Filed under:

# Statistical In-Ferentz, Week 8: On the Horns of a Bielema

An interesting strategic situation arose in the Michigan State-Wisconsin game this past weekend, and considering that the Iowa-Indiana game was not particularly close, I thought I would look at the Bielema-D'Antonio chess match checkers match game of Yahtzee instead of the usual Hawkeye concerns. I'll use the NFL win probability framework to analyze these decisions.* The basic question is this: should Bret Bielema have called those timeouts to get the ball back in regulation, or should he have let Michigan State kneel on the ball and gone to overtime?  Here is the play-by-play of the game if you need to refresh your memory about the sequence of events.

* The usual caveats apply: 1) that the Advanced NFL Stats WP calculator I use derives from historical NFL stats and doesn't totally apply to college (which might be especially relevant in the last minutes of close games, when clock stoppages and the quality of kickers becomes so important) and 2) the generic model I'm using doesn't consider timeouts (going into the final drive, Michigan State had two, Wisconsin three). Also, sorry for the lateness of this post -- computer problems.

Decision Point #1: Game tied 31-31, Michigan State receives kickoff at own 22, 1:26 left.

The decision here was a familiar one for Iowa fans: try to score or go for overtime?  In the WP model, a team in Michigan State's situation has a 59% chance of winning the game. NFL overtime games are essentially 50/50 toss-ups, so the model suggests that teams with the ball in that situation do, on average, have a better chance of winning in regulation. That's the NFL, where a coin flip literally does significantly determine the outcome of overtime, but similar logic applies to the college. It's hard to know what chances Michigan State had of winning in overtime, but as long as their chances were less than 59%, going for the win in regulation was probably the better choice. Given how well Wisconsin had been playing to end the game (and how poorly Michigan State had been playing), it's hard to imagine Michigan State's chances of winning an overtime contest would have been much better than 50%, so going for the score in regulation was probably the better play.

Decision Point #2: Kirk Cousins sacked, fumbles, recovered by Michigan State at its own 24; Michigan State faces 2nd and 20 with 40 seconds left; Wisconsin calls time-out.

Of course, going for it in regulation has certain risks, such as your quarterback inexplicably holding the ball for 10 seconds while a defensive end bears down on him from the blindside. If you turn the ball over in your own territory, the game is essentially over. And it's possible that this kind of catastrophe happens more often when easily-rattled college quarterbacks/coaches are leading last-minute drives, not professionals. I still would recommend going for it, just because overtime is no guarantee, but the percentages could be much closer than the NFL model suggests, and there are situations where settling for overtime is advisable. But the interesting strategic question came after Michigan State recovered the fumble. Bret Bielema immediately called a timeout, hoping that Michigan State would recognize the futility of trying to continue the drive given such poor down and distance and would instead essentially kneel on the ball twice. Assuming each of Michigan State's running plays would take about 3-5 seconds, Wisconsin would then receive the punt with about 30-35 seconds left and a reasonable chance for Russell Wilson to drive them into field goal range (if Wisconsin didn't block the punt or return it for a touchdown, both of which they had almost done earlier in the game). It was a smart move by Bielema: at this point, with 2nd and 20, Sparty truly was behind the eight ball, with a WP of .47 in the generic model; an NFL team in their situation would lose more often than in overtime, suggesting that the opposition usually has a chance to get the ball back and score before overtime in these situations.

Decision Point #3: Kirk Cousins pass complete to B.J. Cunningham for 12 yards to Michigan State 36 with 30 seconds left; Wisconsin calls timeout.

The first timeout was a good move by Bielema, but it did force D'Antonio somewhat into a place of desperation. Now Michigan State knew that a conservative play wouldn't just send the game to overtime, it would give the ball back to Russell Wilson with time left to lead a final scoring drive. This knowledge effectively forced D'Antonio to be more courageous than he perhaps otherwise would have wanted to be, because the consequences of timidity were clear: give the ball back to Wilson and watch him win the game.

If Wisconsin had managed to force an incomplete pass on second down, they would have been in great shape, but they didn't -- Michigan State gained 12 yards -- and that had a surprisingly large effect on the strategy: it shifted the WP from .47 to .55 for MSU.  So that single play shifted the odds from Wisconsin's side to Michigan State's, and reflected the fact that it was now more likely that if anyone was going to score in regulation, it was probably going to be Michigan State. This is partially due to the manageable down and distance, but also, if I had to guess, has to do with the sheer lack of time left for Wisconsin to score. Assuming everything went according to plan and Michigan State threw an incomplete pass on third down, that would still mean that Wisconsin would get the ball back with something like 20 seconds left at its own 25-35 yard-line. Even with a timeout, it's just not that likely that Wilson could have led a drive into field goal range given those circumstances. So whereas the first timeout was a reasonable play, the second timeout was probably a mistake. There were just so many more ways that 3rd and 8 could lead to a Michigan State win rather than a Wisconsin win, especially when you consider that Michigan State still had two timeouts. It eventually wound up taking a Hail Mary for Michigan State to win, but that is one possibility you leave yourself open to by extending the game at that point.

If Bielema had not called the timeout, Michigan State would have been forced to make a very tough calculation about whether the risk of giving the ball back to Wisconsin was worth taking a shot to extend the drive. D'Antonio very well may have called a timeout himself, or let the clock tick down to zero, both of which would have been preferable from Wisconsin's side of things to using their own timeout.

Everyone knows how the script played out: Michigan State converted on third down, called a timeout of their own, ran three pretty inept and slow-developing plays, all incomplete passes, then connected on a 44-yard Hail Mary. The last play was not particularly likely, but we can't consider what happened after the initial decision point when evaluating Bielema's decision. When he made his decision, Michigan State was facing a very makeable 3rd and 8, only needed 30-35 yards to get in field goal range, and had two timeouts. It was not unthinkable that Michigan State might score from that position, while it was pretty improbable that Wisconsin would get the ball back with any time left to score. Bielema probably would have been better off just letting the game go to overtime, or at least forcing D'Antonio to use his own timeouts to continue the MSU drive. I'd say this was an example of Bret getting a little too cutesy and reaching for a gamble (Wilson getting the ball back and miraculously scoring in 15-20 seconds) that wasn't all that likely to pay off.

There is one other consideration that's not reflected in the model: if Wisconsin had good reason to think that their chances of winning in overtime would have been considerably better than 50%, then the bar would have been that much higher for any decision to extend regulation. Calling the first timeout might have given Wisconsin a slight edge, but did it really make their chances of winning better than, say, 60%?  I'm not saying that's exactly what Wisconsin's chances of winning in overtime would have been, but if Wisconsin thought they were the stronger team heading into overtime (which they sure looked like), then all their decisions should have been made with that fact in mind.

Alternate timeline

As I was watching the game unfold, I was dead sure that Bielema would take a gamble to try to win the game before overtime, but not the one he wound up taking. Specifically, I thought for sure that he would go for two after Wisconsin's final score. That kind of high-stakes, riverboat gambler risk was the kind I could see Bielema taking with relish. But would that have made any sense? Let's look at the alternatives in WP terms:

1) Kick extra point, kick off to MSU at their 25: .41 WP for Wisconsin

2) Go for 2, convert, kick off to MSU at their 25: .73 WP for Wisconsin

3a) Go for 2, fail, recover onside kick (20% of success on onside kick): .31WP for Wisconsin

3b) Go for 2, fail, don't recover onside kick (80% of failure on onside kick): .03WP for Wisconsin

So for the onside kick to make sense, the expected value of the "Go for 2" options would need to exceed .41 WP:

.73p + .086*(1-p) = .41

p = .61

(The .086 here is the weighted value of the "Go for 2, fail" options.) So the upshot is that Wisconsin would have been better going for two if they felt they could convert more than 61% of the time. That's a tough gamble to take, but for a team with Wisconsin's excellent offensive line, running backs and quarterback, getting two yards 61% of the time seems like a very reasonable bet. It's not a clear-cut mistake or anything (Bielema could have been quite reasonably assuming that Michigan State was dead-set on choking the game away, and that tying the game amounted to giving  Sparty more rope to complete the choke job), but it would have been a decent gamble to take the lead and force Michigan State into even more desperate straits.