Miami at Virginia Statistical Preview: Trending Upward?

Peter Casey-US PRESSWIRE

This post uses a stat-based analysis to predict the final score of the Miami game, while also attempting to account for Virginia's recent improvement.

I last walked out of Scott Stadium on November 10, 2007 after an unfortunate loss to the Hokies. I strayed far from Grounds after graduation, and in the four and a half years since, have maintained what connection I could with the University and its sports teams. A fascination with sports statistics and a fortunate transition to the DC area have kept me more engaged with the Hoos over the past year. Still, it's been a long time since I watched The Adventures of Cavman, sang the Good Old Song1, or loudly lamented the decisions of [insert Virginia coach and / or coordinator here] from the student section. This weekend, I'm going back.

A decisive victory over NC State has roused a fan base aching for some positive signs.2 To be completely honest, as recently as a week ago, I wasn't too excited about the prospect of witnessing live one of the performances that I had the luxury of turning off at home. But against NC State, the defense forced five turnovers. The offense converted a yards advantage into a points advantage.3 The rotating QB system actually worked.4 My pessimism is subsiding5 just in time for my long-awaited return to Grounds.

Awash in wistful Charlottesville nostalgia, this post will attempt to predict the outcome when the Miami Hurricanes blow into town on Saturday.6 I'll start with a tried and marginally successful method of predicting the final point difference, modifying this result by attempting to project the team's post-bye improvement. Has the Chick-fil-A Bowl hangover finally ended? Will the team's apparent turnaround be enough to secure a second ACC win?

[Sagarin Rankings and Point Difference]

Earlier this season I devised a method for predicting a final score margin based on Sagarin rankings and past results.7 I fit a line to the Sagarin rank difference versus point difference data for a given team's 2012 games. From the perspective of Team A, a high positive rank difference with respect to Team B should give a large negative point difference,8 while a small negative rank difference should yield a small positive point difference. A rank difference in an upcoming game can then be input into the model to predict a point margin. These graphs for Virginia and Miami appear below.

Graph1_medium

The Miami portion of the plot clusters rather closely around the fit line, and if you notice the origin of the graph, the Hurricanes have beaten every team they outrank while losing to every team they're ranked below.9 Sagarin ranks Virginia at 92 and Miami at 55. From Miami's perspective, these ranks suggest a 10 point win.

Virginia's 2012 inconsistencies have made the blue portion look more like splattered paint than a set of data points. The Hoos almost need to be ranked 25 spots ahead of their opponent to win, but would only lose by 10 when ranked 50 spots lower. From Virginia's odd data, the respective team ranks suggest an 8 point loss.

The averaged Sagarin rank method thus predicts a 9 point Miami win.10 The lack of Virginia trend that could be skewing the overall result, though, is thanks in large part to last weekend's significantly improved performance against NC State. So I thought, could the predicted point margin be improved by quantifying the recent performance trends of each team?

[The Sagarin Point Gap Over Time]

I used the point difference equations for each team to predict the margins of their previous 2012 games. The difference between the predicted and actual point difference then gives a "Sagarin Point Gap (SPG)," or a notion of the team performance relative to the Sagarin prediction over time.11 Charts depicting the Sagarin Point Gap for each team appear below.

Chart2_medium

Chart3_medium

The chart data is plotted in the following graph. To establish recent trends, I used performances over the last five games, or roughly the more recent half of the season.

Graph4_medium

First, the good news. After the Hoos bottomed out against Duke in game 6, the still-bad Maryland and Wake Forest games were actually an improvement of nearly 20 points relative to the Sagarin prediction. The huge NC State win pulls the recent improvement trend even higher. The fit line, with an input of game 10, outputs a predicted +24 SPG against Miami.12 This suggests that the Hoos would defeat Miami by 16 if the Hurricanes were treading water in recent games ...

Now, the bad news. In the slopes of the lines, we can see that the Hurricanes are improving at only a slightly slower rate than the Hoos. Miami has been at least +8 over the Sagarin prediction in each of their last 3 games; the equation predicts Miami at +22 SPG on Saturday.

[Prediction]

So although the Hoos have shown dramatic improvement, so have the Hurricanes, and Virginia only gains two points in the recent performance analysis. This gives a final point margin prediction of Miami by 7.

To predict the actual points scored, I'm just going to use reason, since tracing numbers in the past hasn't been particularly accurate.13 Football Outsiders ranks the Miami offense somewhere around Duke, Penn State, and NC State. In other words, 42, 17, and 6 points scored against the Hoos defense. So that tells us basically nothing. The Outsiders rank the Miami defense worse than Maryland, but better than Wake Forest and NC State. I'll predict the Hoos come back down to earth from the NC State output and score 21.

FINAL SCORE: Miami 28 - 21 Virginia

Win or lose, I'll be at Scott Stadium on Saturday with a cheered heart and warmed blood14,15 shouting and roaring for the Hoos. It's a football weekend.

_____________________________________________________

1 at least within the confines of Grounds; there was a spirited rendition at my wedding reception.

2 enough to shift the betting from Miami by 3 to Virginia by 1.5

3 now sort of starting to doubt that this game actually happened.

4 ok, now this is just getting ridiculous

5 drinking the orange and blue Kool-Aid, seeing the world through orange and blue glasses; whichever you prefer.

6 see what I did there?

7 back on my old website, which will be imported here.

8 for example: if Team A was ranked 50 and Team B was ranked 10, (Team A - Team B) = +40. we would then expect Team A to lose, possibly by 30-10, or -20 points (Team A 10 - Team B 30).

9 go Sagarin!

10 a note on the aforementioned marginal success of this method: in some form, it correctly predicted the winner 5/6 games, and picked correctly vs. the spread 3/6 times. One of the "wrong" spread picks was TCU by 17 when the spread was 17.5. They won by 20. The method correctly predicted a 1 point win vs. Penn State when the spread was Hoos by 10.

11 a method reminiscent of the Standardized London Performance Gap in my last post

12 this seemed really high at first, but consider that the NC State game was +33. We would probably expect some drop-off from that level, but the team has undoubtedly been improving. Maybe +24 isn't that crazy.

13 and the margin is much more important anyway

14 warmed either by a jacket or the liquid equivalent thereof

15 on a related note, if someone could go ahead and tell Dmitri to start lining up the Gus Burgers, that'd be great

X
Log In Sign Up

forgot?
Log In Sign Up

Forgot password?

We'll email you a reset link.

If you signed up using a 3rd party account like Facebook or Twitter, please login with it instead.

Forgot password?

Try another email?

Almost done,

By becoming a registered user, you are also agreeing to our Terms and confirming that you have read our Privacy Policy.

Join Streaking The Lawn

You must be a member of Streaking The Lawn to participate.

We have our own Community Guidelines at Streaking The Lawn. You should read them.

Join Streaking The Lawn

You must be a member of Streaking The Lawn to participate.

We have our own Community Guidelines at Streaking The Lawn. You should read them.

Spinner.vc97ec6e

Authenticating

Great!

Choose an available username to complete sign up.

In order to provide our users with a better overall experience, we ask for more information from Facebook when using it to login so that we can learn more about our audience and provide you with the best possible experience. We do not store specific user data and the sharing of it is not required to login with Facebook.

tracking_pixel_9347_tracker