Nice effort and welcome to posting.
I thought that myself when I read that article that talked about a Data Scientist(tm)
[Editor's note: I am headed to a wedding in Milwaukee this weekend, and thus won't be able to put together mgo-worthy content for the rest of today. Instead, enjoy this excellent, research-heavy diary.]
First time poster, long time lurker. While procrastinating on various work projects, I’ve been dithering around with a data set on college football win-loss records. I’m the sort of guy who actually thinks HBS case studies are kinda neat, so futzing with this seemed like fun in its own twisted way. Then one of our fellow mgobloggers put up a really nice monte carlo simulation of the 2009 season (using a $500K piece of software no less) and I felt a little guilty about not posting some of my stuff online. So, here goes …
What sent me down this path was the whole topic of what our expectations should be for next year’s win total. Seems like the general consensus is somewhere between a 3-5 win improvement for next year. I am a Bill James devotee, so I began to wonder how realistic that was in a historical context (i.e., how many teams really improve that much in one year). My instinct was such occasions were not all that common. So, I spent a couple of hours pulling some data (30 years worth of W-L records for every D1A team, to be exact). Here are some summary conclusions and some things that I intend to research a little further and post about whenever the procrastination bug strikes again.
Turns out that big improvements in win totals from one year to another are more common than I thought. There were roughly 300 such cases since 1980. Considering that the data set is about 3300 team seasons, I thought this was pretty remarkable. In essence, the average team has at least one 4+ win improvement season every decade. Score one for optimism here.
Digging a little further, I took a look at extreme win total improvements (+6 wins or more). There were quite a few of these as well – 63 (or more than 2 per season). And they weren’t all MAC and Sun Belt teams either. 28 of those seasons were from teams in one of the six BCS conferences. The Big Ten had 6 such seasons:
For those dreaming about the possibility of warm weather for New Years, at least there’s some historical precedent. Also, RichRod is responsible for one of those 63, West Virginia '02, which was +6. Which also happened in his second year after a 3-8 first season. I’m not sayin’, I’m just sayin’ …
For those who are curious, the biggest turnaround in the study (and most likely in college football history) is Hawaii 99 (+9). At +8 were Florida 80, San Jose St 86, Bowling Green 91, South Carolina 00, and Central Florida 05.
I also was wondering if the optimism about a BCS game in two years was rooted in reality. [Editor's note: I would have gone with "vague hope" instead of optimism.] For this, I am assuming the requirement is an 11+ win season. So, if we start with the assumption that this year ends up at 7 wins, we need another 4 win improvement next season to reach that.
Here, the historical precedent is less encouraging*. In the study there are only 5 cases of teams recording 2 consecutive seasons of 4+ win improvements: Colorado St in 89 and 90, Fullerton St in 84 and 85, Georgia Tech in 89 and 90 (culminating with a shared national title), UNLV in 83 and 84, and one other.
This last one should be your reason for optimism, as it is Tulane in 97 and 98. With OC Rich Rodriguez. 2-9 Tulane 96 became 12-0 Tulane 98 under that offense. The historical numbers may be stacked against us, but this staff has defied them once before.
Lastly, I wanted to highlight the three Michigan seasons that were caught by this analysis: 85, 97, and 06 (all +4).
See any themes here? If you named the three best defensive teams of the last 30 years, this is probably the list (with the possible exception of the 80 team). Does this sound like the 09 team to you? Yeah, me neither. Bummer.
I do see some decent parallels to the 85 and 06 teams, though. Both the 84 and 05 teams were plagued by injury hell and were extremely young, as was 08. Both the 85 and 06 teams ushered in a new era of offensive strategy (Bo discovered the forward pass with Harbaugh in 85, the zone stretch changed the running game in 06). For all intents and purposes, 09 is the first real glimpse we’ll have at RichRod’s full playbook so I’m willing to buy that as a philosophical change.
The defense going into both 85 and 06 had only two established stars (Mike Hammerstein/Brad Cochran, Lamarr Woodley/Leon Hall) and a bunch of question marks. Could Graham/Warren count as established stars? The 85 and 06 teams had unheralded defensive players become stars (Mark Messner, Andy Moeller in 85, Alan Branch, David Harris in 06). Could that happen here? Mouton? Martin? Spinner/deathbacker to be named later? Surprise freshman stud (Turner? Campbell? Emelien?) Maybe it’s a stretch, but if you want to be an optimist, I think this is what you look to …
So, that’s all I have for now. I am going to take a closer look at the impact of coaching changes on the big spikes in W/L. Current hypothesis says you get the biggest pop in year 2, but let’s see what the data says. Also going to look at the other side of this coin, seasons of -4 wins or more. Could give some retrospective insight into the whole WTF situation that was 08. If you want me to look at anything else, I am open to suggestions.
* Really this should look at 8+ win improvements over two years, not just consecutive 4+ win years. However, it is late and I’m too tired to do that now. Maybe next time.
[Editor's note: Something struck me as I read this: check out those bounce-back seasons there. Minnesota was 1-11. Illinois was 2-10. Northwestern was 3-7-1, Purdue 3-8, Penn State 4-7. Only OSU -- 7-5 in 2001 -- went from mediocre to very good, and the 2002 OSU team were the luckiest sonsabitches in recent college football history.
Everyone else was bouncing up from horrible to average, which seems much easier to do than to go from average to very good. So, yeah, a crappy bowl beckons.]
Nice effort and welcome to posting.
You seem to be a well informed statistician and researcher, (the fact that you are doing this at 2:00 am is a little freaky but whatever). The post was filled with facts and a lot of optimism which I tend to like to read (especially when it is about my beloved UM). All in all a well written article and I hope you are right about the 2 season improvement. I would love to see UM get back to a BCS bowl, but more than that I would love to see them win a BCS bowl as that has not happened in quite some time not to mention the rest of the B11.
One question: Do teams really need 11 wins for BCS games? I know that would be ideal, but last year OSU got to one with 10 wins and VT with 9. Not that I expect us to go to a BCS game in 2010 but maybe the road to one doesn't need to be that tough.
10 wins is usually enough for a BCS conference team to get into a BCS bowl as an at-large invitee.
I think we have an outside shot at getting 10 wins in 2010, however we play ND, PSU and OSU on the road so we will need to win at least one of those "tough" games. I think PSU will be the most vulnerable in 2010 as D. Clark and E. Royster (if he goes pro early as I suspect) will be gone. We may be facing sophomore Kevin Newsome at QB for PSU in Happy Valley come 2010. I would love to pummel PSU with Newsome at the helm.
Nice effort and welcome to posting.
First of all, fantastic post. Thanks for all the work.
You mentioned that the teams Michigan has had go +4 from one season to the next were some of the best defensive teams we've had. I wonder if the reason that needed to be the case was because of Carr's offensive philosophy. Having a defense he could really lean on meant that he could afford not to take chances towards the end of a game if he didn't have to.
Does that change with Rodriguez? Clearly defense is incredibly important, but does his philosophy of scoring whenever possible reduce the pressure on the defense at all? It seems that with Rodriguez we may be able to get that kind of improvement without having the best defense in the country.
It also doesn't hurt that the +4 threshold is being taken from a baseline of 3 wins either... much easier to go from 3 to 7 than from 6 to 9 imo.
"Does that change with Rodriguez?"
I would argue that it does, as point differential still decides football games. The more points you can score, the more you can afford to give up.
Now that is what sets this blog apart. I am married to a woman who majored in stats and is working on a graduate degree in bio stats so I appreciate what you did here.....I have no clue how you stats people do it*, but great work. And honey, if you are reading this, thanks for winning the bread too!
*By it I mean crunch numbers/data for hours. The only thing I crunch for hours are potato chips. Maybe that is why she brings home the better paycheck......hmmmm....maybe I am better at analysis than I thought....I can do this stats thing! It would be really cool if I could give you the standard deviation of chips:body mass.
That was an impressive post. Not just the magnitude of the number crunching, but what you analyzed and the interpretation of the results. This provides some more hope for '09, but also recalled an old Woody Hayes' quote: "Statistics always remind me of a fellow who drowned in a river where the average depth was only three feet."
Excellent work, I think. Of course, I wouldn't know bad statistical work if it bit me on the leg, but an unqualified compliment is still a compliment after all. I think.
Anyway. I know everyone is intent upon finding rays of sunshine and optimism. And that is kewl. The spring game geeked up a whole slew of fans. I am not in this camp, however. To me, the spring game was like a carnival game. It was rigged. And, how can anyone judge Tate's skills when he was immune from being hit? Every opponent is going to target our QB and try to rattle them, maybe even break him into two separate parts. I mean, i guess he could rush for 1,000 yards and pass for 3,000 yards and cure halitosis and kick Chuck Norris' ass. I hope so. I don't really like Chuck Norris. But then again, if he goes all Sheridan or Cone on us, oh well. Wait till next year, right?
So, nice post. Hope it actually means something. Maybe you could find some stats on how often freshman QB's get kilt.
is the whole "regression toward the mean" concept. That is, although a 5 win improvement next season may seem like a statistical outlier to some extent, for Michigan it would simply be a return to their average performance. Stated differently, it was last year's record that was the clear statistical outlier for Michigan, so an average Michigan season in comparison will appear as an extraordinary jump in terms of win/loss record, even though for Michigan it would still just be an average season. You already sort of touched on this with the "other side of the coin" idea at the end of your post. But it should not be particularly surprising to find that great improvements in the win/loss column are most common, and are probably in fact quite likely, following an unusually bad season (again...regression to the mean).
I certainly would agree that regression to the mean is probably a big explanatory effect in many cases.
I think these three factors will describe most of the season to season improvements across teams:
A Team with an established high level of performance falls off for one season for some combination of factors and then snaps back
B Program undergoes a successful transition in philosophy from a coaching perspective and suddenly gets better results
C Marginal program gets lucky and winds up with an elite player who temporarily takes team to new heights
Just looking at the Big Ten data, Minnesota 08 fits A from a trend line perspective, although there was a coaching philosphy change in the mix as well since Brewster's system is very different from Mason's.
Ohio St 02 and Penn St 05 could fit explanation A as well, although the established performance level for both schools was considerably lower in the 2-3 preceding years than their historical reputation. Ohio St also seemed to benefit from a coaching change, as Cooper seemed to be losing some institutional control at tOSU and may never have had the ability to reattain his previous successes in the mid-late 90s. IIRC, JoePa started delegating a lot more to his staff around 2005, which could be called a coaching change of sorts.
I hate to give explanation B for Illinois (Zooker as genius?) but he did raise the talent level across the board rather than just get lucky with one player, so B is a better fit than C.
Purdue and Northwestern were clearly B.
C type explanations are almost by definition for non-BCS schools.
Another study for another time, I guess.
There isn't much in this squad to instill confidence in me that this will be a top defensive team, but you never know. I don't think that much was expected of the '97 team and look what they did. Also, I believe that was Hermann's first go round as the DC if I'm not mistaken. Coach Robinson has been around the block and has had some success molding good defenses; in terms of choices to replace Schaffer, I'm not sure we could have done better. I think his experience will complement both the squad and the coaching staff.
Offensively, if Tate stays healthy and Denard gets some good experience, I do have confidence we will be successful. I know there are some folks out there that will disagree with my sentiments-with good reason, we are coming off 3-9- but I feel 7-5 is readily achievable and that expectations should begin there, not 6-6.
Nice effort and welcome to posting.
Triple posting hours apart? WTF?
he's a robot
this helps my pessimism
I, like you in the past, have just stuck to lurking about ommenting on others work but feel an entry in my furture given that I am finally over 20 points (at least at the moment).
I have been quite optimistic about this season for some time now and your hard work has given me even more reason to think that we should have a much improved season in '09. I think with the talent of the new recruits and the experience of the vast numbers of returning players Michigan should make some very big strides this year and hopefully, by 2010 will be in contention for the Big Ten Title again (maybe not winning but at least in contention).
Reality: we're starting a true freshman QB, it's looking like our defense will be improved but maddening yet again, and we're probably headed to the Motor City bowl at best.
Fantasy: Tate is accurate as F(!) and can run well(!), the running backs are experienced and deep, the offensive line will be experienced, G Rob is one hell of a DC, outside of OSU and PSU the rest of the B10 is average, and we're headed to a late-December, dare I say, January (?!?) bowl.
I really love my fantasy world but I have prepared myself to accept reality world.
but, yeah, what others have said: great analysis. Thank you. This goes to others posting similar great posts in recent weeks (and to Bri-guy, without whom I'd be wandering alone in the ewilderness).
What is the 6-30-08 circle of trust?
It is the date that Mgoblog went from the old format to this one. I added my signature tag (as a joke) when there was much weeping and gnashing of teeth about the new point system and how those who stayed should get perks, etc., etc.
did we decide on a secret handshake yet?
at the super secret annual meeting.
Is there any way to find a list of who is in that special club of "first day" account makers?
... live in the South now and although found some good regional beers, this makes me pine for some days gone by. Do they still make Two-Hearted?
...is to look at how many teams worsen to the extent Michigan did from 07 to 08 and see what the average rebound time is. I realize 07 might be an outlier year, so maybe take M's average record over the previous 10 years (I would guess about 8.6-3.2) and use that as the baseline. As a previous commenter alluded to, being a crappy team with a ringer is different from being a perennially good team with a down year. I'd be interested to see this type of analysis.
I would agree with you if you had been starting Forcier last year.
Freshman quarterbacks, fit for the system or not, do not have a stellar history. Even Pryor (who had one of the best records of freshmen quarterbacks in history last year) was playing with one of the best defenses in the country, and with one of the best running backs, and experienced receivers. And he still looked really bad much of the time.
I expect Forcier to be an improvement over 3-9, but with the defense definitely still in "rebuilding" mode, I'd be very surprised to see him do anywhere close to that well. 6 wins is a distinct possibility, but I'd be much more optimistic about the "+8 in two years" possibility than the +5 in one year one. Even with some good improvements, you're looking at likely losses versus Penn State and Ohio State, and very possible losses against Michigan State, Notre Dame, Illinois, Iowa, and possibly Wisconsin depending on just how much of an idiot Bielema is.
Still, good luck. I'm excited to see what RichRod comes up with when Forcier (or Robinson, or whoever) has a year of experience under his belt.
Excellent post, and the supporting analysis just bolsters my optimistic expectations of coach Rodrigues. He has done his thing and brought in his changes and after a period of transition has taken programs that were once mediocre to bad and made them respectable to excellent. Each time he has done this he has improved his talent and also upgraded his competition. I beleive that he has his sights set on nothing less than putting himself as a coach in the class of Pete Carol and building a program that will be a modern juggernaught.
Even though the gradual malaise of the program was exposed and the transition to a very intensive agressive program was bumpy and the results on the field were terrible, on the whole I like the "feel" of what was happening underneath the surface. Ken Dryden, one time Montreal great and former Leafs president once made the point that often the teams that have breakout seasons tend to suprise people when they do, as the "on ice" results up to the point of the breakout were disapointing. But for those really paying attention, it is all the little things happening beneath the surface that then come together and bubble up all at once.
Coach Rodrigues strikes me as someone obsessed about coaching, fundamentals, and the work ethic of his players, all that beneath the surface "team culture" stuff. When all the pieces are in place and the cultural change has taken place...this program will be a machine...it is all there right now happening beneath the surface.
If you looked at the entire field of teams that underwent a +4 transition, I imagine there's a very high correlation (>75%) with the baseline win total being 4 or less. It's much easier to improve when you're empirical crap to begin with.
So if a +4 transition has a very small correlation with a baseline total of 8 wins or more, doesn't that sort of already answer the question about the BCS in 2 years? Within 2 years is probably not going to happen, folks.
What about in 3 years? Well, I imagine there's probably a much higher incidence of going through a +4 transition over 2 years, and I would be very interested to see how often that happened if the starting baseline is, say, 7 wins. Again: I'd be very interested to see how often a team with a baseline of 7 wins undergoes a +4 transition over the next 2 years (hint hint please do more of this hint hint).
Fifth-year senior producers (2004)
Tim Jamison, Will Johnson, Morgan Trent, John Thompson
Fourth-year producers (2005)
Terrance Taylor, David Moosman, Zoltan Mesko, ~Kevin Grady, ~LaTerrell Savoy
Fifth-year senior producers (2005)
David Moosman, Zoltan Mesko, ~Kevin Grady, ~LaTerrell Savoy
Fourth-year producers (2006)
Brandon Graham, Brandon Minor, Stephen Schilling, Obi Ezeh, Jonas Mouton, Steve Brown, Carlos Brown, John Ferrara, Perry Dorrestein
With all due respect to the individuals, in aggregate that's one of the smaller talent graduations we've seen at Michigan.
I hate to admit this, but I am having nightmares of a 5-7 season. I want UofM to go undefeated obviously, but I am worried about playing a true freshman QB, as well as a few other positions.
I think we can do much better than a 5-7 season, but I also see the potential for a lot of pain.
If QB play isn't noticeably better than last year we could look at these odds...
WMU -- slightly favored
Notre Dame -- slight underdog
EMU -- favored
Indiana -- slightly favored
@MSU -- slight underdog
@Iowa -- slight underdog
Delaware State -- heavy favorite
Penn State -- underdog
@Illinois -- slight underdog
Purdue -- slight favorite
@Wisconsin -- slight underdog
Ohio State -- heavy underdog
WMU -- slightly favored
Notre Dame -- slight underdog
EMU -- Auto-win (or Ann Arbor burns)
Indiana -- Clear favorite
@MSU -- underdog
@Iowa -- underdog
Delaware State -- Auto-win (or Ann Arbor burns)
Penn State -- heavy underdog
@Illinois -- slight underdog
Purdue -- Clear favorite
@Wisconsin -- slight underdog
Ohio State -- heavy underdog
Rivals rankings in parenthesis (Mich is 46)
2 Auto-wins EMU(117) & Delaware St(FCS)
2 Clearly probable wins Indiana(86) & Purdue(76)
1 Should win WMU(58)
3 slight dogs ND(25ish?), @Ill(35), @Wisc(49)
2 dogs @MSU(25ish), @Iowa(20ish)
2 heavy dogs PSU(10-15ish), OSU(5-10ish)
If we open with a win vs WMU, I think five wins are fairly safe with seven chances to pick a few more off. I'm predicting 7-5 most likely, with a seat-of-the-pants, rounded distribution of:
3 wins or less (03%)
4 wins (10%)
10 or more (02%)
But although this seems encouraging on the whole, I would wager that there were far more teams that do not improve 5+ games after losing seasons. I guess the overall probability of teams that do show that kind of improvement is around 20% (3300 team seasons, 300 instances of teams with 5+ games improvement, 1650 losing seasons? and approx probablity = 300/1650 ~= 0.2). So there is still an 80% chance that we don't see that sort of improvement... I think.
The reason this is because is that a team that does that badly has a number of systemic flaws that do not usually get corrected in the course of a single summer. I am encouraged by this post (fantastic post as others have pointed out), but I will still be extremely pleased with anything approaching a .500 season with incremental improvements in years to come.
Again, thanks for the effort!
many people seem to underestimate the running game. RR's offense is based on the running game (better OL, more RB options, etc.), therefore if Tate displays better decisions, ability, than last years QBs, UofM should see a minimum of 6 wins--even with the questionable D.
Great job on the post with all the statistics. Basically I think that we will probally win around 6 to 8 games this year. If Forcier can just make opponents respect the pass we should be much improved.
Next season will be a struggle to get to 5-6 wins. Yet the defenders of Rodriguez will still be as fervent as ever as we spiral further and further down the drain.
Sorry, still not drinking the Rich Rodriguez Kool-Aid.
Does that taste like chicken? ...after all... Maybe the Machines couldn't figure out what chicken tasted like and that is why everything tastes like chicken. Makes you think doesn't it?
I think 5-6 wins is probably about right. We're a lot like Notre Dame last season, except we're starting a true freshman QB.
The big thing -- I just don't see where the big improvement on defensive is going to be. Our defensive line won't be as good. Our LB play last year was pretty atrocious, and we're banking on big improvement from them. But they're not a deep group. Dear God, let's not even bring up safety play.
On offensive we should be better at almost every position -- especially offensive line. But I think our ceiling will still be significantly limited by freshmaniness of Tate.
Unfortunately, I think it's Motor City bowl at best, home fot the holidays at worse.
I'll put the over/under at 5.5 wins.
I'm hoping the big improvement on defense comes from the offense. I pretty much agree with your assessment of the defense except that I'd say Mouton had a pretty good second half of the year.
I'm hoping that an improved offense will help mask what will probably be at best an average defense. By improved offense I'm not talking a juggernaunt but rather one that hopefully averages more points, has a lot fewer 3 and outs to give the defense some rest and help with the field position battle, and doesn't turn the ball over as much (special teams would fall into the last catagory as well).
...is team camaraderie and unity. You won't find that too often at the pro level but it's important at the college level and critical at the high school level.
I've waffled back and forth about the upcoming season and the whole RR regime. One day I'm optimistic and the next, I'm like, "Fuck, we are SO screwed."
But, one thing I KNOW RR has done in the off season is improve team unity and camaraderie tremendously. I've been to a couple spring practices. The kids (AND visiting recruits) love the guy. He relates to them. He communicates well with them. He's demanding as hell but not an asshole (for the most part). And most importantly, the kids want to PLAY for him.
One statement from last year that really struck a chord with me was when Brandon Minor said (and I paraphrase) "we got rid of a bunch of guys who didn't buy into this system."
That speaks volumes, imo. It's something you don't often hear or read (especially coming out of A2), but it made perfect sense. We had some talent, but we had guys who went through the motions and were just trying to play out the season.
The 84 team was a disaster. One thing Bo harped on after that season was the complete lack of senior leadership. From that day forward, it was "The Team. The Team. The Team."
If there's one thing I'm optimistic about for the upcoming season, it's better TEAM PLAY. If these guys have bought into the system, and are willing to "pay the price", I think we will be sufficiently pleased with the improvement of the 09 team.
I'm certainly optimistic for the team showing some improvement in the W/L column. Even with the comedy of turnovers, with the worst defense most of us have ever seen, with all that went on the only thing separating us from a 5 or 6 win season was halfway decent quarterback play. If we'd received that we probably win Utah, Toledo and Northwestern.