Saturday, March 24, 2012

2011 ACC SDPI

I know a lot of regular readers were becoming worried that the offseason SDPI recaps had slowly gone the way of the buffalo. Fear not, SDPI returns! For those who don't know what SDPI is, here is the Reader's Digest synopsis. SDPI measures how many standard deviations a team is above or below average at gaining and preventing yards. Since conference play occurs in a vacuum, teams are rated against their conference mates and not against the nation at large. Since there are 11 conferences, you can look forward to 11 posts on how SDPI played out in 2011. We'll begin with the six BCS conferences and tackle those in alphabetical order before moving on to the mid-major leagues. As always, we begin with the ACC. Here's the link to last year's ACC post.

This first paragraph will explain how SDPI is calculated. So if you want the meat of this article skip on down. In the 2011 ACC regular season, conference play only, championship games excluded, the average ACC team gained and allowed 3043.75 yards. The standard deviation for yards gained (offense) was 309.47 yards. The standard deviation for yards allowed (defense) was 360.91 yards. NC State gained 2716 yards and allowed 2601 yards. Their offensive SDPI was -1.06 = [(2716-3043.75)/309.47]. Their defensive SDPI was 1.23 = [(3043.75-2601)/360.91]. Their total SDPI was 0.17. This number ranked 8th in the ACC.

Here are the 2011 ACC standings.Now here are the 2011 ACC SDPI standings. The standings are sorted by division by total SDPI with ranking for each category (out of 12 teams) in parentheses.Clemson and Virginia Tech ended up in a virtual tie atop the SDPI ratings, and lo and behold both teams represented their respective divisions in the ACC Championship Game. Clemson routed the Hokies in Charlotte, winning their first ACC title since 1991. Joining Clemson and Virginia Tech in the upper-tier of the ACC was Florida State. After the Seminoles, five teams (Georgia Tech, North Carolina, Miami, Virginia, and NC State) were within four tenths of a standard deviation of each other. Perhaps not surprisingly, those five teams finished a combined 20-20 in the ACC.

So Who Was Better Than Their Record Showed?
Florida State won four conference games by 20 points or more. That was the most of any team in the ACC. However, the Seminoles were not quite as fortunate in close games, losing one-score decisions to Clemson, Wake Forest, and Virginia leaving them with a somewhat deceiving 5-3 conference mark. A few lucky breaks here or there, and perhaps Florida State is playing in the Orange Bowl and not embarrassing the conference.

So Who Was Worse Than Their Record Showed?
Wake Forest used some Jim Grobe magic (patent pending) and nearly stole the division away from Clemson. Four of the Deacons five league wins came by eight points or fewer. Unfortunately, in the games that were not close, the Deacons ended up on the losing end. In two of their three conference losses, the Deacons were beaten by a combined 46 points by Virginia Tech and North Carolina.

Conference Superlatives:

Best Offense: Clemson 1.64
Here's what a high school offense can do. The Tigers topped 500 yards in three of their eight conference games after doing so just once from 2007-2010.

Worst Offense: Boston College -2.26
While Clemson rose (meteorically I might add) from the offensive abyss they occupied in 2010, Boston College dug deeper. The Eagles were held below 300 yards in five of their eight conference games and topped out with a 404 yard effort against Maryland.

Best Defense: Florida State 1.43
After opening conference play with consecutive defeats to Clemson and Wake Forest, during which they allowed a combined 834 yards (hardly a terrible amount), the Seminoles put the ACC on lockdown. Their last six league opponents averaged just 282 yards per game.

Worst Defense: Maryland -1.80
The Temple debacle was a chilling vision of things to come. The Terps, prohibitive favorites against the Owls, allowed 285 yards on the ground and were bludgeoned in their own stadium. In their ACC games. the Terps twice allowed opponents to rush for over 300 yards against them (Clemson and Boston College) and held just two conference opponents under 400 yards (Miami and Georgia Tech). Perhaps not coincidentally, the Terps beat Miami and suffered their closest ACC loss to Georgia Tech.

Florida State Football: Brought to You by Hype Williams
2011 began like many Florida State seasons before it, with the Seminoles ranked in the AP Top-10. Unfortunately, it ended like many Florida State seasons, with the Seminoles ranked outside the AP Top-10. In fact, since 2001 (the year after their last national championship game appearance), Florida State has begun the season in the top-10 and finished outside the top-10 four times, tied for the most nationally with Florida and Oklahoma.The Gators of course, can revel in their two national titles and the Sooners can take solace in their three national championship game appearances since 2001, Florida State has come nowhere near those heights. Florida State may not begin the 2012 season in the top-10, but if they are, this likely represents their best chance to live up to the hype.

Saturday, March 10, 2012

The Predictive Powers of Pythagoras

Regardless of what any talking head or former player may say on ESPN regarding intangibles and heart, team success is a function of the points they score and the points they allow. With that in mind, I want to take a fresh look at the Pythagorean Theorem and its predictive powers in college football. I've looked at the Pythagorean Theorem on this blog before. In fact, in my very first blog post nearly seven years ago (time flies), I accurately surmised that the 2005 Tennessee Vols were not national title contenders despite their top-5 preseason ranking. For those who don't know what the Pythagorean Theorem is (at least as to how it relates to sports), here's a snippet from that first post:

'In 1980 noted baseball analyst and the founder of sabermetrics, Bill James, developed a theory for predicting a team's winning percentage based on the number of runs a team scored and allowed. He called it the Pythagorean Theorem of baseball and it looked something like this:

(Runs Scored)^2
------------------------------------
(Runs Scored)^2 +(Runs Allowed)^2

The resulting ratio was a team's estimated winning percentage. Over the course of a 162 game season the Pythagorean Theorem predicted the final record for most teams within three or four games of their actual performance. The Pythagorean Theorem has application beyond baseball (and some triangles I hear). It could also be applied to predict a football team's actual performance.'

The other day I was brainstorming ways to make the Pythagorean Theorem more accurate. Football has a unique scoring system. Touchdowns are worth six points, field goals three, and safeties two. In addition, after touchdowns, teams can attempt to kick for one point or cross the goal again for two additional points. By contrast, in baseball and hockey, scoring events, runs and goals, are worth the same amount each time. Basketball is a little more complex, with free throws worth one point, field goals worth two points, and shots behind a certain arc worth three points. Basketball also has more scoring events than baseball and hockey. With that in mind, I wanted to look for a scoring event in football that is analogous to a run in baseball or a goal in hockey. Its actually pretty easy to find. Offensive touchdowns. Teams score offensive touchdowns at about the same ratio that teams score runs in baseball and goals in hockey. In addition, offensive touchdowns help us get rid of some of the statistical 'noise' generated during a football game. Field goals (both for and against) are heavily dependent on one individual player in a sport where team play matters most of all. Special teams and defensive touchdowns, while they play a huge role in the winner of an individual game, do not have a great deal of predictability going forward. Notice I said 'some' of the noise, but certainly not all. Defensive and special team ability can make offensive touchdowns easier by setting up the offense with a short field. Still, ignoring all scoring plays except for offensive touchdowns should give us a good idea about a team's strength.

So how does this relate to the Pythagorean Theorem? Well instead of looking at a teams ratio of points scored to points allowed, we'll look at a teams ratio of offensive touchdowns to offensive touchdowns allowed. The formula will look a little something like this:

(Offensive Touchdowns)^2.37
-----------------------------------------------------------------------
(Offensive Touchdowns)^2.37 +(Offensive Touchdowns Allowed)^2.37

We'll call the resulting record a team's Adjusted Pythagorean Record (or APR). I used 2.37 as the exponent because research has shown it makes the resulting ratio a more accurate predictor.

Now lets take a look at the Adjusted Pythagorean Theorem in action.

In 2007, Oklahoma State enjoyed a rather middling season in the Big 12. There were some highlights of course. They beat Nebraska in Lincoln for the first time since 1960, upended a Kansas State team ranked in the top-25 at the time, and gave Texas a run for their money in Stillwater. They were also trounced by Oklahoma and beaten by Kansas in Stillwater for the first time since 1995. All in all, the Cowboys went 4-4 in the Big 12. How did they do according to the Adjusted Pythagorean Theorem? Well, in their eight conference games, they scored 35 offensive touchdowns. Not bad. That tied for the third highest number in the conference. Unfortunately, teams also scored 35 touchdowns against their defense (tenth in the conference). All in all, the Cowboys record based on their offensive touchdowns scored and offensive touchdowns allowed was 4-4. And lo and behold, that was their record. Of course, this is just one cherry picked example. the Adjusted Pythagorean Theorem rarely hits the nail directly on the head. However, it does usually correctly predict a team's record within one game of their actual record. Since 2007, the Adjusted Pythagorean Theorem predicted each team's conference record within one game for nearly two-thirds (386 of 582 team seasons or 66.3%) of all teams. So if the Adjusted Pythagorean Theorem does a pretty good job of predicting a teams record based solely on how many offensive touchdowns they scored and allowed, what happened to those teams that significantly over or underperformed their APR? I thought you might ask that. To answer that question, I looked at all teams from 2007-2010 with a conference record at least one and a half games different (either better or worse) than their record as predicted by the Adjusted Pythagorean Theorem. I then looked at how they did in conference play the following season. And now I will share with you the results. As with the previous two blog entries, I have separated BCS and non-BCS conference teams to see if there appear to be any differences. First off, the BCS conference teams that exceeded their APR.
Of the 15 teams that exceeded their APR in conference, 13 declined the following season. One team stayed the same, and only Oregon State circa 2007 improved. Many of the declines were significant. 11 of the 15 teams declined by at least two games in conference play, and some (Texas, Auburn, and Cincinnati) totally fell off a cliff. On average, these teams declined by 2.27 games in conference play the following season.

Now the BCS conference teams that underperformed their APR.
Of the 20 teams that underperformed their APR in conference, 13 improved the following season. Three stayed the same and four finished with worse conference records. Overall, the teams improved by just a shade over one game (1.025) in conference play the following season.

Now the non-BCS conference teams that exceeded their APR.
Of the 15 teams that exceeded their APR in conference, 12 declined the following season. One team stayed the same and two improved. Overall, these teams decline by 1.8 games in conference play the following season.

Now the non-BCS conference teams that underperformed their APR.
Of the 14 teams that underperformed their APR in conference, 11 improved the following season. Two stayed the same and only one team (New Mexico circa 2008) declined. Overall, the teams improved by nearly two games (1.964) in conference play the following season.

So now the next question on your mind should be: Who under and over-performed their APR in 2011? Eight BCS conference teams exceeded their APR in conference play. In the ACC, Virginia allowed more offensive touchdowns than they scored, yet managed a 5-3 ACC record. Mike London may find it hard to take the Cavaliers to back-to-back bowl games for the first time since Al Groh did it in 2004 and 2005. In the SEC, Auburn allowed 10 more offensive touchdowns (27) than they scored (17), yet still somehow managed to win half their league games! With Texas A&M joining the SEC West and Clemson on the non-conference schedule, the Tigers could miss the postseason altogether in 2012. In the Big 12, Kansas State allowed more offensive touchdowns than they scored yet managed a 7-2 conference record, finishing all alone in second place. A reversal of fortune is almost certainly in store in 2012. The Big 10 had three teams drastically exceed their APR. Penn State won six Big 10 games and was in contention until the final weekend of the regular season. However, the Lions allowed more offensive touchdowns than they scored and will be breaking in a new coach for the first time since 1965. Michigan State won seven conference games and nearly earned their first Rose Bowl bid since 1987. However, based on APR, the Spartans should have only won a shade over five league games. Nebraska underachieved all season and actually allowed more offensive touchdowns than they scored. That should make Michigan the favorite in the Legends Division in 2012. UCLA 'won' their division in the Pac-12 despite allowing more offensive touchdowns than they scored. Jim Mora may be in for a long season during his first year at the helm. And finally, Louisville was a surprise co-Big East champ in Charlie Strong's second season. However, the Cardinals allowed more offensive touchdowns than they scored (sensing a theme?), so even with West Virginia out of the picture, a BCS bid is still likely a pipe dream.

Six BCS conference teams significantly underperformed their APR in conference play. Miami scored more offensive touchdowns than they allowed, yet finished with a losing record in the ACC. Perhaps Al Golden can coax a solid showing out of this under-performing group. North Carolina also scored more touchdowns than they allowed in league play, yet finished with a losing record. Larry Fedora steps into a pretty good situation assuming the NCAA investigation has run its course. In the SEC, Vanderbilt showed some moxie and was much better than their 2-6 SEC record. Could James Franklin take then to consecutive bowls for the first time ever? Texas A&M was expected to be a contender during their last run through the Big 12. A plethora of second-half collapses caused the Aggies to finish 4-5 despite scoring seven more touchdowns than they allowed. They move to an insanely tough division and must contend with Alabama, Arkansas, and LSU, but could position themselves in the middle of the division. South Florida lost seven of eight after a 4-0 start and finished tied for last in the Big East. No one would be shocked if the Bulls went from worst to first. Syracuse also drastically underperformed their expected Big East record. Syracuse probably won't be winning the league in 2012, but a return to the postseason is likely.

Five non-BCS conference teams exceed their APR in conference play. Marshall finished 5-3 in Conference USA despite allowing more offensive touchdowns than they scored. The Herd will find returning to the postseason quite challenging in 2012. UAB won three games in Conference USA despite allowing twice as many offensive touchdowns as they scored (those teams typically only win about one game). Despite the relative good fortune, the coach was still canned. New coach Garrick McGee has his work cut out for him. Northern Illinois won the MAC Championship despite posting the profile of a solid, but hardly elite MAC team. Dave Doeren deserves credit for leading the Huskies to the summit in his first season, but a repeat is highly unlikely. Arkansas State finished unbeaten in the Sun Belt and got coach Hugh Freeze a job with Ole Miss. New coach Gus Malzahn may yet make the Redwolves into a Boise State-like power, but there is likely to be some regression in his first season. And speaking of Sun Belt powers, Louisiana-Lafayette also significantly exceeded their expected Sun Belt record. A second straight bowl appearance would likely have coach Mark Hudspeth in line for some big time jobs.

Four non-BCS conference teams significantly underperformed relative to their APR in conference play. UCF followed up their top-25 final ranking in 2010 with a losing season in 2011. They'll be back at or near the top of the league in 2012. Miami of Ohio followed up their MAC Championship in 2010 with a losing season in 2011. With Temple gone to the Big East in 2012, the MAC East is wide open. Air Force scored more offensive touchdowns than they allowed, yet finished with a losing record in the Mountain West. The Falcons could be poised to take flight in 2012. And finally, Louisiana-Monroe scored more offensive touchdowns than they allowed in the Sun Belt, but finished just 3-5 in the league. With some breaks, they could overtake their in-state brethren and get to a bowl.

Saturday, March 03, 2012

Turnovers = Turnaround

Last time out, I published a piece examining how extremely lucky or unlucky teams perform the season after they experienced extreme fortune (either good or bad). Now I want to turn to another method that can be used to identify potential sleeper teams and teams that are likely to disappoint their collective fan base.

One of the preseason magazines I cannot do without is published by Phil Steele. In the fall, with the television on while on slaving away at this blog, I heard a commercial for a movie starring Hugh Jackman and thought they were making a documentary based on the college football guru. Imagine my disappointment when I discovered it was a derivative boxing movie featuring Rocky + robots. Oh well, I digress. One of the items Mr. Steele harps on in his preseason magazine is how big a role turnovers play in team performance. His magazine always features a section titled 'Turnovers = Turnaround'. Steele's contention is that teams with extreme turnover ratios (either good or bad) will see their margin trend back toward average the following season. Consequently, teams with great turnover margins tend to see their records decline and teams with poor turnover margins tend to see their records improve. I wanted to test his theory in a similar fashion to the way I tested close game performance. Thus, I looked at every college football team (excluding independents) from 2007-2010 and calculated their turnover margin in conference play. I looked only at conference games because teams often play extremely divergent non-conference schedules. Conference schedule strength on the other hand, remains relatively constant. I also decided to differentiate between BCS and non-BCS conferences based on the assumption that the non-BCS conferences would have greater shifts in year-to-year conference record thanks to a much less intrinsic ruling class. I then chose the most 'extreme' teams, those with a turnover margin greater than or equal to +10 or less than or equal to -10, and examined how they performed (in conference) in the season during which they had a great or poor turnover margin. Next I looked at how they performed the following season and calculated the difference. Here are the results. First the BCS conference teams with great turnover margins.
Of the 18 BCS conference teams with great turnover margins, 12 (or two thirds) declined by at least one game in conference play the next season. Four stayed the same and two improved. Overall, the teams declined by about 1.2 games in conference play the following season.

Now the BCS conference teams with poor turnover margins.
Of the 20 BCS conference teams with poor turnover margins, 14 improved by at least one game in conference play the next season. Two stayed the same and four declined. Overall, the teams improved by about 1.6 games in conference play the following season.

Now the non-BCS conference teams with great turnover margins.
Of the 19 non-BCS conference teams with great turnover margins, 13 declined by at least one game in conference play the next season. Three stayed the same and three improved. Overall, the teams declined by just a shade over one game in conference play the following season (-1.026). That number is not very far from the 1.2 game decline felt by similar BCS conference teams.

Now the non-BCS conference teams with poor turnover margins.
Of the 15 non-BCS conference teams with poor turnover margins, 11 improved by at least one game in conference play the next season. One team improved by half a game, two teams stayed the same, and only one declined. Overall, the teams improved by 2.3 games in conference play the following season. While the decline suffered by BCS and non-BCS conference teams was very similar, the improvement by the non-BCS conference teams is vastly superior (nearly 50% more) to that enjoyed by BCS conference teams.

So what teams should we keep an eye on in 2012? Who could be slated for improvement and who may fail to meet preseason expectations? BCS teams with great in-conference turnover margins last season include Oklahoma State (+17), Wisconsin (+13), Kansas State (+13), LSU (+12), and Arizona State (+10). Oklahoma State, Wisconsin, Kansas State, and LSU went a combined 29-5 against their conference foes, so some regression is certainly in order here. Meanwhile, Arizona State went just 4-5 in the Pac-12, so while they do not have as far to fall, they probably will not be contending for the division title either.

Only two BCS conference teams from last season meet the criteria for poor in-conference turnover margin. Ole Miss (-11) and Illinois (-10) went a combined 2-14 in their respective leagues. Perhaps not coincidentally, they also have new coaches for 2012. At the very least, those new coaches should see at least modest improvement (and they can take all the credit).

Six non-BCS conference teams had great in-conference turnover margins in 2011. Memphis (+12), Toledo (+12), Houston (+11), Louisiana Tech (+11), Arkansas State (+10), and Wyoming (+10) posted double-digit margins. Four of those teams (Arkansas State, Houston, Louisiana Tech, and Toledo) combined to lose just two conference games. Wyoming also had a solid year in the Mountain West, going 5-2. But Memphis...The Tigers went 1-7 despite a great turnover margin. Arkansas State, Houston, and Toledo all lost their head coaches after their successful seasons so the new hires may be unfairly evaluated if their teams dip somewhat.

Three non-BCS conference teams had poor in-conference turnover margins in 2011. SMU (-13), East Carolina (-11), and Florida Atlantic (-10) were all bitten by the turnover bug. Amazingly, East Carolina and SMU both finished with at least .500 marks in league play, a sign that they may be contenders in Conference USA in 2012. Meanwhile, the Owls from Florida Atlantic bottomed out at 0-8 in the Sun Belt, so there is no place to go but up for new coach Carl Pelini.