Tuesday, July 07, 2009

Introducing Negative Grohmentum

So I was reading Phil Steele the other night and ruminating on the demise of my beloved Charles Rogers Theorem when something dawned on me. I was looking at the conference stat pages and was reminded of the fact that, at this time last year, Sylvester Croom was the reigning SEC Coach of the Year. A lot of good that did him. And then I remembered that Al Groh has won the ACC Coach of the Year not once, not twice, but thrice. 20 minutes later, I was ploughing through my Phil Steele archive, which dates back to 2001.

An observation started to form: winning the coach of the year award is a curse because the winning coach's team almost inevitably regresses the following year. Negative Grohmentum was born. So how tight is the correlation between a head honcho winning coach of the year in a BCS conference and then watching his team get worse.


SEC

2008 - Nick Saban / Bobby Johnson / Houston Nutt

2007 - Sylvester Croom - 3.5 games worse

2006 - Houston Nutt - 1.5 games worse

2005 - Mark Richt - 1 game worse

2004 - Tommy Tuberville - 3.5 games worse

2003 - Nick Saban - 3 games worse

2002 - Mark Richt - 2 games worse

2001 - Houston Nutt - 1.5 games better

2000 - Lou Holtz - 1 game better

Total - six of eight teams regressed; mean of 1.5 game regression; median of 1.75 game regression.


Big Ten

2008 - Joe Paterno

2007 - Ron Zook - 3.5 games worse

2006 - Bret Bielema - 3 games worse

2005 - Joe Paterno - 2.5 games worse

2004 - Kirk Ferentz - 3 games worse

2003 - John L. Smith - 2.5 games worse

2002 - Kirk Ferentz - 1 game worse

2001 - Ron Turner - 5 games worse

2000 - Randy Walker - 3.5 games worse

Total - all eight teams regressed; mean of 3 game regression; median of 3 game regression.


Big XII

2008 - Bob Stoops / Mike Leach

2007 - Mark Mangino - 4 games worse

2006 - Bob Stoops - no change

2005 - Mack Brown - 3 games worse

2004 - Gary Barnett - 1 game worse

2003 - Bill Snyder - 5 games worse

2002 - Les Miles - 1 game better

2001 - Gary Barnett - 1.5 games worse

2000 - Bob Stoops - 2 games worse

Total - six of eight teams regressed; mean of 1.94 game regression; median of 1.75 game regression.


ACC

2008 - Paul Johnson

2007 - Al Groh - 3.5 games worse

2006 - Jim Grobe - 1.5 games worse

2005 - Frank Beamer - 1 game worse

2004 - Al Groh - 1 game worse

2003 - Tommy Bowden - 2 games worse

2002 - Al Groh - .5 games worse

2001 - Ralph Friedgen - unchanged

2000 - George O'Leary - 1.5 games worse

Total - seven of eight teams regressed; mean of 1.37 game regression; median of 1.25 game regression.


Pac Ten

2008 - Mike Riley

2007 - Dennis Erickson - 4.5 games worse

2006 - Pete Carroll - unchanged

2005 - Pete Carroll - 1 game worse / Karl Dorrell - 3.5 games worse

2004 - Jeff Tedford - 2 games worse

2003 - Pete Carroll - 1 game better / Bill Doba - 4 games worse

2002 - Jeff Tedford - unchanged

2001 - Mike Price - .5 games worse

2000 - Dennis Erickson - 5.5 games worse

Total - seven of ten teams regressed; mean of 2 game regression; median of 1.5 game regression.


Big East

2008 - Brian Kelly

2007 - Brian Kelly - .5 games better

2006 - Greg Schiano - 3 games worse

2005 - Rich Rodriguez - .5 games worse

2004 - Walt Harris - 2.5 games worse

2003 - Rich Rodriguez - .5 games better

2002 - Larry Coker - 1 game worse

2001 - Larry Coker - .5 games worse

2000 - Butch Davis - 1 game better

Total - five of eight teams regressed; mean of .69 game regression; median of .5 game regression.


Grand Total - 39 of 50 teams regressed; mean of 1.76 game regression; median of 1.5 game regression.


So what's going on here?

To a certain extent, this is good ol' regression to the mean. A coach wins coach of the year in a very good season and then his team will typically take a step back the following year. There isn't much room for improvement after a very successful season. A coach will usually win coach of the year when the voters determine that a program is at its absolute apex: an unbeaten season for Texas, eight wins for Virginia, etc. A team might do well with an experienced roster and then fall back to earth the following year with younger replacements. Schedule can also play into the equation. A team might have an excellent record because of a favorable schedule and then regress when they rotate to tougher opponents or more critical road games in the following season.

That said, these numbers demonstrate a misunderstanding that a lot of people have about what truly matters in college football: talent. I'm going to step out of the objective world for a second and speculate that the sort of coach that wins coach of the year is often one whose team was lucky. Not to keep picking on Croom or Groh, but they won coach of the year after their teams won a bunch of close games. The media looked at their teams and concluded "that team had no business winning eight games, so the coach must have done a great job." What the media should be saying is "that team had no business winning eight games, so they were lucky as hell and are going to take a step back." In other words, the coach of the year award is a pronouncement that a team really wasn't as talented as its record. It's an unintentional veiled insult that mistakes good fortune with good coaching.

There's no doubt that we have a strong correlation between a coach winning coach of the year and then his team getting worse. 78% of the teams in this situation this decade have seen their record regress the following year. 34% of the teams in the sample saw their record get worse by at least three games. By way of comparison, Phil Steele likes to look at net close wins and yards per point in finding teams that were especially lucky or unlucky in the previous season and are therefore due for a correction. (Page 299 if you're following along at home.) Teams with three net close wins have been weaker or the same the next year 76.7% of the time. Teams with 11.56 offensive yards per point or less have been weaker or the same 72.3% of the time. Teams with 19.85 defensive yards per point or more have been weaker or the same 77.6% of the time. Again, 78% of the teams whose coach won coach of the year have been weaker (not just weaker or the same, but weaker full-stop) the next year.

Random Thoughts on the Data



So which teams are the lucky ones who have a four-in-five chance of seeing their records get worse this year? Alabama, Ole Miss, Vandy, Penn State, Texas Tech, Oklahoma, Georgia Tech, Oregon State, and Cincinnati.

Negative Grohmentum has been especially pronounced in recent years. In the the last four seasons, the only team with a reigning conference coach of the year that has improved was last year's Cincinnati team, which went from 10-3 to 11-3.

Negative Grohmentum is also especially pronounced in the Big Ten, mainly because the voters have not given the award to Michigan or Ohio State coaches this decade, most likely because of an assumption that those programs have natural advantages, but those natural advantages make those two programs the consistent winners in the conference. (Insert "3-9!" joke here.) Michigan and Ohio State have also been consistent winners this decade and Big Ten voters appear to be attracted by major changes in a won-loss record. How else does one explain the fact that Jim Tressel's teams have won or shared four Big Ten titles, but he has never won the conference coach of the year. Meanwhile, Joe Paterno has won the conference coach of the year twice because his teams have followed the decent-decent-very good-decent-decent-very good pattern, whereas Tressel's teams have been consistently very good. (Insert "SEC Speed Killz!" joke here.) Tressel is penalized because he almost never has to come back from a mediocre season. Illinois has finished over .500 exactly twice this decade and its coaches have won coach of the year both times. "Hey, you've taken a program that is routinely referred to as a sleeping giant and won enough games to make a bowl in which you'll be slaughtered. Here's a trophy for your trouble!" In case you're interested, the last Big Ten team with a reigning coach of the year that did not regress: the 1992 Michigan Wolverines. 1992 is also the last season in which a Michigan or Ohio State coach won the Big Ten coach of the year, despite the fact that those programs have won or shared the conference title in 12 of the 16 seasons since.

Urban Meyer: two national titles in four years, no SEC coach of the year awards. Hell, does anyone want to take bets on whether Florida mimics the '95 Huskers this fall and Meyer is beaten out for coach of the year by Bobby Petrino because Arkansas goes 8-4?

Coaches who have won three coach of the year titles this decade: Pete Carroll, Bob Stoops, Houston Nutt, and Al Groh. The former two are on just about every list of the best coaches in America; the third has been essentially fired; and the fourth is on the hot seat.

By my unofficial count, 12 of the coaches on this list have been fired or forced out of the jobs they held where they won coach of the year: Nutt, Tuberville, Croom, Smith, Turner, Barnett, Bowden, Dorrell, Doba, Erickson, Harris, and Coker.

4 comments:

Senator Blutarsky said...

You had me at "Grohmentum".

Jesse said...

This is very interesting, especially in the case of Paul Johnson and GT. If net close wins is a FG or less, then GT had four such wins and one such loss (though I'm not sure how much close losses play into this), but if it is a TD or under then that number goes up to five such wins and two loses. Taking the percentages you already posted, combined with what some already consider a more difficult schedule than last year as well as the major losses on defense, and it looks as though Paul Johnson and GT are heading for a regression in wins this year.

However, if my math is correct, GT had 14.28yds/pt offensively and 14.21yds/pt deffensively, both of which exceed the stated break point favorably for them to not regress, or at least the chance for them to not regress. Considering that it was the first year in Paul Johnson's offense, the first year in the new deffensive scheme, with an all new coachng staff, I would think that the chance of GT staying the course or getting better would be a likely outcome.

Obviously, neither of these chances can ever take into account blind luck (turnovers, etc) or injuries, but it does give you a good starting point when making somewhat subjective projections. I would be interested in seeing where these numbers place GT percentage-wise with regards to possibly regressing.

Also, do your numbers account for first year coaches or switches in offensive designs? I was just wondering if there was a mean to regress to if those were the circumstances involved when the coach was selected as COY. I'm not sure it would be correct in assuming the GT would regress to the standard mean that was when Gailey was around. Just a thought though.

Dawg19 said...

Well done, Sir.

Excellent read.

consejo comprar yate said...

Here, I do not actually consider this will have success.