Good to have you back, Mathlete.
[Ed: this marks the debut of the Mathlete in an official capcity for the site.]
I guess when I was born and my parents named me Brett they know that I wouldn’t ever mean it if I said I was retired. An unwillingness to stay out of the game comes with the birth certificate.
When I last left we had no coach, no momentum and the school formerly known as Ohio State was coming off a big win over the SEC after avoiding serious repercussins from a minor tattoo incident. Today Hoke-a-mania rules all, a probable top 5 recruiting class is just about wrapped up and Ohio is staring into a great unknown with a new QB and head coach.
When I ran down the prospects from a host of candidates in January, Brady Hoke provided two nice charts about which I had this to say about his time at Ball State:
Better than I expected, actually. Slow steady growth taking the program from terrible to average over four years and then a big leap forward. The team obviously fell apart in two years under Parrish. The good news is that the team progressed well over a long period of time, the bad news is that during all the period at the helm, Hoke only produced one above average team.
Then at San Diego State:
San Diego St has shown nice improvement during Hoke’s time there. The Aztecs have improved by over 7 points each of the last two seasons. The big jump has been repairing a terrible defense (-11 in two preceding years) and turning them into an above average group by year 2. The offense has improved as well, but the majority of change has been driven by the defense.
So we have a track record for Coach Hoke at turning bad teams into good teams, what do the first year prospects look like for a new coach inheriting a team with a season like Michigan just put in the books.
The numbers you saw above and most you will see from me are based on my Points Above Normal (PAN) metric. It is a simple number that is exactly what the descriptor indicates; it tells you how many points above an average team a team or unit is. It adjusts for quality of opponent, excludes 1AA cupcake games and any plays where the lead is 17+ in the second half. +7 will probably get you in the top 25. +14 should put you in a BCS bowl and +21 is typically good enough for a title shot.
Last year Michigan finished +2 with a +10 on offense and a –6 on defense (the remaining gap comes from special teams). Because the offense/defense spread was the one of the ten biggest over the last 8 years, it will make more specific comparisons difficult and we’ll restrict the study to teams around +2 overall.
Since 2005, 18 BCS conference schools have undergone a coaching change after a season between +0 and +4.
|Team||Season||Conf||Change||New Coach||Old Coach|
|Arizona St||2007||PAC 10||7||7||1|
|Kansas St||2006||Big XII||0||1||1|
|S Florida||2010||Big East||0||2||2|
|Oklahoma St||2005||Big XII||-7||-3||4|
Dennis Erickson, Gene Chizik, Jimbo Fisher and Brian Kelly all pushed their new teams ahead by at least five points in their first seasons while Dan Hawkins, Turner Gill, Ed Orgeron and Mike Gundy all saw their teams take the biggest dips in year one.
On average, teams regressed by about a point per game in the first year of a new coach versus the previous year under the departed coach.
In general, a new coach coming into a BCS program coming off a season similar to Michigan’s don’t trend toward major changes in either direction, but some big swings have come under similar situations.
Regression to the mean will be the friend and foe of Michigan this year. The offense will be hard pressed to maintain the high levels of success and the defense will almost certainly make a step forward. The question is how much in each direction.
In my database of the last 8 years, there have been 58 BCS teams that had defenses within 2 points of Michigan’s lowly –6 from last year. Across those teams, the following year saw teams improve on the defensive side by about 4 points. 21 of the 58 teams showed improvements of a touchdown or more.
The offensive side shows similar numbers. Only 35 teams over the timeframe were within two points of Michigan’s +10 last season. Of those 35, 8 improved from there and the other 27 declined. The average change mirrored the defense at 4 points to average. The teams who were able to buck the trend were truly elite offenses. Of the four teams to go from Michigan’s range and improve by more than 2 points, two were the 2005 finalists Texas and USC, Oklahoma’s basketball on grass of 2008 and Florida in Tebow’s Heisman winning season of 2007.
Between a coaching change and where the offense and defense landed from last season, the strongest indicator that Michigan will move forward is in the success that Brady Hoke has had taking mid- and low-level programs, and consistently moving them forward. Beyond that, the optimistic scenario is ride the Hoke wave, Mattison gets a returner-heavy defense to make the leap and finding a combination of new and old on offense to hold on to all the success we can. A realistically optimistic outcome is probably a 5-point improvement generated on the back of the defense, a top 25 finish and the table set for 2012. The historical mean points to a worse offense, better defense but ultimately similar 7-8 wins.
No team in my database history has lost a coach after the kind of season (+16) that Ohio had in 2010. Five schools have been at least +10 and gone through a coaching change.
|Team||Season||Conf||Change||New Coach||Old Coach|
|W Virginia||2008||Big East||-11||1||12|
Of those five, Chip Kelly is the only one to push the team forward. Les Miles was able to keep LSU at a very high level while the successors to Brian Kelly, Rich Rodriguez and Bobby Petrino all saw significant drops in their first seasons on the job. A quick look says Ohio is most likely to fall somewhere between Miles and the Big East schools.
Good to have you back, Mathlete.
This is a fantastic addition to the stable of talent available at MGoBlog. You can count on my annual $50 in the beveled guilt dept. around Christmas time. Hopefully though, this doesn't mean the rumours of TomVH departing are true (good for Tom though).
read. As always, good work
It seems to me that the quarterback situation will be a contributing factor as well. There has to be some consideration for a returning starter versus a freshman recruit.
I'm just thinking about the Michigan and Ohio situations.
I'd think that's why OSU would be projected to fall back, but not all the way to mediocre. Regardless of how rough the situation is at OSU, they'll have the best average recruiting class (even subracting the suspended players and Pryor) of anyone they face unless things go really well and they make a BCS game.
How's life with the triplets? Hopefully you'll have time to catch some games this fall and give us your mathematical insight.
I appreciate the easily-comprehensible, yet meaningful metrics you use to take a look backward & take a stab at guessing the form of the future. pleasure to read :D
you weren't reading the data with your maize colored glasses on.
Remember, just like all weight gains/losses at this point are good, all stats that show a possible increase in success are fact, cause you know, math doesn't lie.
What accounts for Coach Hoke's first year woes at Ball State and SDSU?
What was the +/- for Ball St. and SDSU the year before Coach Hoke's arrival?
Would you be willing to generalize on where Coach Hoke would rank when at Ball St. and SDSU within your 18 team chart (as the background for a generalized sort of comparison - feel free to comment on the fairness), even though I imagine his first year seasons did not come after a season where the previous coach finished between +0 and +4?
The big question I have: how does Coach Hoke's first year challenges at Ball St. and SDSU compare to the conditions he faces this year with so many variables involved? (Feel free to offer us an opinion if you have one which relates to your statistics with use of terminology which is a little more concrete than "realistically optimistic".)
Also, my impression of Coach Hoke's performance based on my understanding of stats. you provide for him at both Ball St. and SDSU (especially if one removes the first year), is something I see not so much as slow growth, but a sky-rocket! I know he probably inherited below average teams. But, if an average team or a +2 team like Michigan begins to rise anywhere near as rapidly as the teams at Ball St. and SDSU under Coach Hoke (+19 in five years and +13 in three years)...would it not correlate somewhat (how strongly in your opinion?) to the differences between a +7, +14, and +21 team? Where does such an upward trend lead beyond 2012, and why not think in these terms, too?
There are few examples of good teams geting better under new coaches. Charlie Weis did it in 2005, Jimbo Fisher did it in 2010, Chris Petersen (the only one who took a good team to greatness) in 2006 and Gary Crowton in 2001.
Mathlete...can you link the explanatory diary for PAN? I don't see it on a quick search.
There are several posts where you run it down...IIRC you had an explanatory diary when you first started this metric...? NBD if it takes more than 20s...I was just curious when you started that analysis and the particulars of your method.
I think it stands for "Points Above Normal" and it is a standardized measure he created in order to assign simple but descriptive +/- integer values to different units (e.g. offense, defense, special teams).
Truly great analysis, as always.
The "regression to the mean" perfectly describes what I've been trying unsuccessfully to say any time someone asks me "How is Michigan going to be this year?". Of course the Offense will be worse. And the defense will be better. By how much? That is the question. I tend to think the offense will be talented enough to regress less than the defense will improve. The drive-killing inability to convert a 3rd and 10 and the ability to have a 12 play drive will help the defense as well. And improvement in special teams (we hope) will mean no more going for it on 4th and 5 from the other team's 30 yard line. Also helping the defense.
Thanks again and keep up the great work.
Is the concept of regressing to the mean really applicable to football teams? We're not talking about coin flips or lottery balls. It seems to me that success in football is mostly non-random, making "regression to the mean" a largley inapplicable concept.
I agree that M's offense is likely to be worse with the defense better, FWIW, but I don't think that that will have much to do with randomness.
Why is it presumed by everyone that the offense will be worse? Sure, from a statistical perspective, it may be worse for some of the following reasons:
1. less possessions because the defense will not be scored on so quickly
2. hopefully better field position, not requiring 98.5 yards to score
3. there will not be a need to score on every possession to keep up with the defense, so no Ilinois 67 point performance.
But, I don't automatically assume that our offense will be worse in terms of game performance. While we will not likely have 500 yards of total offense against our OOC opposition, perhaps the offense will be more consistent across the season, more able to score when the game is close or against B10 teams (see the OSU game).
I think that the difficulty of adjusting to a new scheme will outweigh the gains made from the maturity of the players on offense.
I don't think that will happen on defense b/c I think that the new defensive scheme and coaching will be vastly superior to the old one (while I think that the best we can hope for on offense on that front is a wash) and b/c the defense is gaining not just maturity but talent in the form of a healthy Troy Woolfolk and healthy JT Floyd. The talent, maturity, and coaching gains will outweigh the struggle of adjusting to a new way of doing things, or so I think.
I'm not knocking Borges or the offense. I expect every unit to struggle when it's adjusting to a new scheme.
Also, FWIW, I didn't label you as "flamebait," as you currently are.
On the struggling with an adjustment to a new offense, I totally agree. I do expect there to be growing pains, especially when the OL is asked to block man to man (which, even ifw do some zone, they will still be doing at least part of the time), and I definitely expect Denard to have a tough first few weeks, until he becomes "automatic" with such things as footwork, playcalling, etc.
All, fair points, and I guess in that regards, we will be seeing some regression.
Fortunately, aside from ND (the game that I am most afraid of on the schedule), our first 4 games are very winnable. By the time the B10 season starts, I would expect that most of the adjustments will be smoothed out and that we will continue to improve over the course of the season as ND did last year during Kelly's first season.
I hope you're right. I certainly wouldn't want to put a lot of money on betting against Denard.
7 wins if we are lucky.
You are very un-dude like
I love this kind of analysis. It's fun to let the Mathlete take us on a little data journey. Thank you.
Congratulations on becoming official. Was Brian wearing a robe at any point? Was there a paddle involved? Mathlete, were you ever placed in a coffin and then re-birthed?
Seriously though did they haze you?
My eyes usually glaze at the stats on MGoBlog even though I have plenty of math experience. Usually the metrics often require a bunch of acronym deciphering and leaps about normal distributions that I usually don't even try to understand them. This explanation was great:
Glad to see you are back Math. I think Brady Hoke is already improving the UofM football team in ways that are not tangible, and he will soon have the tangible results to back it up.
"A quick look says Ohio is most likely to fall somewhere between Miles and the Big East schools."
I would agree if they didn't lose tp and the other dudes for 5 games. They have to be worse statistically now.
one has to just hope for the best...