The Weekly Maths: Tightly Contested

Submitted by The Mathlete on

Let’s head straight to a revamped chart. Now fixed to time, as opposed to play, to give a better feel for the flow of the game.

image

What jumped out at me right away was how this game was played between 25% and 75% virtually the whole way. In fact, the first play run with either team have a 75% or greater win likelihood was Denard’s completion to The Threat. I combed through my database and Saturday’s game was the longest a game had stayed within that range in the last ten years. No other game had gone 59.5 minutes with neither team being closer to winning than being even. Of course as soon as Michigan’s odds dipped on Toussaint’s ill-advised reception, the offense comes through with a huge completion to set up a 65% chance of hitting the game winning field goal.

Biggest swing plays

Michigan would have been looking at about 70% win odds, but the 26 yards and a new set of downs on Sparty’s fake punt brought Michigan St back to square at 50%.

Andrew Maxwell had a third down and four on Michigan’s side of the field when he threw the ball straight to Jordan Kovacs. Prior to the snap Michigan was at its current low for the game around 39% but the pick and return quickly pushed them to about 53%. The number would have been about 5% higher if part of the return hadn’t been called back.

Michigan was down 1 with the ball at their own 25 with about 5 minutes left. Denard found some room and went 44 yards for Michigan’s longest play of the day. That jumped the game from 44% to 67% in Michigan’s favor.

With less than a minute to go Denard couldn’t find anyone open downfield so he chose to dump it off to Fitzgerald Toussaint a yard behind the line of scrimmage. The ball was low and Toussaint instinctively went down to catch it, which he unfortunately did. The loss of a down, yardage and time pushed the win percent down from 32% to 15%, the first time all game either side crossed the 75% mark.

Michigan would bounce right back and Denard’s strike to The Threat would reverse that 15% in no time. With only a field goal attempt left, the offense handed the game to Gibbons with a kick an average kicker would make 65% of the time.

And of course he did. +35% to Gibbons and all the brunette girls.

[Hit THE JUMP for an updated season projection, Dumb Punt of the Week, Nebraska prediction, and more.]

Win Percent Model

This game and a prompting from Seth got me to launch a beta version of my Win Percent Model to the public. There are still some small glitches on there but it should be fairly robust. I am still working to add an overtime portion to it and it still doesn’t recognize timeouts for the offensive team (it does for the defensive team if the opponent is trying to run out the clock). Feel free and hit me up on twitter if you see any issues or things that feel off.

Several people have asked if I am using anything from Advanced NFL Stats’ calculator and I just wanted to follow up that this is 100% unique research and is completely based on college data.

Season Projection

Michigan State was one of the three swing games left on the schedule. With the tight win, Michigan still stands at 8 wins as its most likely outcome, but the up-side definitely moved in their favor.

image

Remaining games against Minnesota, Northwestern and Iowa are allocated in the 80-90% percent range of Michigan winning. Strong favorites but definitely not sure things. Road games versus Nebraska and Ohio State are both in the underdog but winnable category of 30-40% win odds. I have the odds at winning out at about 1 in 14 and the odds of winning at least 9 at about 2 in 5.

Dumb Punt of the Week

Since this week’s article has been so Win Percent-centric, last week’s winner gave us the perfect example to review and decide if the conventional wisdom decision to punt was correct or dumb.

Randy Edsall’s Terrapins led NC State by 1 and took over near midfield with 2:40 left. After three consecutive Wes Brown runs for 3 yards and three NC State timeouts, Maryland was faced with 4th and 1 at the NC St 47 with 2:26 on the clock and the Wolfpack out of timeouts. A first down ends the game. A punt is the conventional choice and the one that Edsall made. Let’s go to the numbers:

Punt from opponent’s 47. Typically results in 1st and 10 at 18 yard line. 64% expected win.

Go and fail. Give NC State a first and 10 at their own 47. 42% expected win.

Go and succeed. NC State can’t stop the clock. 100% expected win.

The break even point is 38% success on fourth down. At this point it’s a neutral decision. In the fourth quarter of close games, fourth and ones are converted 74% when in non-goal line situations. The decision to go for it would be an 85% chance of winning, 21 percentage points better than punting. Randy Edsall is who he is and punted, NC State drove and hit a 43 yard field goal to win by two.

Randy Edsall is your back-to-back Ron Zook Memorial Dumb Punt of the Week winner. To the punters go the spoils.

Prediction

EV+, National Rank (leader), B1G Rank (leader)

Michigan Rush Offense

Michigan O: +2, 25th (Oregon), 4th (Nebraska)

vs

Nebraska D: +1, 53rd, 7th

Denard Robinson: +3, 5th (Manziel), 2nd (Miller)

Fitzgerald Toussaint: –2 125th (Murray, USF), 12th (Weisman)

Michigan Pass Offense

Michigan O: +2, 30th (Texas Tech), 2nd (Nebraska)

vs

Nebraska D: +0, 58th, 10th

Denard Robinson: +3, 24th (Doege), 2nd (Martinez)

Nebraska Rush Offense

Michigan D: +2, 40th (MSU), 6th (MSU)

vs

Nebraska O: +5, 3rd, 1st

Taylor Martinez: +3, 6th, 3rd

Ameer Abdullah: +1, 37th, 5th

Nebraska Pass Offense

Michigan D: +3, 28th (MSU), 5th (MSU)

vs

Nebraska O: +4, 18th, 1st

Taylor Martinez: +4, 18th, 1st

Kenny Bell: +5, 21st, 1st

Special Teams

Michigan: +0, 66th, 7th

Nebraska: +0, 53rd, 4th

The numbers are about square for this one with Nebraska’s home field giving them the bump. Michigan’s defense has been outstanding for the last five games but for the most part the offenses they have faced haven’t been spectacular which has limited how much the EV+ model will credit them. Saturday will be a big test. If Michigan’s defense can hold up against as an elite of an offense as the Big Ten has, the game should go Michigan’s way. My pick goes by the numbers but I still think Michigan has a great shot at breaking open the Legends division Saturday.

Nebraska 29 Michigan 25

Comments

UofM-StL

October 25th, 2012 at 2:17 PM ^

The link to the Win Percentage Calculator is broken. Looks like you left out the colon after "https" so it reads "https//docs.google" when it should be "https://docs.google"

That calculator is incredibly awesome by the way. I have a feeling that I'm going to be spending a lot of time on Saturdays compulsively entering data to see everyone's chances in every game I watch. Thanks for that.

UofM-StL

October 25th, 2012 at 2:34 PM ^

As I said above, super awesome. One oddity, since it's a Google Doc, multiple people can edit it at the same time, and that gets a little funky.

This seems like something that could be turned into a web form fairly easily with a little JavaScript, and for a tool as awesome as this I'd certainly offer my services to help set it up if you're interested.

Hannibal.

October 25th, 2012 at 2:55 PM ^

Every close game, I'm like "Wooo!  Big Denard run, we're going to win!" and then "Oh fuck, Denard threw a pick, we're going to lose!"  The win percentage graph totally validates my mood swings during these games. 

I'll be a bit disappointed if we don't score more than 25 against Nebraska, unless their defense has massively been overhauled and improved since they got torn to shreds by UCLA and Ohio State.  We might be slight underdogs, but I feel like the upside is in our favor here.  I could see us losing, but I could see us putting up 45 again too and blowing them out.  It would really surprise me if we lose this game by more than a TD. 

BK-bloo

October 25th, 2012 at 3:07 PM ^

i'd imagine a graph that tracks the WOOOOOOs and FUUUUUUUUCKs in realtime on the liveblog would also look pretty similar.

yeah, if we lose to nebraska it'll be close. it's still a weird feeling to feel good about our defense and completely unsure of what'll happen with the O.

UMaD

October 25th, 2012 at 2:52 PM ^

I'd love to see an MGoBlog Analytics exclusive, live, in-game, updated Michigan Win Probability.  Put it in the sidebar next year.  Way more in-game hits for Mgoblog, I'd imagine.

Across 110th St

October 25th, 2012 at 3:12 PM ^

Great stuff as always.  Many thanks.  A short question that can hopefully be given a relatively short answer; how do you think about/account for variance?  It seems like your model gives you point estimates for performace around various aspects of team and individual play - do you calculate variance around the estimates?  If so, do they tighten as the season progresses?

 

 

 

BK-bloo

October 25th, 2012 at 3:22 PM ^

it's really cool to be able to identify specific plays and their impact on the outcome probabilities. it would be sweet if the vertical lines that represent jumps or dips of 10% or more could be keyed to the UFR.  that could be a pretty big project, but would be an awesome melding of things that make this blog so great.

UPMichigan

October 25th, 2012 at 5:10 PM ^

Well I got $400 riding on Michigan winning on Saturday. I'm feeling pretty confident about that too. Nebraska's padded stats showed it's true colors when Ohio blew their doors off.

LewanHatesDonkeys

October 25th, 2012 at 5:36 PM ^

Is it possible to throw a marker somehow on the graph to kinda point out which big swing is which when you talk about them later?  I am capable of finding them but some kind of legend would help if its not crazy hard to do.