Given some of the decisions this past weekend, it appears the Body Snatcherization of Kirk Ferentz is nearly complete. "Go for it on fourth and one? Why not?" "Fourth and seven? No problem." Sure, there was still a decision to run out the clock with a minute left and timeouts in his pocket, but a complete Les Miles personality transplant takes time, so maybe by the Nebraska game we'll see this happen. In this week's look at the statistical side of the game, I'll look at some of those bold choices, as well as the other single plays that turned out to be most crucial to the outcome in terms of changing the win probability of the game.
* Win probability is a figure that is used to estimate the chances that a team will win a football game given a certain game situation (the down, distance, yard marker, time remaining, and score). WP figures derived from the Advanced NFL Stats generic model. The usual caveats apply about an NFL model not totally applying to college football, etc. etc. (see previous versions of this post for more on WP, or look around the Advanced NFL Stats website).
1) Iowa ball, 3rd and 8 from the Iowa 26, 12:30 left in first quarter, 0-0; James Vandenberg completes pass to Keenan Davis for 44 yards to Michigan 30: +.10 WP for Iowa
This was a huge play for Iowa, not only because it gained so many yards, but also because it converted a long 3rd down play. How did it happen? It was a confluence of a soft zone call by Michigan and a shallow crossing route by Davis:
It looks like the Wolverines were in a quarter-quarter-half type zone in back with their linebackers dropping in zone underneath. The middle linebacker for Michigan drops a little too far back and is then out of position for the tackle when Davis catches the pass near the line of scrimmage. One interesting wrinkle is how Kevonte Martin-Manley's "route" effectively sets him up to block the outside linebacker after Davis makes the catch, suggesting that this was a read to Davis all the way. Add in some speed by Davis and a bad angle by a cornerback, and you have a 44-yard catch. This play was followed up by another pass play that was almost as significant, a 20-yard completion to Marvin McNutt that was worth +.08 WP.
2) Iowa ball, 4th and 1 from the Michigan 39 yard-line, 7:47 left in first quarter, 7-6 Iowa; Iowa fails to convert 4th down: -.04 WP for Iowa
This was an interesting sequence of events. On 3rd and 3, Marcus Coker ran the ball for what looked to be either a first down or inches away from it. Here's a screen shot of his forward progress before the referees waved the play dead:
Maybe the angle is wrong or the ESPN first down line is off a bit, but it looks like Coker is very, very close to a first down. So it's understandable that Kirk Ferentz thought a quick QB sneak would almost certainly be enough to get the first down. How hard is it to get a few inches, right?
Well, somehow between the end of the 3rd down play and the beginning of the 4th down play, a yard disappeared for Iowa:
Now when Vandenberg attempts the QB sneak and gains maybe a few inches, Iowa is short the first down by about a yard. I call shenanigans.
With all that said, Iowa has shown a decided affection for the quick QB sneak on 4th down, and I just don't know if it's justified. There's a certain gain from surprise, but there are also times where the decision is unnecessarily rushed and important information goes unnoticed, like, say, that the refs shorted you by about a yard, or that Michigan has left the entire left side of the offensive line uncovered. It also reflects a kind of pathology among coaches that one needs to call a play that gets exactly the number of yards required in short yardage conversions.* Defensive coaches, not being dummies, react by placing 9 or 10 men at the line, creating the bizarre situation where the defense is offering the offense 10 or 20 yard plays in exchange for stopping 1-yard plays. It's risky to throw down-field in these situations, of course (no one likes to see a pass flutter incomplete on 4th down), but the payoff for a completed pass is potentially huge. At the very least, it might pay to try some runs away from the massed concentration of the defense.
* This happened late in the Alabama-LSU game, where Alabama had a 3rd in 2 in LSU territory and obediently ran right into the LSU line for a one-yard loss. Alabama then punted and the game went to overtime, where they lost. Also important about the Alabama-LSU game: some more proof that kicking field goals in college is a sucker's game: the Crimson Tide missed four field goals, including one taken from the LSU two yard-line (two of the field goals were pretty much no-brainers, though, given the yardage needed).
One last note: even if the execution wasn't great, the call for the conversion was a good one. Iowa stood to gain +.08 WP with a conversion, and lose .04 if they failed to convert, while a punt to the Michigan 10 would have left the WP situation unchanged for Iowa. This translates into a break-even percentage of 33% to go for it; i.e. if Iowa felt they could get that one yard more than a third of the time, they should have gone for it.
3) Michigan ball, extra point try, 2:14 left in first quarter, 7-6 Iowa; Michigan fails to convert: +.03 WP for Iowa
This failed extra point was not a huge deal, but still cost Michigan a little bit. The fact that Michigan would have needed to convert a two-point conversion at the end of the game in order to tie goes back to this mistake.
4) Iowa ball, 4th and 7 from the Michigan 34, 9:00 left in second quarter, 7-6 Iowa; Vandenberg completes pass to Martin-Manley for 8 yards and 1st down: +.08 WP for Iowa
This was, from my perspective, the most surprising and delightful decision of the game by Kirk Ferentz and co. And actually, it still shocks me that Iowa took this gamble. The payoff was considerable (+.08 WP in the generic model) and the costs were not quite as high (-.06 WP). Also, the alternative (punting) was not that attractive (a punt to the Michigan 10 would have been worth -.01 WP). The break-even percentage to go for it rather than punt winds up being 35%, so this was a very reasonable bet by Iowa's coaches. The only other real option would have been to kick the field goal, but that would have been a slightly worse bet due to the lower payoff for making a field goal rather than keeping the drive alive. The break-even percentage for kicking a field goal versus punting works out about the same, though, so if Iowa's coaches had felt Michael Meyer could make that field goal more than about a third of the time, that would have been an equally reasonable bet. The key point, though, is that a punt in this situation is just not worth it, and even the worst case scenario -- turning the ball over at the opponent's 34 -- is really only slightly worse than punting, so why not take a chance? Iowa did here, and it worked out beautifully, allowing them to continue their drive and eventually score a touchdown.
Here's how the play broke down: Michigan blitzed two linebackers up the middle, but
James Ferentz and
Matt Tobin did an admirable job absorbing the extra men, leaving Vandenberg plenty of time to find Martin-Manley on an out route (KMM was facing man coverage thanks to the blitz). Martin-Manely got open easily, and Vandenberg makes a very accurate and
long throw to the sideline. In general, the pass blocking seemed much more secure this week than last, and Vandenberg made some good throws under pressure all game.
5) Michigan ball, 3rd and 10 from the Michigan 47, 4:07 left in second quarter, 14-6 Iowa; Denard Robinson fumbles, recovered by Tyler Nielsen at the Michigan 31: +.10 WP for Iowa.
Robinson really cost his team on a couple of plays in the second quarter, and this was the more inexplicable of the two. Credit Nielsen for applying pressure and Norm Parker for calling a blitz on an obvious passing down. The fumble marked a huge shift in field position and possession, and led directly to three Iowa points.
6) Iowa ball, 4th and 3 from the Michigan 24, 2:11 left in second quarter, 14-6 Iowa; Myer makes 42-yard field goal: +.02 WP for Iowa.
I highlight this play precisely because the impact on Iowa's chances was so limited. A kick in this situation gave Iowa a slight boost, but you have to balance that against the fact that even a 42-yard kick was no guarantee. If we assume that it was pretty likely -- say, 80% -- then the weighted value of trying the kick works out to be only +.004 WP. Meanwhile, converting on fourth down would have been worth +.06 WP and failing to convert would have been worth -.05 WP, leading to a break-even percentage on the 4th-down conversion of 49%. Could Iowa have made three yards half of the time? Maybe. They had, after all, just made eight yards on fourth down. Iowa's coaches were probably feeling like they had finished their gambling for the day, though, and were glad to take the (pretty) sure three points.
7) Michigan ball, 2nd and 8 from the Iowa 11, 1:02 left in second quarter, 17-6 Iowa; Robinson throws interception to Christian Kirksey, returned to Iowa 5: +.10 WP for Iowa.
This play followed the inexplicable semi-squib kick by
Trent Mossbrucker, which in itself cost Iowa .03 WP (if you assume that the alternative was a Michigan return to their own 30). The interception was another costly play by Robinson, although it was not as clear of a mistake as the fumble.
Micah Hyde may have gotten away with some pass interference on the Michigan receiver, and the throw was deflected into the air into the waiting arms of Kirksey. The play stalled a promising Michigan drive and left the half-time lead at 11. In the generic NFL model, a team in Iowa's position would have an 85% chance of winning the game.
8) Michigan ball, 3rd and 8 from the Michigan 39, 1:30 left in third quarter, 17-9 Iowa; Broderick Binns sacks Devin Gardner for 12-yard loss: +.05 WP for Iowa.
It was fun to compare the way Binns treated Robinson in the pocket and how he treated Gardner. With Robinson, Binns always stayed a respectful five yards away, hoping to both contain Robinson's dangerous rushing ability and swat away passes with his
Goro arms. With Gardner, Binns went right after him and planted him in the ground. Gardner assisted him by not throwing the ball away when he had the chance. The fact that Gardner was even playing is attributable to another sneakily significant play, the blitz by
Tanner Miller on the previous drive that temporarily knocked Robinson out of the game (and should have led to an interception). That doesn't really show up in WP terms, but Michigan's offense certainly didn't seem as dangerous with Gardner in charge.
9) Iowa ball, 3rd and 1 from the Michigan 43, 7:53 left in fourth quarter, 24-16 Iowa; Marcus Coker runs for no yards: -.02 WP for Iowa.
If Iowa had one consistent problem on offense, it was a tendency to lose yards on running plays in short yardage situations. Brandon Scherff, in particular, whiffed several times diving for blocks, often trying to slow down Michigan DT Mike Martin. On this play, if Iowa had converted, their chances would have jumped up to 98% from 94%. As it was, getting no yards here dropped Iowa's chances to 92% in the generic model. The subsequent penalty on James Ferentz dropped them another 1%, but Eric Guthrie's punt (and Shaun Prater's excellent save at the four yard-line) bumped the odds back up to 93%.
It's an interesting question whether Iowa should have even been trying for the first down on that 4th and 1. Converting would have increased the WP in the generic model to 99% from 92%, while failing would have dropped a team in Iowa's position to 88%. Also, a punt, even one as good as the one Iowa wound up getting, would only increase their chances by 1%. It works out that the break-even percentage to go for it would have been 45%. Getting one yard 45% of the time seems like a reasonable bet to make, especially when you consider that the game did wind up still in doubt on the final play. Converting here would have taken time off the clock and increased the chances of going up two scores. It's not open and shut, but given that Iowa's defense had started to flag by this point in the game, I can understand the inclination to finish the game right there.
10) Those four plays from the three for Michigan: ????
I put question marks here, because the Advanced NFL Stats model spits out some nonsensical results here. Basically, it tells us that when Michigan completed that pass to the Iowa 3 yard-line with 16 seconds left, their chances of winning dropped from 15% to 1% when. This seems a little unreasonable, so I'm assuming it's an artifact of the fact that there have just not been that many 8 point games at that yardage with that little time left on the clock. Still, it's not like Michigan's chances were excellent heading into that first down. There was very little time left, and a two-point conversion was still needed. If you put Michigan's chances of scoring at 75%, converting the two-point conversion at 50%, then winning in overtime at 50%, the Wolverines chances would have been 19% on first down. That may be a little low, but it's hard to see their chances as above 30% at that point in the game, given all the things that would have needed to break their way.
Of course, this means that if
B.J. Lowery was guilty of pass interference on that 4th and 3, that play could have been quite significant in WP terms. If you assume that Michigan had a WP heading into the play of something like 15%-20%, and that their chances of scoring on the final untimed down would been good (say 75%), then their chances of winning overall would improve to something like 30% (.75*.75*.5) the benefit of a potential PI call would have been a 10%-15% in their chances of winning. Another way to look at it is that a no-call dropped Michigan's chances of winning to zero from 15%-20%. And maybe the best way to look at it is to take into account the cost of the call
and the opportunity cost of the non-call -- that is, that getting the call might have increased their chances by 15% and not getting the call reduced their chances by 15%, so that single play was worth a hefty 30% swing in Iowa's favor. So I can understand why Michigan fans are upset about the call, but that just raises the question: was it pass interference? I didn't see anything egregious there, but I am of course biased. What's not in doubt is that the final play was indeed very crucial to the outcome of the game, given the leverage involved.
