BREAKING(?): Wahoo Metrics taking talents to Streaking the Lawn

For those who haven’t already heard, I’ve decided to transfer[1] my articles and join the group of talented Virginia sports writers over at Streaking the Lawn![2] Check out my intro post.

I hope you’ll join me at the new site, where you’ll have access to a plethora of non-stat related reports and opinions on Hoos sports. All my old posts will also make the switch. Look for my first posts soon as part of the highly anticipated Midnight Madness.

  1. [1] Tony Bennett’s worst nightmare
  2. [2] after being offered a very lucrative transfer package (not really)
More Wahoo Sports:
University of Virginia Cavaliers IDPPP vs. the RPI Top-100

2012 Football Season: Taking on Water?

On April 10, 1912, Captain Edward Smith set off on a voyage, brimming with pride and optimism. On September 1, 2012, Mike London led a young Hoos team onto the field to kick off the 2012 season, after ending the previous season in a prestigious bowl. Captain Smith’s unsinkable ship infamously struck an iceberg and promptly sank.[1] London’s team is at least lost in the North Atlantic.[2]

Last week’s loss to Duke was one of the more devastating I’ve seen. The halftime lead evaporated as the previously unstoppable offense suddenly couldn’t muster a first down.[3] Even past employees and physical structures associated with the football team felt the ripples of this monumentally awfully loss; Al Groh was fired from his defensive coordinator job at Georgia Tech, and the football practice facility roof literally caught on fire. This post will grapple with the uncertain state of the 2012 season following the devastating loss to Duke: are the Hoos “just” in serious danger, or are they already sinking?[4]

{Hoos 2012 Performance Margin}

I’ll attempt to evaluate the trend in football performances by standardizing the Hoos FBS opponents. The goal is to discern whether, from week to week, the Hoos are performing worse against this statistically standardized opponent. Are the Hoos in the midst of a steady decline, evidencing minimal hope of a rebound, or do they just have a couple miserable performances amongst games against some strong FCS teams?

To standardize the opponents, I began by comparing each opponent’s current rank to the Hoos current Sagarin rank of 89 in the chart below. I then averaged the rank differences and final point differences in the games against each of these teams.[5] The listed gain is the ratio of average rank difference to average actual point difference.[6] The concept is, by multiplying the actual point difference by this gain,[7] then adding it to the rank difference, the average resulting value, listed as the Performance Margin,[8] should be zero.

To illustrate the concept, consider game 4 against TCU.[9] TCU is ranked 70 spots ahead of the Hoos (+70). The Hoos lost by 20, which when multiplied by the gain, gives -49.6. Adding this to the rank difference of +70 gives an overall performance margin of 20.4, which is good because its positive. This effectively means that the Hoos outperformed their rank in this game. An average performance of a team with the same rank deficit would result in a PM of 0; corresponding to a 28ish point loss.[10] This calculation method allows a comparison of losses to losses over the course of the season.

The chart shows the wide range of Performance Margins for the Hoos in 2012, which are plotted in the graph below. An ideal graph would show consistent performances that only vary slightly around zero. The worst case scenario would be a steadily decreasing line that starts with a high positive and ends with a low negative, which would show that the Hoos are getting worse with each passing game. Using our TCU example, a steadily decreasing line would indicate that, if the Hoos were to somehow play TCU every week, they would lose by larger and larger margins in each successive week.

The graph reveals some interesting aspects of the Hoos performances. The two ACC losses are by far the worst games of the season, and the Georgia Tech loss is somewhat worse than the loss to Duke. As expected, the win against Penn State is the best performance, but the loss to LA Tech isn’t that far behind, and was a “better” loss than the one at TCU. Also note that the two highest points are the two home games.[11]

Overall, the Hoos do not appear to be taking on water and sinking into the Atlantic, as they played competitive games against very strong TCU and LA Tech teams after defeating a rising Penn State team. The ACC performances, however, leave much to be desired. A return to good ‘ol Scott Stadium for the next two ACC games could be exactly what the Hoos need to right the ship. Despite the current state of the team, I would be very, very surprised if the Hoos lost both the Maryland and Wake Forest games.[12] A small winning streak and a return to .500 in both overall and ACC records is not entirely out of the question.

{Epilogue: Extremely Abbreviated Maryland Preview}

My method of extrapolating Sagarin rankings to predict the final score has been the more accurate of my two attempts at final margin predictions.[13] Maryland’s rank of 72, with the Hoos way down at 89, predicts an 8 point win for the Terps. Since we just finished discussing the Hoos stronger home field performances, I’ll add the customary 3 points in favor of the home team to the final margin. My final prediction is: Maryland 30 – 25 Hoos.[14]

  1. [1] forever immortalized by a decent movie and the cheesiest Celine Dion song ever written (http://youtu.be/zmbw8OycJrE <– you’re welcome), and that’s saying a lot
  2. [2] I’ll admit, the Hoos football team was never really Titanic-like
  3. [3] still bothered by the fact that, when asked about what halftime adjustments Duke made, London shrugged it off as the difficulties of playing from behind. I hope he realizes that Duke did in fact make a number of strategic adjustments, and that he can prevent that same adjustment from burning the Hoos again.
  4. [4] with a quick Maryland preview at the end
  5. [5] under the grid portion
  6. [6] well, the absolute value of the actual point difference
  7. [7] to set it at the same magnitude level as the rank difference …. hopefully this is making sense
  8. [8] PM in the chart
  9. [9] still referring to the chart
  10. [10] make sense now?
  11. [11] and the Hoos are returning for two straight home games
  12. [12] with Wake being the more likely win
  13. [13] if you haven’t seen this method, click on (Football > Game Predictions), it’ll be in any of those posts
  14. [14] Vegas has the Hoos by 1.5
More Wahoo Sports:
Win Probability and the Florida State Game
Virginia Tech @ UVA Football Preview: Fan Edition
Fading Defense and the Chick-fil-A Bowl

Game 6 at Duke Preview: Bedeviled

The mixture of emotions surrounding the Virginia football program is enough to turn the most even-keeled fan into a mess of optimism and despondence. On one hand, current savior of the offense Phillip Sims will start his first game this weekend in place of previous savior of the offense Mike Rocco.[1] On the depressing side of things, the 2-3 Hoos are traveling to North Carolina to face a 4-1 Duke team[2] in a game they probably need to win to have a hope of becoming bowl eligible.[3] I’m already holding my breath.

The less informed of us might laugh and arrogantly wave our hands in our patterned orange pants and v-sabre bowties[4] and say “Duke has long been the laughingstock of ACC football, it’s the perfect opportunity for the Hoos to right the ship down in Durham.”[5] A sobering look at recent seasons reveals that the Hoos have lost three of their last four games against the Blue Devils. [6] I’m now holding my breath even more.

This post will attempt to predict the outcome of Saturday’s game.[7] As per usual, I’m going to slightly alter the method this week. I used two separate methods in predicting the TCU game, but was shortsighted enough to apply my own discussion of intangibles to the resulting range of possible scores; simply averaging the two scores would have predicted an almost exactly correct final margin. So that’s what I’ll do this time. The final prediction follows the descriptions[8] and individual predictions[9] for each of the two methods.

{Prediction Method 1: Overall Rank Difference vs. Point Difference and Points Scored}

The first method calculates the overall rank difference, in Sagarin rankings, for the two teams involved in the game and plots this value against the final point margin and points scored. The idea is that any rank difference between two teams can then suggest a final score difference and a point total from the perspective of one team. In this iteration, I limited the games creating this plot to Hoos and Duke games this season.

The above games created the below plot:

The Sagarin rankings have Duke ranked at 86 and the Hoos at 85, thus predicting the Hoos will score 28 points[10] and win by 1.[11] Method 1 predicts: Hoos 28 – 27 Duke. As mentioned above, I wanted to verify this score by comparing it to a completely separate final prediction …

{Prediction Method 2: Offensive and Defensive Rank Differences vs. Points Scored}

Football Outsiders separately ranks both offenses and defenses in the FBS. For this method, I plotted the rank difference between Team A’s offense and Team B’s defense versus the number of points that Team A scored in the game. This also works in reverse to create a whole separate set of data points; the difference between Team B’s offense and Team A’s defense and the points scored by Team B.[12] The following chart shows this data for the same set of games:

This method gives a graph and equation of:

The Duke offense is ranked 72nd, while the Hoos defense is ranked 46th.[13] Based on these rankings, the above equation predicts 24 points for Duke. On the other side of the ball, the Hoos offense is ranked 78th, while the Duke defense is ranked 105th;[14] predicting that the Hoos will score 43 points[15] against Duke. Method 2 gives Hoos 43 – 24 Duke.

{Conclusion}

The Vegas line currently has the Blue Devils winning by 1.5. My two methods produced drastically different results, but at least both picked the “under” on the final score margin. I’ll stick with my averaging method despite the large variation and hope that the errors in these two methods perfectly offset each other.

Final Prediction: Hoos 35 – 26 Duke[16]

  1. [1] the title might be a bit of a stretch for Rocco, but he was at least considered good … for a while
  2. [2] mandatory note that Duke has largely played really terrible teams to amass this impressive record
  3. [3] although the rest of the ACC isn’t that great either, so who knows. Also, “any given Saturday” etc. etc.
  4. [4] as you can see, I’ll be commenting on the sports knowledge of a terrible UVA caricature advanced by other members of the ACC. A person who is undoubtedly not you.
  5. [5] since this caricature would also use nautical metaphors.
  6. [6] Sad. So very sad.
  7. [7] My success in predicting previous games can be found here.
  8. [8] which are worth skipping if you’ve read the descriptions before
  9. [9] probably shouldn’t skip these
  10. [10] per the orange line
  11. [11] the blue line.
  12. [12] so even though the third column says “D rank difference,” it’s still from the perspective of the opposing team’s offense.
  13. [13] interestingly, it’s gone up since the TCU game
  14. [14] yikes. Hopefully a field day for Sims “the savior of the offense” ™
  15. [15] oh snap!
  16. [16] note that any discrepancies from averaging are due to fractions of points that are calculated but not shown in this post
More Wahoo Sports:
Season Preview - Offense: Good Old-fashioned QB Controversy, Part II
Season Preview - Offense: Receivers, the Critical Unknown
Game 4 at TCU Preview: The Formidable Horned Frogs