Love your work. Do you have a Hindsight analysis?
Hoke was top notch at this aspect of his job.
Once again, Connor Cook and I disagree on the Spartans outlook entering the season
This is the third year running I’ve crunched the numbers and simulated the schedule to see which teams I would buy and which I would sell. All predictions are based purely on numerical calculations I have run the last 5 offseasons to predict the following season and compared against Vegas’s win total predictions. The main inputs are players/ contribution returning, last season’s performance, recruiting rankings of players on the roster weighted by experience along with coach and program history.
There are only two teams in the Big Ten I have a major disagreement with Vegas on. 11 of 14 teams I’m projecting within 1 win of the Vegas total and Penn State is only a slightly larger gap. I may not be sold on Tim Beckman, but my numbers are high on his Illinois team this year. 762 out of 1,000 scenarios run have the Illini exceeding their projected 4.5 wins this year, with over half putting Illinois in line for a bowl bid.
Just like two years ago, one of my biggest sells of the season are the Spartans with only 5.5% of simulations seeing the Spartans exceed their 9.5 win projection. Michigan State’s three-year defensive run is downright amazing, especially considering how elite defenses are typically so reliant on highly touted talent being present. The Spartans have never had to replace as much as they do this year and although the talent pool has been rising, it is still another couple years away from four stars dominating the 2-deep.
[After the jump: mathy things.]
Some people think that the defense can regress because Connor Cook and the offense emerged in the last part of last season. Yes, the offense did improve, but mostly because the first four games couldn’t have been much worse. One quick check I do is to see how many points a team scored/gave up, versus what is expected based on each drive’s starting field position. This measure is opponent agnostic, just measures for each whether the points you scored verse allowed were more due to starting field position or value generated by the respective offense.
Here is MSU’s game by game look at the 2013 season:
The offense contributed more than the defense against Indiana, Illinois, Nebraska and Ohio State and less in every other game. Over the final six games of the season, the defense was +51 and offense was +7. The offense will likely be better this season, but the offense has a long hill to climb. I have no reason to think Cook can’t be a solid quarterback, but the idea that he is line to be the best quarterback in the Big Ten is purely a function of the defense wearing the same jersey as he does.
Ultimately, my prediction stems from a belief that its going to be very hard for the defense to be as elite as it was last year (I had them ranked #1 in the country) and I don’t see the offense getting enough better to cover the loss on defense.
Mack Brown after realizing he let me down again, or after realizing who he put in charge of his defense.
A big ball of mediocre. Even more pronounced than the Big Ten, I have everyone but BC (over 4.5 wins) within 1.2 games of the Vegas win total. On top of that, 9/14 teams are predicted between 3-5 and 5-3 in conference.
While my preseason prediction model has had some really nice sleeper calls, no team has foiled it more often than the Texas Longhorns. For the last two years it was convinced the Longhorns were coming back to national elite status but for it never to materialize. Mack Brown is gone because of it, and the model is just as high on Charlie Strong’s Longhorns. The model is predicting about 9 wins and a virtual tie with Oklahoma, right behind predicted frontrunner Baylor.
There is good consensus at the top between Vegas and the model, but at the bottom, the numbers like Cal a lot more and Colorado a lot less than the projected win totals. The top is packed tightly but my projected order is: Oregon, Stanford, USC, Washington and UCLA
After picking Tennessee to breakout in previous years, the model has given up on the Volunteers this season, along with Les Miles’ LSU squad. Two teams projected to overachieve, are league favorite Auburn, which despite a brutal schedule, the model pegs at 10.5 wins, a full 1.5 wins above Vegas along with the rebuilding Kentucky Wildcats.
Back in the spring I looked at the BCS era national champs to look for common threads. Each of the last ten ranked in the top in roster talent (recruiting rankings of the roster paired with seniority) with all but 2 teams landing in the top 4 on either offense or defense. Based only on the this secret sauce metric, your national title short list of roster talent is Top 4 O/D noted in ():
1. Alabama (O/D)
2. Texas (O/D)
3. Auburn (D)
4. Ohio St (O)
5. USC (O)
7. Florida St (D)
If you’re looking for some longshots, there are 4 full-fledged longshots on that list, including Michigan. Oregon, Oklahoma and UCLA are betting favorites not on the list, rank 16, 12 and 19 respectively in roster talent.
A Michigan contention for a national title is definitely a long shot, but nearly 20% of my simulations had Michigan with at least 10 wins, with the average outcome right around 8.
Love your work. Do you have a Hindsight analysis?
Is no. 10 supposed to read Michigan or Michigan State? 8 win projection doesn't sound like a top ten team.
As always, interesting stuff!
It's two different pieces. I am picking Michigan to win 8 games, but based on the National Championship Secret Sauce article, each of the last 10 national champs were rated in the top 10 of Roster Talent and Michigan is #10 this year in that metric.
Got it now. Thanks! Reading comprehension fail on my part.
To reconcile the two, the peak of Michigan's win distribution curve is lower (fewer wins) than MSU's, but the distribution itself is flatter. Lower projection, higher ceiling (but also equally higher chance of another unfortunate implosion)?
Michigan is, to be blunt, more unpredictable. Predictions vary wildly between 6-6 and 10-2, with most hedging at 8-4. The team is super talented and could threaten a NC if it all comes together, but too much needs to go right to expect that just yet.
Out of curiousity (and not looking back at previous editions), what was your projected win total last year for UM? Also, who do you have the 4 losses coming against? I'm guessing ND, MSU and OSU...who is the fourth loss?
Edit: Disregard my first question. I see now you predicted 8 wins last year.
I think schedule strength is part of it, but I'm pretty sure it doesn't actually run through game by game to get those win estimates. Could be wrong.
Given that he's projecting win distributions, there is no game-by-game "fourth loss." He's using a Monte Carlo simulation, so this isn't exactly what is happening, but to simplify things a little bit, let's say that he projects a 75% chance of losing to ND, MSU, and OSU, a 30% chance of losing to PSU, Maryland, and Utah, and a 15% chance of losing to each of the remaining six opponents on the schedule. This would result in a most likely estimate of four losses ((.75*3)+(.30*3)+(.15*6)=4.05), but there is no "fourth loss" picked out - just a chance of losing a game to enough different teams on the schedule that it's likely that we lose to one of them.
The systems engineer part of me is totally geeking out with this. Yay expected values
As noted above, I am mostly in line with the Vegas line of 8.5 wins for this Michigan season, but barring a Gardner injury, there is definitely more upside to downside. Michigan has 8 games where they should be a solid favorite. At their projected level, odds are that one of them finds a way to get away, but if they can win all 8, that leaves coin flip games against Nebraska, Michigan State, Notre Dame and Ohio State. If Michigan is better than expected at all, those 8 games should move to virtual locks and make a double digit win season a very real season. Without a major change event, there is a very solid downside firewall in place for this season, at least if you think, like me, that Penn State and Northwestern are generally overrated entering the season.
for Roster Talent?
Yes, they were 9th last season.
Now if you can tell me Notre Dame was in the top 10 two years ago, I will be convinced that we can lose the National Championship this year, and that would be much better than the board's current expectations - considering Brian's recent OL piece.
8 is the most common outcome, but there's certainly variation in that. The top 10 thing seems somewhat tangential to the win-loss outcomes, no? That top 10 list is for the amount of talent alone (star rankings and experience). It's one of the requirements a team seems to need to win it all.
EDIT: I forgot I loaded this article and then did other stuff without refreshing, so like 5 people have already answered you haha.
Also, we still don't have many upperclassmen, but we're already top 10 in roster talent. Next year should see us significanlty more loaded.
Thanks for your work
Great work. Yay statistical analysis.
Question: Is there any evidence to suggest that directives aimed at referees (such as PI, defensive holding, etc) have an impact? My guess is that it would be hard to specify that in a function but would be interesting to see.
the Mack Brown picture or the personification of the model "giving up on the Volunteers".
This is always one of my favorite articles. I love the analysis. I don't understand the top ten list though. Is that your projected end of season ranking?
Always love your posts. Any chance you can put up the numbers?
Also, does your analysis show which games Michigan is likely to lose?
Also, when you say Penn State is slightly out of the Vegas range, in which direction?
Sorry to be so demanding.
ALL OF THE UPVOTES
Looking at that graph, I would love to see our offense progress like MSU's did last season. It can even start out every bit as bad as long as the improvement's there. I'd love a good offense as well but I'm gonna protect my heart against a true frosh starting LT.
So I don't care if we can't even muster a TD against App. State as long as we win. I don't see their offense getting in the end zone against our back seven even if their veteran OL fights our three-deep DL to a stalemate. I know I know, complacency, Akron, Toledo, Teh Horror, etc., but FFS there's jinx and then there's reality. Mattison has the first two-deep defense since he arrived and this ain't the same App. State that came to Michigan off back-to-back national titles. If we win 15-6 but get a ton of IZ reps against a D-line that isn't wearing winged helmets I'll be happy as long as we get that sharp upward trajectory from WMU to Indiana that Sparty fans saw last season. Their first few games were so awful it dragged down their stats but they won most of them, over the last few games they held their own, and they wound up winning the Rose Bowl. I'll take that.
As for STAEE, I wouldn't count them out. I doubt they'll repeat a one-loss season but they're not replacing last year's defense with freshmen. The new group is probably just as coached up as the last, so I see them picking up another 1-2 losses not because of genuine issues so much as regression to the mean.
If we do ultimately follow State's season from last year I would be ecstatic. I actually asked that question in the Chicago Alumni football preview about whether it is possible. Of course I asked a similar question for 2013 regarding 2012 ND so go figure.That would be the best season we've experienced full stop since I've started following in 2006.However, there would be some justifiable angst on here if we only beat this App State team 15-6.
To clarify, I don't envision 15-6 where we pull all the stops. Again, a lot of MGoContributors intend to measure this season based on how the team plays, not the outcome. I'm in that camp, is what I mean by it.
I said I'd be "happy" with 15-6 if it becomes quickly evident that our defense is not only deep but able to keep a mediocre FCS team from moving the ball. Recent history acknowledged, but I maintain that's not a tall order. The defense is almost 3-deep so they're in no danger of wearing out. Those conditions in mind, this would allow Nuss to stick to base plays that -- hey, let's be honest -- still need work. They could go 30-40 yards a drive and settle for field goals all game as long as A) the OL makes the most of the experience and B) we win without using the back pages of the playbook. If we meet those goals I don't really care about the points. I'm happy with 40 but beyond wins I want to see progress, not desperation. If we score only 15 points with Nuss showing all his cards I'll be very upset.
The Akron game didn't piss me off because of the score, per se. It pissed me off because the O-line regressed and Mattison had to use one of his trump cards on the last play to seal the win. It was an aggressive called play. I'm sure he wanted to save it for later, but using that play was a symptom of a bigger issue. Ditto with UConn. I'm all for linebackers making circus catch interceptions in any situation but that we needed a circus catch interception to win the game was a bad sign. Replace the urgency with control and I'm happy to let the OL improve at whatever pace doesn't compromise the D's efforts.
Very interesting. It seems like your model values recruiting very highly, and coaching not as much (which I understand is supported by the data). Some of the teams that have bucked your predictions have the reputation of being coached very well or very poorly. Arguably Michigan State has benefited from good coaching, while Texas has suffered from bad coaching. Maybe exceptional or putrid coaching is real, but rare, and has a huge impact on a few teams?
There was more thought/commentary on the issue you bring up in the offseason "Secret Sauce" posts, but basically the idea that seniority is a good thing is partly an estimation of the importance of coaching (experience, coaching exposure, and physical maturity are all reasons why seniority matters), and managing a roster and program well enough to build seniority is a major part of coaching. So I am not sure I agree entirely that his model doesn't value coaching on a similar level with talent.
My thoughts as well, esp when I saw UM ranked #10 at the end. Considering it has had two top 10 classes in the past 3 I think he emphasizes recruiting very highly. In this regard I think the same analysis would not look so highly on UM basketball via the models. When in fact great coaching has turned it into a stellar program.
Of course, the analysis was purely NCAA Div I football-based, not meant to be applied to the NFL, NCAA BB, NCAA hockey, etc.
Also, from an empirical standpoint regarding preseason picks, we know that many top 10 teams won't stay there. There will be some really big surprise teams trending up and down. Last year bucky was supposed to go undefeated and they ended the seaon losing 2 and almost 3 straight. Sparty wasn't even ranked and was supposed to be in a rebuilding year with a new QB while Mich was picked by most to contend for the BgTn title. Also Gardner was supposed to be a dark horse heisman candidate. We know how that all panned out. Nationally, you had teams like Florda & Texas expected to be contenders had disastrous seasons while teams like Auburn, Missou, and UCF all unranked in the preseason ended up top 10 teams.
Looks like Florida has too much roster talent for Muschamp to survive another underachieving season.
Thanks for doing this, love your work. Just out of curiousity - where are you pulling the Vegas win O/U numbers from?
Well, he is The Mathlete and not The Grammarlete or The CorrectUsageOfApostropheslete after all
(seriously though, good stuff as always)
LOVE the analysis
srsly hoping that apostrophe is removed
the hype surrounding cook was always because of the defense. He seems to be a good QB but not one that can outright take a game over like Miller could.
Or Gardner even with one good foot.
MSU was somewhere above 2 deviations above a typical major conference defense last year. A regression to just 1 deviation above, while still excellent, will require a significantly better offense for the overall team to be elite this year.
MSU has an excellent program and may have taken the mantle from Wisconsin as the best "program" in the conference. They are great at talent identification and development in a way that Michigan hasn't been for 25 years. It's the reason Wisconsin has been good for so long and why OSU dominated the conference during Tressel's run.
However, I don't believe in that team as a national contender. They can certainly win the Big 10 and are probably most likely to do so. I just don't see them making a title run. The Big 10 is going to be a muddle this year and I'm guessing the winner has 2 losses in conference.
Sparty ... "They are great at talent identification and development in a way that Michigan hasn't been for 25 years. It's the reason Wisconsin has been good for so long ... "
I maintain that scheme is what has brought Sparty to unseen heights. Their defensive style is unique (rather like the Bears 46 defense of the 80's) in the college game. It's a high risk style that has seemed to escape most OC's ability to call plays. They recruit to this scheme. Same for Wisconsin, huge lineman and typically quicker RB's (no more Ron Dayne's - who never got 100yrs against us!).
Michigan is still developing its "scheme" so we don't even know if its going to work. 3 changes in schemes in 4 years will cause any program to sputter. Throw in AB (who I still hold personally responsible for the PSU loss - AND ALWAYS WILL) and you have a disaster.
I have a lot of hope for this season as we are starting to have REAL depth. I'll take the over +
It's not my primary hope going into the season, but I'll be so glad if you're right about State.
I'm am buying your selling of MSU. I don't believe that they are a powerhouse program here to stay and will win 11 games every season with top 5 defenses. I think they can probably maintain consistent 8-9 win seasons, but winning the conference and Rose Bowl are exceptions and not the norm.
Me too. History is filled with teams that were great despite relatively low talent, and rarely were they great for long. Putting together elite teams without elite talent is *hard*, and especially hard when you lose a bunch of multiyear starters from your team strength.
I don't think they'll be bad though, just merely very good instead of elite like last year. I think the closest analog is probably Wisconsin, who also consistently puts together good teams despite mediocre recruiting thanks to program stability and coaching. But Wisconsin rarely finishes in the top 5 (only once in the last decade, 2006, and only in one poll), and only slightly more often in the top 10. They're usually in the top 25 though, which is good for that level of recruiting. The top 10 years are the exception, not the norm, and I expect things to be similar for MSU.
Thanks for the analysis -
Connor and friend
This is overwhelming.
I remember two years ago when some MSU fans freaked out after the Mathlete selling MSU after back to back 11 win seasons. "But we beat Georgia in the bowl! Will Gholston!", they said. MSU went 6-6 in the regular season that year.
Fine work, Signor Mathlete. Complimenti.
So, what win totals do you project for each B1G team, and who have your top 25 projected win totals?
P.S. You're awesome.
Great stuff. Love the MSU chart, and its interesting to see that UM still has top talent despite down years. I suspect they'll start making strides this year, and 8-9 wins should be the expected outcome. I will say, Connor Cook being ABT would shock me, even with Miller out.
Not to be a hater or a homer but I don't see anyway Cook has a better than decent year. The D will be good in EL, we know that but it won't be as good as last year. Also you can only drill 12-13 defenders in the numbers before one of them catches the ball.