the gap between the two was worth nearly two points a game and this year it is projecting to narrow slightly. This projection is probably on the pessimistic side for Michigan as UConn has four consecutive years of decline on rush defense and nothing would indicate that Michigan would see a drop versus last season’s performance on the ground.
Since Rodriguez had experience against UConn while at West Virginia, those matchups provide another, better data point of comparison. In four games from 2004-2007 West Virginia averaged 6 PAN/game offensively and UConn averaged 0 PAN/game defensively. In other words, West Virginia’s ground game average 6 points per game more than the average team that played UConn and the Huskies defended the Mountaineers about on par with the average team.
Based on both West Virginia and Michigan experience, the numbers indicate that Michigan should have an opportunity to do some damage on the ground on Saturday.
When Michigan Passes
Michigan was pretty average passing the ball last year but UConn wasn’t great at covering the pass. The historical numbers are a bit all over the map, the Huskies had a 10 point negative swing from 2008 to 2009.
UConn returns a lot of their defense from last year but the one position group that will be replacing players is the secondary. In 2009 the team had to deal with the midseason murder of starting cornerback Jasper Howard, putting a little perspective on the mostly on-field issues Michigan’s secondary has faced. Of the top 11 UConn players in points taken last year, the only three not returning this year are cornerback Robert McClain, 25 PT, 2nd on team and first among DBs, DE Lindsey Witten, 20 PT, 4th on team and first among DL and S Robert Vaughn, 15 PT and 2nd among DBs.
With the year to year variance these two teams have shown in passing and defending the pass, it is difficult to tell who will pick up the advantage when Michigan puts the ball in the air.
When UConn Runs
Michigan saw their first dip into negative PAN against the run last year, while UConn is coming off back to back strong seasons on the ground.
The UConn running back situation is one where PAN sheds an interesting light that is hidden by tradition stats. Last year UConn split the carries almost evenly between Jordan Todman and Andre Dixon (235 vs 239). Todman ran for 1188 yards and 14 TDs while Dixon had 1093 yards and 14 TDs as well. Despite those very similar stat lines, Todman’s performance was worth 16 points and Dixon’s nearly offset the gains with –15 points.
Unfortunately for Michigan Todman is back and Dixon is gone. The historical trend indicates that Michigan should have the advantage, but with a quality back in Todman returning, Michigan will need a much improved defense performance to limit the UConn rushing attack.
When UConn Passes
After a dreadful stretch through the air in 2005-2008, UConn bounced back last year with their best showing in five years.
UConn has two QB’s with starting experience coming back. Cody Endres who took over in mid-season after an injury, was a modest 1.1 PAN whereas this year’s starter Zach Frazer was a worse –1.5 PAN in action at the beginning and end of the season. Frazer posted a similar –1.6 in 4 games in 2008.
Despite the higher value, Frazer beat out Endres again for the job this season and Endres went on to get suspended for the opener, leaving UConn with the sole experienced QB for Michigan. Unfortunately, Michigan’s secondary will make this matchup interesting, but at least the Huskies are able to trot out a world beater at QB even if he does have 2 years of experience.
History in Openers
When factoring in quality of opponent, Michigan best two games of the Rodriguez era have been the openers. 2008 felt very disappointing at the time, but taking an eventually undefeated and Alabama crushing Utah team to the wire, was the best performance of the season. 2009 saw a much much weaker opponent in Western Michigan, but the utter dismantling Michigan displayed made the 2009 the highest rated game Rodriguez had at Michigan to date. Success in openers had been the norm for Rodriguez at West Virginia. 3 of his last 4 were double digit PAN and two were over 20.
UConn’s sample size is much smaller. 3 of the last 7 years they have opened with 1AA opponents and the four years have seen performance within 5 points or so of average.
Head to Head
In the last four meetings Rodriguez and West Virginia owned UConn. West Virginia average a PAN of 13 while UConn came in at –5 PAN. Even after giving the Huskies a break for how good West Virginia was for several years, they still did worse than average against them.
The 2007 game is a bit of anomaly on this chart. It looks like UConn outplayed West Virginia but the Mountaineers completely dominated the Huskies in the game. The PAN is off because two first half fumbles by UConn meant the offense didn’t have to do much heavy lifting to build a 17 point lead after the first drive of the second half. A 17 point lead means that the plays stop counting towards the PAN, but WVU just kept going. To the tune of nearly 400 yards, 29 PAN all after they already had a 17 point lead. So in other words, 2007 looks like a good performance by UConn, but in reality a couple fluke plays got them in a hole and once they were there, West Virginia buried them.
The All In Look
The history is on Michigan’s side, the two year trend is on Michigan’s side, the strength in openers is on Michigan’s side, the head to head coaching matchup is on Michigan’s side and with homefield, I have Michigan pegged at about a touchdown favorite with about a 75% chance of starting the year off in the win column.
*PAN is calculated by assigning every play a value based on how much the play helped or hurt the offense’s chances of scoring. Every down, distance and line of scrimmage combination is assigned an expected value, the average points scored across college football in that same situation. If a play increases the expected value, the respective teams and players are credited with the amount of increase.
All plays are then adjusted based on strength of opponent. Plays against weak opponents are penalized and downgraded while plays against strong opponents are bumped to reflect the degree of difficulty.
Only games against FBS (D1A) opponents, games against FCS (1AA) opponents are non-existent in any numbers used in this work.
Qualifying Plays (QP) are all plays in the first half and plays in the second half when the game is within two touchdowns. End of half run out the clock drives are also excluded.