this may be of some local interest
Support MGoBlog: buy stuff at Amazon
m1jjb00
Scoring Draftageddon via All B1G/AA teams
I get that few people enjoyed Drafageddon ’14, and even fewer are interested in revisiting the topic. Maybe one interesting piece out of it is the number and quality of people who weren’t picked, some of whom came from nowhere, as both a coda on the 2014 season and early hint at 2015.
To score the teams, I use the Allconference and AllAmerican teams. They provide an objective, easy metric to judge performances that also treat all positions similarly. Anyone reading this is now listing all the problems with their use. I agree, but I think they’re still the best option.
As a scoring system, each pick gets 3 points for making one of the coaches’ or media’s first team AllBig 10 team for a possible maximum of six, 2 points for each 2^{nd} team, and 1 point for honorable mention. In addition if a player averaged better than second team All American among five AllAmerican teams (AP, Walter Camp, USA Today, Football Writers, and Sports Illustrated) he gets a twopoint boost. If a player otherwise averages better than third team (or honorable mention) he gets one point. For instance, Brad Craddock was second team AA on four teams and first on the Football Writers’ team; hence he squeaks by with two extra points, something that is way more important than I thought it would be when dreaming up the scoring system.
Based on this scoring system BISB wins the season, which was also the consensus at the start of the year. Ace beats out Brian by one measly point. Seth’s strategy of a hyperspread team was not rewarded by the All Conference teams as Maxx Williams easily outpointed his entire set of skilled offensive players. He did, however, have the second highest defensive point total. Injuries and suspensions were obviously an important part of the outcome.
Draftageddon 2014 


Points 
B1G 
AAmer 
Ace 
65 
58 
7 
BISB 
69 
63 
6 
Brian 
64 
60 
4 
Seth 
49 
47 
2 
2020 
87 
87 
0 
The four Mgo’ers left a lot of points on the table. A perfectforesight team (2020) picked at the end of the season from those left over includes six guys who made at least one 1stteam All B1G team. Lots of points could have been had by showing more love for the Gophers. On the other hand, too many Michigan players were taken. Nine were drafted who didn’t make any All Conference teams; Northwestern came in second at five. Wisconsin, Maryland, Michigan State and Ohio State had four such players.
Quarterbacks

Ace 
BISB 
Brian 
Seth 
2020 
QB 
4 
0 
0 
0 
6 
JT Barrett was consensus first team, unpicked by the four because, obviously, given OSU QBs’ proclivities for injury there was the risk that he would drop too early and Cardale Jones would scoop up all the points. Connor Cook was consensus second team, and no other quarterback merited honorable mention. That’s probably the best defense of the All Big Ten teams as a scoring system.
Running backs

Ace 
BISB 
Brian 
Seth 
2020 
RB 
5 
10 
8 
0 
5 
This position is the exact opposite of quarterbacks with three guys earning extra AA points. The only points undrafted were David Cobb (MIN) as consensus 2^{nd} team, and Elliot (OSU) and Jackson (NW) with one honorable mention. Institute a rule for next year that an Mgo’er automatically gets the backup in case of an “injury”, and Seth would have been on the board with Jackson.
Wide Receivers

Ace 
BISB 
Brian 
Seth 
2020 
WR 
3 
6 
9 
2 
10 
Carroo (Rutgers), Dudek (Illinois) and Hamilton (PSU) make a nice receiving corps, though eschewing freshmen receivers is understandable. Lippett was an inspired choice relatively late in the process. In retrospect there were too many Spartans, Terps and Wildcats chosen at this position.
Tight ends

Ace 
BISB 
Brian 
Seth 
2020 
TE 
7 
0 
4 
2 
2 
Ace’s Maxx pick looks better on this scale than others thought in real time. It wasn’t a banner year for Big Ten tight ends otherwise. Heuerman, who was a consensus 2^{nd} teamer, had 17 receptions for 207 yards and 2 touchdowns.
Offensive line

Ace 
BISB 
Brian 
Seth 
2020 
OL 
12 
17 
19 
16 
12 
I thought that Ace had the best oline originally, but his picks of Johnson (RU), Marz (UW) and Walsh (Iowa) earned only four points. Lindsay didn’t see the field. All but one of Brian’s offensive line points were via Wisconsin players. Seth got at least one point at all five positions. A bunch of guys not picked earned two points from dual honorable mention awards, but only Travis Jackson (MSU), who was a consensus 2^{nd} team All B1G, could be considered a miss.
Defensive line

Ace 
BISB 
Brian 
Seth 
2020 
DL 
12 
11 
7 
13 
13 
BISB and Brian were hurt by offthefield issues with Spence (OSU) and Clark (UM). Hamilton (RU) and Washington (OSU) didn’t even make the coaches’ team. Benntett (OSU) was the secondbest interior linemen but was squeezed out of the 1^{st} team by three ends. To my mind, Zettel (PSU), who was undrafted, is the second biggest surprise of the year, deservedly earning firstteam honors. Ott (IA) was also undrafted and made the media’s secondteam.
Linebacker

Ace 
BISB 
Brian 
Seth 
2020 
LB 
5 
13 
0 
5 
10 
Brian took the collar on linebackers with two injuries and Bolden & Ross being only ok. Seth reasonably had higher hopes for Jones (MSU, 3pts) and Longa (RU, 0 pts), while BISB picked the two best backers in Hull (PSU) and Ryan (UM). I thought Phil Steele’s 2^{nd}team preseason pick of Landisch (UW), who wasn’t drafted, was nuts at the time. He earned 5 points. Biegel (UW) also made the coaches 2^{nd} team.
Defensive backs

Ace 
BISB 
Brian 
Seth 
2020 
DB 
9 
8 
17 
11 
19 
Brian’s DB picks looked good ex ante and proved to be ex post. In retrospect the rest of the gang took too many Michigan and Michigan State backs and too few Gophers. BodyCalhoun and Wilson each earned five points. Hilary (WIS), Williams (PUR) and Van Hoose (NW) all made one 2^{nd} team and honorable mention otherwise.
Kickers

Ace 
BISB 
Brian 
Seth 
2020 
P/K 
8 
4 
0 
0 
10 
Ace’s pick of Brad Craddock, who by virtue of getting one 1^{st} team All American selection, pushed him past Brian. BISB picked a couple guys earning dual honorable mention in Johnston (OSU) and Sadler (MSU). Mortell (1^{st} team) and Ficken (PSU) (2^{nd} team) would have been a good punter/kicker tandem that weren’t drafted. DuVernois (Ill) was a consensus 2^{nd} team punter who also went unpicked.
Updated analysis of football injuries in the Big Ten
As can be seen in the table, Michigan is second to Purdue in missed starts, updated through Saturday’s game. The third column takes starts lost (SL) divided by games played (GP). I’ve changed the methodology a bit due to popular demand and explain the changes below.

GP 
SL 
%GM 
Non QB 
QB 
Steele 
Adj 
Purdue 
10 
65 
6.5 
39 
0 
26 

Michigan 
10 
64 
6.4 
46 
0 
18 

Maryland 
9 
46 
5.1 
46 
0 
18 
18 
Minnesota 
9 
43 
4.8 
27 
1 
15 

Northwestern 
9 
43 
4.8 
16 
0 
27 

OhioState 
9 
37 
4.1 
14 
9 
12 
2 
Rutgers 
9 
37 
4.1 
28 
0 
9 

Indiana 
9 
35 
3.9 
13 
2 
20 

Nebraska 
9 
33 
3.7 
24 
0 
9 

PennState 
9 
33 
3.7 
13 
0 
20 
0 
Wisconsin 
9 
30 
3.3 
22 
4 
4 

Illinois 
9 
25 
2.8 
14 
4 
7 

MSU 
9 
23 
2.6 
13 
0 
10 

Iowa 
9 
20 
2.2 
19 
1 
0 

I calculated the total of starts lost (SL) as the sum of four columns.
 Non QB counts starts lost as previously. For every player with at least one start, I sum up across players the difference between the number of games the team played versus the number of games the player appeared in. This is calculated only for non quarterbacks and is nonedited, so it should be replicable.
 QB counts the starts lost by quarterbacks. The second table presents the relevant data. Lunt, Sudfeld, Rudock, Leidner and Stave had verifiable injuries. I’m counting Stave’s yips as missed starts for the first four games. I assume that games not played by the others were coaches’ decisions. Yes, Morris had an injury, but I don’t think he not playing in the games after Minnesota truly represents lost games by starters. Ohio State’s 9 lost starts are Braxton Miller.
 The third column, Steele, is a response to previous comments. This lists all the games “lost” by players that were thought to be starters preseason but haven’t started a game yet. (Preseason starters who started a game but also missed some games are already counted in the Non QB column.) I used Phil Steele’s preseason magazine as I wanted a consistent, easily available source. The third table presents all the Steele starters who didn’t start a single game so far, how many games they played, and hence the games missed. All of this is unedited and therefore should be easily replicable.

The fourth column represents some sensible adjustments I made to three teams.
 Steele lists two freshmen as starters on the offensive line for Maryland, but Gray and Prince appear to be headed for redshirts. So, it doesn’t seem like their missed games should be counted.
 The NCAA data count Noah Spence as having started a game and played in two. Hence, the rest of the games he missed are counted in the Non QB column. But, a poster said that he didn’t play in those games that the NCAA thinks he appeared in. So I added 2. If I’m wrong, let me know.
 I made offsetting adjustments to Penn State. Thompkins is listed as a starter by Steele, but he’s headed for a redshirt. So, I subtracted the 9 missed games. On the other hand Diefenbach wasn’t listed as a starter by Steele because he knew he was out before publication. But, it seems like those should be counted anyway, so I added back 9.
Starting Big Ten Quarterbacks 



GP 
GS 
GM 
Illinois 
O'Toole 
9 
4 
0 
Illinois 
Lunt 
5 
5 
4 
Indiana 
Sudfeld 
7 
6 
2 
Indiana 
Diamont 
3 
3 
0 
Iowa 
Rudock 
8 
8 
1 
Iowa 
Beathard 
6 
1 
0 
Maryland 
Brown 
9 
9 
0 
Michigan 
Gardner 
10 
9 
0 
Michigan 
Morris 
5 
1 
0 
MSU 
Cook 
9 
9 
0 
Minnesota 
Leidner 
8 
8 
1 
Minnesota 
Streveler 
5 
1 
0 
Nebraska 
Armstrong 
9 
9 
0 
Northwestern 
Siemian 
9 
9 
0 
OhioState 
Barrett 
9 
9 
0 
OhioState 
Miller 
0 
0 
9 
PennState 
Hackenberg 
9 
9 
0 
Purdue 
Appleby 
8 
5 
0 
Purdue 
Etling 
5 
5 
0 
Rutgers 
Nova 
9 
9 
0 
Wisconsin 
McEvoy 
9 
5 
0 
Wisconsin 
Stave 
5 
4 
4 
Phil Steele Starters in the Big Ten with No Starts 



GP 
GM 
Class 
Illinois 
Day 
9 
0 
So 
Illinois 
James 
2 
7 
Fr 
Illinois 
Schmidt 
9 
0 
So 
Illinois 
Smoot 
9 
0 
So 
Indiana 
Eckert 
5 
4 
Sr 
Indiana 
Friend 
0 
9 
So 
Indiana 
Green 
9 
0 
Jr 
Indiana 
Taylor 
2 
7 
Sr 
Iowa 
Powell 
9 
0 
Sr 
Maryland 
Gray 
0 
9 
Fr 
Maryland 
Prince 
0 
9 
Fr 
Michigan 
Bosch 
1 
9 
So 
Michigan 
Morgan 
1 
9 
Sr 
MichiganState 
Arnett 
3 
6 
Sr 
MichiganState 
Knox 
5 
4 
Jr 
Minnesota 
Bak 
0 
9 
Sr 
Minnesota 
Bobek 
3 
6 
Jr 
Nebraska 
Reeves 
9 
0 
Jr 
Nebraska 
Rose, J 
9 
0 
Jr 
Nebraska 
Rose, M 
0 
9 
So 
Northwestern 
Jones, C 
0 
9 
Sr 
Northwestern 
Mark 
0 
9 
Sr 
Northwestern 
McEvilly 
0 
9 
Sr 
OhioState 
Lindsay 
0 
9 
Sr 
OhioState 
Reeves 
8 
1 
Jr 
OhioState 
Underwood 
7 
2 
Jr 
PennState 
Anderson 
0 
9 
So 
PennState 
Thompkins 
0 
9 
Fr 
PennState 
Zwinak 
7 
2 
Sr 
Purdue 
Clements 
7 
3 
Jr 
Purdue 
Gregory 
8 
2 
So 
Purdue 
Monteroso 
2 
8 
So 
Purdue 
Newton 
0 
10 
So 
Purdue 
Robinson, G 
9 
1 
Fr 
Purdue 
Yancy 
8 
2 
Fr 
Rutgers 
Lambert 
9 
0 
So 
Rutgers 
Peele 
0 
9 
So 
Wisconsin 
Hudson 
8 
1 
Fr 
Wisconsin 
Keefer 
8 
1 
Jr 
Wisconsin 
Maly 
8 
1 
Jr 
Wisconsin 
Wheelwright 
8 
1 
So 
Comparing injuries across the Big Ten
Alum 65 asked what did we do to the injury gods. I was thinking the same thing as it seemed like a lot, but then I asked myself how it compares to other schools. Some Internet searches convinced me that reporting was too inconsistent across schools to have a good, comparable measure. Plus, it was really time consuming.
The table below presents an imperfect measure of starts lost. Michigan appears second to Maryland in terms of injuries in the Big 10.
“Starts” lost* 


Games Played 
Starts lost 
as % of gm. 
Maryland 
6 
21 
3.5 
Michigan 
7 
22 
3.1 
Purdue 
7 
20 
2.9 
Minnesota 
6 
17 
2.8 
Rutgers 
6 
15 
2.5 
Wisconsin 
6 
12 
2.0 
Nebraska 
6 
12 
2.0 
Illinois 
7 
11 
1.6 
Iowa 
6 
8 
1.3 
Northwestern 
6 
7 
1.2 
Indiana 
6 
6 
1.0 
MichiganState 
6 
5 
0.8 
PennState 
6 
4 
0.7 
OhioState 
5 
2 
0.4 
* excludes secondstring quarterbacks 
For each player on the team that records at least one start in the NCAA database, the table sums up the number of games played according to the same database less the number of games the team played. So, for instance, Michigan has played seven games. Jehu Chesson started at least one game (four actually, which doesn’t matter) and has played in six. So, he contributes 1 to Michigan’s total. Kyle Kalis started only three games, but because he has played all seven he doesn’t contribute to the measure. The theory is that if a player has started at least one game, he probably plays in all the games even if he doesn’t start unless he is injured or suspended. You can’t just take everyone on the roster and calculate how many games weren’t played because you’ll end up just with an estimate of how coach’s play their benches and how many blowouts there were.
The one adjustment I made to the count was that I treated quarterbacks differently. If a starter is replaced, he may not be hurt but not play in the game. Moreover, if you end up starting a secondstring QB in one game, you’ll end up treating him as hurt in all the other games that he may not have played. Shane Morris is such an example, as he started against Minnesota, but did not play against Notre Dame when he wasn’t hurt. So, he otherwise would have been counted as injured for Notre Dame had I not just excluded secondstring quarterbacks. I did go back and add in any games not played by obvious 1^{st} string quarterbacks. Rudock, Leidner, and Lunt missed games due to injury, while it doesn’t appear that Etling, who missed two games, was hurt. (The theory is that reporting on starting quarterbacks is good enough across teams.) I didn’t add back in an injury for Morris for the Penn State game, which he probably should be counted, just like I didn’t do any comparable searches for other secondstring quarterbacks who may be injured.
It’s useful to acknowledge all of the other imperfections in this measure.
 It includes suspensions, so Glasgow adds one to Michigan.
 It doesn’t include obvious starters lost for the entire year before the season started, such as Braxton Miller. Note that Noah Spence also doesn’t show up, as although he played in one game, he didn’t start in that game.
 It doesn’t include injuries to important people on special teams or important backups that in a perfect world would be counted.
 No adjustments are made for how good the player was that was lost. Nor does it weigh injuries if for instance they occur in the same area, which may compound the damage.
 No allowances are made for guys who play in some games but are obviously not 100% (Jake Butt?).
 Teams that have instability in their starting lineups will have a bigger pool to have a chance to show up as injured.
The advantage is that it doesn’t depend on a detailed knowledge of the teams, which might then be biased as I at least know more about some teams more than others.
Uncommitted Players by Region
With Mike Weber committed, there are very few top 300 players left in the Midwest. According to the 247 Composite there are only 4 left out of 25 prospects. The top guy in the Midwest, Terry Beckner, is uncommitted, with the next highest guy being Asmar Bilal at #198. This is way different than the West where over half the guys in the top 300 still uncommitted. The table reports the number of uncommitted players by five different regions. There’s one Canadian prospect that’s otherwise kept separate from the analysis.
One hypothesis as to why this is so is that relative to the number of top guys, there are more major schools in the Midwest than elsewhere. The correlation between the uncommitted rate and the number of available players per major school is 0.67. Of course, this is a perfect example of the wrong way to do statistics: look at the data and then design a hypothesis around the fact. Indeed, drop the Midwest and the correlation goes away. On the other hand, if you saw a scatter plot of the data you’d see that it’s actually the South that looks like the outlier; the Midwest and the other three regions draw a pretty nice regression line, while the uncommitted rate in the South looks low relative to the number of top players per major school. .

Uncom . 
247 Com. 300 
Uncom. Rate 
Major Schools 
247 Com 300 per Major 
Midwest 
4 
25 
0.16 
12 
2.1 
East 
13 
35 
0.37 
9 
3.9 
Midlands 
17 
54 
0.31 
12 
4.5 
South 
47 
131 
0.36 
21 
6.2 
West 
28 
54 
0.52 
11 
4.9 
Canada 
1 
1 
1.00 
0 

Total 
106 
275 
0.39 
53 
5.2 
Midwest: Big Ten footprint when the Big 10 had ten schools.
East: VA/WV/PA and points north and east.
Midlands: ND south to TX. NM north to MT, ID + MO.
South: SEC footprint before the last expansion + NC.
West: Pac 12 footprint + AK & HI.
Correcting O/U win totals for moneylines
The fact that over/under win totals come with moneylines provide no real complication for the betting public. Convert the moneylines to a return, ask whether the probability is better than the return and bet accordingly. But, what of rest of the public who want to read the over/under’s as Vegas’s expected wins? How can you convert the combination of the moneylines and O/U’s to an expected win value? To be concrete Michigan is at 7.5 (140 over/+100 under) according to the odds just released. Penn State is at 8.5 but (+100/140). How far apart are they? My results suggest Michigan’s expected number of wins is just a little higher at 7.6, while Penn State’s is a ½ win lower than its O/U at 8.0. In 2013, there were a few teams whose moneylines were so extreme that they implied expected wins a full win less than the O/U.
Let p denote the probability of winning more games than the O/U. Then the implied expected number of wins is p E[winswins>O/U] + (1p) E[winswins<O/U] where the expected values can be read as the average number of wins when a team comes above or below the over/under. The probability p is calculated from the moneylines and Table 1 contains estimates of these averages, which depend on the level of the over/under.
The way to interpret the moneylines depends on whether they’re positive or negative. If positive, the line gives you the extra amount you win if you bet $100. If negative, it gives you the amount you have to bet to win an extra $100. Let p(o) and p(u) denote the probabilities read directly from the moneylines, and denote by m as the absolute value of the number in the moneyline. If the moneyline is positive, then the probability is 100/(m+100); if negative it’s m/(100+m). In Michigan’s case, p(o) = 140/(100+140) = 0.583 and p(u) = 100/(100+100) = 0.500. If you think the probability that Michigan wins more than 71/2 games is greater than 58.3%, bet the over; if you think the probability is less than 50%, bet the under.
Note, however that p(o)+p(u) is greater than one. That’s Vegas’s juice to make money. We want a fair probability for our purposes, so we convert them by setting p = p(o)/(p(o)+p(u)). In Michigan’s case, p = 0.583/(0.583+0.500) = 0.538.
Table 1 contains the applicable expected values for various over/under's calculated under two different methodologies. The first method takes actual over/under’s from 20102013. The second method takes Chris Stassen’s data on preseason magazine ratings and converts them as a decent exante estimate of an O/U. Then for each I calculate the average wins when the teams come in above and when they come in below the totals. For my calculations below I then take an average of the two estimates. In Michigan’s case, the applicable expected values are E[winswins>O/U] = 9.20, which is the midpoint of the range and E[winswins<O/U] = 5.75, likewise. Weighting by 0.538 gives 7.6 wins.
Table 1 Conditional Win Averages 

O/U 
E[winswins>O/U] 
E[winswins<O/U] 
3.5 
5.20  4.80 
2.00  2.17 
4.5 
6.69  7.11 
2.86  2.67 
5.5 
7.18  8.00 
3.26  3.52 
6.5 
8.16  8.00 
4.87  4.40 
7.5 
9.33  9.06 
5.86  5.64 
8.5 
9.86 – 10.24 
6.41  6.05 
9.5 
10.73  10.45 
7.53  7.11 
10.5 
11.00  11.33 
9.50  9.50 
11.5 
12.00  12.00 
n.a.  11.00 
To be sure my estimated expected wins in table 1 could be improved upon, and the method as a whole could be augmented by arguing that the expected values should vary with the moneyline.
Table 2 gives expected wins for all Big Ten teams and Notre Dame for those over/under’s provided by Kegs ‘N Eggs. Note that Notre Dame’s expected wins is 0.7 below the O/U.
Table 2 Big Ten Expected Wins 


O/U 
Moneyline 
E[wins] 
Indiana 
5.5 
+145 over / 185 under 
5.0 
Maryland 
7.5 
+100 over / 140 under 
7.3 
Michigan 
7.5 
140 over / +100 under 
7.6 
MichiganState 
9.5 
155 over / +115 under 
9.2 
Nebraska 
7.5 
130 over / 110 under 
7.5 
Notre Dame 
9.5 
+110 over / 150 under 
8.8 
OhioState 
10.5 
140 over / +100 under 
10.4 
PennState 
8.5 
+100 over / 140 under 
8.0 
Rutgers 
4.5 
110 over / 130 under 
4.8 
Wisconsin 
9.5 
135 over / 105 under 
9.1 
.
Should Michigan fans have rooted for OSU over Northwestern
Summary
A simulation suggests that Ohio State beating Northwestern increases Michigan’s chances at that time of winning the division by 5 percentage points and the Big Ten championship by 11/2 pp. Such a win also improves on average Michigan’s bowl placement when it doesn’t win the division. On the other hand, what’s true for Michigan is also true for Michigan State. And perhaps more importantly, a win over Northwestern greatly improves Ohio State’s chances of a great season.
Details
So, events have overtaken the question, but the argument that I should have rooted for OSU over Northwestern, which played out in some forums in the week before the game, remains stuck in my mind. Before getting to the relevant analysis, I’m compelled to make two points.
 Whom I root for is of no practical consequence to anyone else, especially how Michigan’s season actually plays out.^{1} I can understand an argument that I should give to charity or be nice to my neighbor, but why I “should” do something that doesn’t matter strikes me as a strange concept.
 Being happy that OSU beats Northwestern makes sense because it increases Michigan’s chances of winning the division. Taking pleasure in an OSU loss makes sense because, obviously. The choice depends on how I weigh the tradeoffs. Determining whom to root for solely on how it affects Michigan’s chances of winning the division at the exclusion of all other considerations may work for some people, but there’s no basis to argue that I “should” have those same preferences.^{2}
Still, there’s an unresolved issue that can be analyzed. How much does the outcome of the OSUNorthwestern game affect Michigan’s chances of winning the division? For that matter, how does it affect other considerations, like MSU’s chances, OSU’s chances of going undefeated, etc.?^{3} Armed with those answers, everyone can draw their own conclusions over whom they should have liked to win.
To answer this question, I picked a probability structure for the Big Ten games as scheduled, used a random number generator to establish outcomes, programmed the Big Ten’s tiebreaking procedures, and repeated 5,000 times. To calculate the effect of the OSUNW game, I let the rest of the season stand and calculated the season outcomes when NW wins and when OSU wins. I also let stand the actual outcomes of the MinnesotaIowa and WisconsinOhio State games that occurred the week before.
The lefthand side of the Table 1 details the results for Michigan. When Northwestern wins the game, Michigan wins the division about 25% of the time; when Ohio State wins, Michigan wins the division about 30%, for a 5 percentagepoint differential. (Keep in mind that with four teams in the division that at the time were thought to have reasonable chances of winning the division, any one team’s odds aren’t going to be large.)The differential for the Big Ten championship, however, is much smaller at 11/2 points. That’s because an OSU win increases its chances of winning the Leaders Division, and Michigan’s chances of beating OSU in the championship games are assumed to be smaller than those of beating anyone else out of that division.
Table 1 Probabilities of U of M / MSU Season Outcome 


Michigan 

MichiganState 


NW W 
OSU W 

NW W 
OSU W 
B1G winner 
11.6 
13.1 

4.2 
4.8 
Division winner 
25.3 
30.3 

12.9 
16.7 
Helped 
0.6 
5.5 

0.2 
4.0 
Unaffected 
24.7 

12.7 







B1G order, nondiv winner 





Helped 
7.0 
11.2 

2.9 
11.5 
Unaffected 
50.7 

63.9 
It’s interesting to observe that there are a small number of times when a Northwestern win helps Michigan win the Legends. That can happen in the following case: MSU beats Michigan but loses to Northwestern and say Iowa. Michigan beats Northwestern and loses to a team in the Leaders Division. And Northwestern loses to one other team other than OSU and Michigan. If Northwestern beats OSU, then there’s a threeway tie that Michigan wins because each of the winners are 11 against the others, but either Michigan has the best division record or at least MSU is knocked out for having the worst division record, and then Michigan wins headtohead against Northwestern. On the other hand, if Northwestern loses to OSU, then it’s just a tie between Michigan and Sparty, which Sparty wins by the headtohead match up. Even so, the more common outcome is that a Northwestern win hurts Michigan’s title hopes, rather than helping.
One may also be interested in knowing how a Northwestern win affects Michigan’s bowl placement when it doesn’t win the division. (I use a simple rule that considers nonconference wins and a tiebreaking procedure that looks at headtohead records to determine the order, but then gives Michigan the benefit of the doubt otherwise.) The outcome of the NWOSU game is of no consequence for Michigan’s bowl placement the large majority of the time. Otherwise, a NW win helps Michigan sometimes, but hurts it about 4 pp more often.
What’s true for Michigan is also true for Michigan State, which may be relevant for some people. An OSU win increases MSU’s chances of winning the division by 33/4 pp and winning the Big Ten by 0.6 pp. As with Michigan it also helps its bowl placement on average.
Obviously, a win against Northwestern also makes it more likely that OSU has an outstanding season, which could also be an important consideration. An OSU win over Northwestern increases the probability OSU wins it division by almost 19 pp and the probability it wins the Big Ten by 13 pp (Table 2). It increases its chances of going to a BCS game by 263/4 pp (Big Ten winner or 102 record or better.) And finally, what some may consider a nightmare scenario, Ohio’s chances of going undefeated go from zero to 15% with a win against Northwestern.
Table 2 Probabilities of OSU Season Outcome 


NW W 
OSU W 
Division winner 
70.3 
89.0 
B1G winner 
42.1 
55.2 
130 
0.0 
15.1 
121 
14.7 
26.8 
112 
21.4 
12.2 
B1G loser 
28.2 
33.8 
121 
0.0 
9.6 
112 
10.0 
16.5 
Division nonwinner 
29.7 
11.0 
111 
0.0 
0.1 
102 
7.8 
5.2 
BCS 
59.9 
86.6 
Methodology
For each scheduled Big Ten game (other than OSUNorthwestern and the two games the week before) I model the probability of the home team winning as depending on the difference in fixed strength parameters plus a home premium. A more sophisticated model could allow for match up issues that mean win probabilities aren’t transitive.
Table 3 describes these parameters, the probability Michigan beats them on a neutral field and the average Big Ten wins in the simulations. They aren’t meant as an accurate gauge of the relative strengths of the teams as much as an expression of what I take to be the relative strengths of the teams as perceived by the fan base at the start of the Big Ten season. Home teams won 56.9% of their Big Ten games in the 20032012 period, so the homefield premium equals 6.9 pp. Home team i beats visitor j in the simulation if the draw of a random variable (distributed uniformly between [0,1]) is less than or equal to ½ + ½ (P_{i } P_{j}) + h. The outcomes of all the games are assumed to be mutually independent. The results are little different quantitatively, and are qualitatively the same, if I shock the strength parameters for each simulation with a uniformly random variable between [0.10, +0.10].
Table 3 Assumed Relative Strength 


Strength 
Probab. v. Mich. 
Avg. Wins 
Iowa 
0.15 
0.18 
2.7 
Michigan 
0.80 

5.1 
MichiganState 
0.55 
0.38 
4.6 
Minnesota 
0.10 
0.15 
1.5 
Nebraska 
0.70 
0.45 
5.2 
Northwestern 
0.80 
0.50 
5.1 
Illinois 
0.10 
0.15 
2.0 
Indiana 
0.40 
0.30 
3.6 
OhioState 
0.95 
0.58 
6.3 
PennState 
0.55 
0.38 
4.2 
Purdue 
0.05 
0.13 
2.1 
Wisconsin 
0.85 
0.53 
5.5 
Probabilities versus Michigan are calculated on a neutral field. Average wins for Northwestern and OSU take the outcome of their game as 5050. 
Endnotes
1. My theology doesn’t leave room for God choosing the victor based on a prayer vote or my prayers alone, but I can see if others go in a different direction. I’m also excluding the metaphysics of Scott Adams of Dilbert fame who argues that you can effect changes in the universe over which you otherwise have no control by simply writing down an affirmation for 15 straight days. See the end of The Dilbert Future: Thriving on stupidity in the 21^{st} century.
2. Economists call a preference structure where one needs to be satiated in one dimension before appreciating other dimensions as lexicographic preferences. They’re not logically excluded, but they’re not how I roll. The spirit of this piece is that life has tradeoffs.
3. There are two things I’m not going to consider. First, I’ll ignore whatever effect the game has on BCS rankings as it’s probably small and related to the order in the Big Ten standings anyway. Second, another reason to root against a team is because you’d like an indication that the team is bad and is more likely to lose future games. An OSU loss will make you feel better about the Game as you update your win probability using Bayes’ Theorem. This thought experiment doesn’t lend itself easily to such considerations. It takes pairwise probabilities as “known” in advance.