Category Archives: Win Probability

Recap and Win Probabilities vs. Penn State

Posted on by
1

Yesterday’s game was an absolutely crucial win in setting a successful course for the 2012 season. With upcoming away games against Georgia Tech and TCU, a loss against Penn State could have set a negative tone easily leading to a hole-digging 1-3 start. It is critically important for every successful team to scrap through at least one flat outing. These sub-par performances are going to happen in every season, so when they come against a solid opponent, emerging victorious is what separates quality teams from those on the cusp. I can’t emphasize this enough; if the Hoos go on to match the success of last season, the ugly Penn State win will be one of the prettiest in hindsight.[1]

This post will graph and comment on the Win Probability of the Hoos throughout the PSU game.[2] Win Probability (“WP”), as the name suggests, is an attempt to measure one team’s likelihood of winning the game as a function of score, time remaining, field position, and down and distance. The formula is an algorithm combining historical data of various game situations and outcomes. There are a number of assumptions[3] inherent in this analysis. Nonetheless, the formula should create an interesting result and give us a baseline for comparing various events within the game. An explanation of the WP of a few major plays follows a graph of the overall WP below.

{Hoos Win Probability Throughout Penn State Game}

Big McGee catch #1: The Hoos faced 3rd and 16 on their half of the field on a drive that needed to end in a touchdown to avoid a crippling loss. Jake McGee’s huge one-handed catch single-handedly[4] improved the Hoos chances of winning the game by 11 percent, keeping our hopes alive.

TD / Big McGee catch #2: McGee struck again, catching a touchdown pass on 3rd and goal from the 6 with 1:31 remaining. As the pass left Rocco’s hand, the Hoos had a 37% chance of winning the game. The Hoos win probability increased to 71% the moment the pass settled into McGee’s grasp. Interestingly, the ensuing Penn State drive leading up to their final FG attempt entirely negated the 36% increase in win percentage resulting from this TD catch.

Final Kick: Poor Sam Ficken and mother nature conspired to contribute the final 69% of win probability the Hoos needed to escape with an as-predicted[5] 1 point victory.

{Short Postgame Thoughts}

Offense: Rocco was bad for most of the game, but the receivers[6] weren’t helping either. He also seems to occasionally make his reads from short to long, instead of vice versa; choosing some ill-advised dump-offs. Inevitably, when the bell tolled, he answered, leading the team on his umpteenth game winning drive. I agree with most that the decision to put Phillip Sims in was mind-boggling. The Hoos offensive success also seems to depend largely on Bill Lazor, as they have great drives when he has time to map out a plan of attack.[7] The offensive line certainly did not live up to expectations, which was a large factor in the lack of running game success.[8]

Defense: Very, very good performance against the decidedly weaker unit of Penn State’s team. Coverage was good, blitzing was timely, and they cleaned up myriad offensive mistakes.[9] Nicholson and Greer had noticeably fantastic games. Drequan Hoskey was a bit over-aggressive on a few tackling opportunities. The heralded Eli Harold also made a small contribution.

Look for a Georgia Tech preview coming later this week.

  1. [1] Also keep in mind: this Penn State defense should easily be highly ranked at the end of the season. The offense was bad, but with good reason. ↩
  2. [2] much like the inaugural post of this blog inspired by the wacky FSU game ↩
  3. [3] Some assumptions: 1. The home team starts the game with a WP of 53%. I borrowed this number from baseball, where advanced statistics are more developed. Probably more accurate than just calling it 50%. 2. The model assumes that only score difference, and not score, matters. Imagine a game where Team B is at the opponent’s 40 with 2 minutes remaining in a tied game. Would you think they are more likely to score if the game is 35-35 as opposed to 0-0? Is this just perception or reality? ↩
  4. [4] oh, puns ↩
  5. [5] didn’t think I’d let that go for the whole recap, did you? ↩
  6. [6] outside of McGee ↩
  7. [7] like the first drive of the second half ↩
  8. [8] although again, the Penn State defense, specifically its linebackers, is very good ↩
  9. [9] although some of their success can be attributed to the ineptitude of Sam Ficken ↩

Fading Defense and the Chick-fil-A Bowl

Posted on by
Reply

Let’s start off by stating the big picture: the Hoos had a fantastic football season, which culminated in a prestigious bowl game against the defending national champion.  Any of us would have been ecstatic to hear that news at the start of the season.  The rough losses to Tech and Auburn to end the season, however, tarnish our perspective.  In the coming weeks we can hopefully appreciate this significant step forward and the potential start of a new trajectory for UVA football.

[UVA WP Throughout the Chick-fil-A Bowl] [1]  [2]  [3]

[The Onside Kick]

Just as the Auburn offense was starting to gain traction, they successfully executed an onside kick after tying the game at 14.[4]  Auburn needed just five plays on the ensuing possession to take a 21-14 lead.  The onside kick alone did not dramatically increase Auburn’s WP.  However, when sandwiched between two quick scoring drives, the three events lessened the Hoos WP by more than 40% in approximately four minutes of game time.  This sequence was precisely as devastating a blow as it seemed live.

[The Fake FG]

The Hoos were crawling back into the game when they faced a 4th and 6 at the Auburn 15.  At the start of that play, their WP stood at 33% with an even 2.00 expected points on the possession.  The resulting failed fake dropped their WP to 26%; a significant drop for a single game event.[5] [6]  If the Hoos had taken the field goal, their WP would have slightly increased their to 34%.  A first down at the Auburn 9 would have yielded a WP of 42% and 4.83 expected points. Anything over a 44% success rate would make going for it a good decision, as the probability of gaining 9 WP vs. losing 7 WP would even out. [7] The historical success rate on similar 4th and 6 scenarios (in the NFL) is 42%.  Consequently, this was technically a bad decision, although it’s really too close to be significant.  Nevertheless, the unsuccessful fake continued the Hoos woes.

[Summary]

Overall, the Hoos defense simply faded after the first quarter; unable to contain the athleticism of the Auburn offense.  That, and Onterio McCalebb is fast.  It would have been interesting to see what impact Greer and Minnifield could have had on the game.

On a positive note, the passing game was surprisingly efficient.  Rocco’s one interception was more of a Parks lateral to an Auburn defender.  The offensive game plan as a whole was well constructed; some trickery without detracting from the team’s ability to execute the offense.  Bill Lazor needs to be confined to Grounds with outside world contact.

While the game technically doesn’t mean anything for the Hoos, the nationwide exposure could have some positive and negative recruiting effects.  On the plus side, recruits saw a team on the rise lead by a charismatic head coach.  Conversely, they also saw a team not yet ready to compete against an established SEC program.  Given Mike London’s recruiting prowess, I think he’ll be able to emphasize the former.  Some major recruits are still undecided, so we’ll see if he can continue his string of strong classes.

Can’t wait for next season.

  1. [1] When the Hoos went up 7-0, the WP calculation includes the ensuing kickoff penalty, placing the Auburn offense at the 40.  A touchback on the kickoff would have increased the Hoos WP by 3% to 72%. ↩
  2. [2] Hurting Auburn’s first QB Moseley was a bad strategic move, as he was 0-2 on drives.  They had TDs on 4 of their next 5. ↩
  3. [3] Also, Barrett Trotter and his handlebar moustache made me angry. ↩
  4. [4] As an objective football fan, that was one of the most impressive onside kicks I have ever seen.  The announcers harped on Hoos “turning and running,” but I’m not sure they would have stopped it if they knew it was coming. ↩
  5. [5] Although the first blocked punt was a -18% on a single play ↩
  6. [6] The play appeared to be blown, as a potential receiver fell down while running out into the flat.  Hodges would have had a run/pass option based on the reaction of the covering CB.  Not sure who fell, but it was costly. ↩
  7. [7] 9 x .44 = 7 x .56 ↩

Win Probability and the Florida State Game

Posted on by
1

Like any devoted Wahoo fan, I watched last night’s game covered in UVA apparel[1] in hopes that my superstitions would help overcome an 18 point spread. What we eventually witnessed was easily one of the best UVA games in history featuring an extremely bizarre ending. The myriad of big plays and momentum shifts made me wonder: how influential were the various plays in the game? Was my defeatism late in the fourth quarter actually warranted or just a product of anxiety?

Win Probability (“WP”), as the name suggests, is an attempt to measure one team’s likelihood of winning the game as a function of score, time remaining, field position, and down and distance. The formula is an algorithm combining historical data of various game situations and outcomes. There are a number of assumptions[2] inherent in this analysis. Nonetheless, the formula should create an interesting result and give us a baseline for comparing various events within the game. A graph of the overall WP and the WP within the last 10 minutes follows an explanation of the WP of a few major plays.

[The Initial UVA Score - Jones TD Reception]

The only threat up to this point in the game was an FSU drive that ended in a fumble and turnover as they neared field goal range. Putting points on the board early in the second quarter gave the Hoos a WP of 71%. This number seems a bit high offhand this early in the game, but any lead is necessarily over 50% WP, and a 7 point lead (almost always) requires two scores to lose. Statistics aside, it was a fantastic play on both ends of the pass by Rocco and Jones Superman.

[The Rest of the First Half – FSU Answers]

As the FSU offense settled into the game they tied the score on a touchdown pass and went ahead on a field goal late in the second quarter. The tying touchdown lessened UVA’s WP to 52%, while the later field goal further dropped the WP to 38%. I find it interesting that the FSU field goal is nearly as important as the touchdown (-19% vs. -14%). The numbers make sense given that the field goal gave FSU the lead, and the comparative WP lends a unique perspective.

[Minnifield's TD-saving Tackle]

FSU was starting to stretch the field in the third quarter and connected on a deep pass over the middle to Bert Reed. Reed sprinted through open space, but was caught at the one yard line by Chase Minnifield, who closed a substantial gap over the final 40 yards to make the tackle. FSU appeared to be taking control of the game, but the UVA defense stood firm and forced a field goal.

To calculate the Win Probability added by Minnifield’s tackle, I compared the resulting field goal to a hypothetical touchdown scored at the time of the tackle. UVA had a WP of 21% immediately after the tackle, which increased to 26% once UVA forced a field goal. If Minnifield had not made the tackle, UVA would have had a WP of 19%. Minnifield’s tackle therefore was an immediate +2% WP that gave the defense an opportunity to gain +7% WP.

[Greer's Sack]

Things were looking darkest just before the dawn UVA’s game winning drive. UVA’s WP bottomed out at 1% with four minutes left in the game. FSU had a first down at the UVA 29 with a 6 point lead. This figure seemed low, but consider: if FSU scores any points, the game is virtually over (i.e. 99.9%ish WP); FSU can score points without gaining any additional yards; even if FSU doesn’t score, UVA must score a TD. One percent starts to seem more reasonable.

One minute and fifteen seconds later, UVA had forced a 3rd and 11 from the 30, raising their WP after two downs to 12%. Greer’s monumental sack on 3rd down was crucial in pushing FSU out of field goal range, further raising their WP to 25%. WP reveals that Greer’s sack was obviously influential at +13% WP. Interestingly, though, the two stout defensive plays prior to the sack were nearly as important at +11% WP combined.

[Last UVA Drive]

This drive was incredible. The offense stalled repeatedly throughout the second half before marching 75 yards in 47 seconds when anything less would have led to a loss. Statistics cannot measure whatever it was that changed on this drive; whether it was a tired FSU defense, inspired play-calling, or sheer will. Whatever it was will henceforth be referred to as the “Mike London Factor” (“MLf”). The brief drive added +57% WP, raising the UVA WP to 76% with 1:16 left in the game. The 5 plays on this drive averaged a +11% WP each. Not considering the punt, UVA had a +82% WP spread quite evenly over their last 8 plays.

[UVA Wins, Four Separate Times]

The WP of 76% was measured after the ensuing kickoff and disconcertingly long return. If the kickoff had resulted in a touchback, UVA would have had a WP of 83%. One late-game kickoff coverage therefore had the same importance as the Minnifield tackle and forced FG. Special teams do matter after all.

Win #1: When FSU faced a 4th and 1 at the 50, UVA’s WP had increased to 99%.[3] Notably, FSU’s clock mismanagement on this final drive was crippling. They posted a -23% WP over 45 seconds despite moving the ball. This WP loss was simply from not getting out of bounds, not calling timeouts, and throwing short of the first down marker.

We all know what happened on 4th and 1. We are also obligated to insist that no facemask penalty occurred, while secretly acknowledging that it probably was the correct call. Part of me wanted this penalty to be overlooked just because Jimbo (seriously?) Fisher was complaining the entire game. But I digress. The fifteen yard penalty on fourth down dropped UVA’s WP -54% to 46% by putting FSU in field goal range. Huge, huge call.

Win #2: FSU continued to attempt to shoot themselves in their collective feet by throwing short of the first down marker, apparently completing a pass in-bounds and running out the clock. The subsequent incomplete pass ruling resulted in a 42% WP for UVA. Again, it was close, but I would have to agree with the call. Same Jimbo Fisher comment applies. This was an even bigger call in WP terms since FSU had shortened the distance of the field goal.

Win #3: FSU appeared to jump offsides while lining up for a field goal, which would have resulted in a 10 second runoff and the end of the game.[4] UVA, however, was flagged for a “disconcerting signal,” instead moving FSU 5 yards closer.[5] There has been some post-game confusion on whether movement or a yelling UVA defense prompted the call. Regardless, consider this: I watch a lot of football; I have therefore seen a lot of last second field goals; I have never seen this call made in a game; I can see no reason to suspect that something unusual was happening on this play. Draw your own conclusions. Decreasing the field goal distance from 47 to 42 yards resulted in -10% WP, giving UVA a 32% WP on the most important snap of the game.

Win #4: 100% WP! This time they actually won due to the missed field goal. I haven’t seen a replay of the kick, but it looked extremely close live. Meanwhile, Jimbo Fisher reacted like someone had just called him John.[6]

[Graphs]

 

[Final Thoughts]

Stay tuned for a preview of the noble Cavaliers vs. the pregnant turkeys.

  1. [1] Including a sweatshirt, t-shirt, pants …. and boxers. The t-shirt commemorated the FSU upset from 2005; the most memorable game from my time as an undergrad. Notably, I also have UVA-colored argyle socks, but didn’t want to go overboard. ↩
  2. [2]  Some assumptions: 1. The home team starts the game with a WP of 53%. I borrowed this number from baseball, where advanced statistics are more developed. Probably more accurate than just calling it 50%. 2. The model assumes that only score difference, and not score, matters. Imagine a game where Team B is at the opponent’s 40 with 2 minutes remaining in a tied game. Would you think they are more likely to score if the game is 35-35 as opposed to 0-0? Is this just perception or reality? ↩
  3. [3]  FSU had a 99% WP with 4 minutes left in the game. UVA had a 99% WP 3 minutes and 30 seconds later. ↩
  4. [4]  Somewhat obscure rule. ↩
  5. [5]  Even more obscure rule. ↩
  6. [6]  Real name: John James Fisher. ↩