2003: An Advanced Stats Year in Review
Images archived from MGoBlue.com
This is the first in a series of posts going back through the years to pull out some of the highlights from an advanced stats perspective of years prior. Glossary at bottom for further detail on specific metrics. @The_Mathlete
The Basics
Record: 10-3
Big Ten: 7-1, Big Ten Champs
Bowl: Rose Bowl loss to USC
Final AP Rank: 6th
Advanced Team Stats
Talent rank: 13th in nation (Texas), 2nd in Big Ten (Ohio)
Offense: +7 EV+ (10th-Minnesota, 2nd)
Defense: +6 EV+ (17th-LSU, 4th-Iowa)
Special Teams: +1 EV+ (33rd-MSU, 5th)
Advanced Player Stats
John Navarre, 1st Team Big Ten, +2 EV+ (27th-Roethlisberger, 3rd-Abdul-Khaliq), 0.13 WPA/G (39th-Frye, 4th-Sorgi)
Top Games
Ohio: 278 yards, +8 EV+ and 0.23 WPA
Minnesota: 369 yards, +5 EV+ and 0.37 WPA (legendary receiving TD not included)
Northwestern: 284 yards, +7 EV+ and 0.18 WPA
Chris Perry, all everything, +2 EV+ (9th-Barber, 3rd), 0.09 WPA/G (6th-Barber, 3rd)
Top Games
Central Michigan: 17 carries for 232 yards, +8 EV+ and 0.25 WPA
Michigan St: 52 carries for 220 yards, +8 EV+ and 0.30 WPA
Houston: 18 carries for 184 yards, +7 EV+ and 0.14 WPA
Braylon Edwards, 1st team Big Ten, +4 EV (51st-Fitzgerald, 2nd-Evans), 0.07 WPA/G (97th-Hackett, 8th-Evans)
Top Games
Ohio: 7 targets, 132 yards, +12 EV and 0.21 WPA
Michigan St: 7 targets, 103 yards, +11 EV and 0.23 WPA
Northwestern: 6 targets, 110 yards, +7 EV and 0.10 WPA
WPA Charts
Oregon
Michigan takes their first loss of the season in a wild one in Eugene. Three special teams touchdowns, Michigan recovering the onside kick with a chance for a dramatic comeback from an 18 point second half deficit that wasn’t to be.
Iowa
Michigan takes control early with a 14 point first quarter lead, only to see Iowa reverse course to go up 30-20 midway through the fourth. Braylon Edwards goes for a 41 yard touchdown and the Wolverine D forces a three and out. But Michigan’s final drive stalls on Iowa’s half of the field.
Minnesota
Michigan’s third game with big swings finally goes their way. Michigan trailed 28-7 at the end of the third quarter with their only touchdown coming on the famed Breaston to Navarre connection. Four fourth quarter touchdowns would tie the game and Garrett Rivas hit from 33 yards out with less than a minute to go to give Michigan it’s final margin.
Ohio
Tressel’s only loss to Michigan in ten tries saw Michigan in control until the fourth quarter. A John Navarre interception gave the Buckeyes the ball back down seven with 13 minutes to play, but a three and out, a scare on a fumbled punt return and an 87 yard touchdown drive would push the lead back to 14 and Ohio wouldn’t threaten again.
Season Summary
In terms of contribution to success, the 2003 Michigan team was highly balanced. Chris Perry would be the player that took home the hardware, those of you have followed my articles are familiar with the lack of love EV+ and WPA have for running backs. Perry certainly had an outstanding season but benefited from a lot of carries and the famed position as lead back at Michigan. In all, Perry and the ground game, a solid Navarre and a pretty good defense would give Michigan its last outright Big Ten title and only victory over a Jim Tressell-led Buckeye team.
For being BIG TEN FOOTBALL, the season featured several wild comeback games. Michigan pulled out a win over Minnesota but dropped games to a solid Iowa team and an unranked Oregon team. The Rose Bowl saw Michigan take on the first USC juggernaut that went on to win the AP crown despite missing out on the BSC title game.
Heisman Re-Vote
2003 was a tough vote for the Heisman. Jason White, Larry Fitzgerald, Eli Manning and Chris Perry were the finalists. My four finalists would have been Ben Roethlisberger, Larry Fitzgerald, Philip Rivers and Matt Leinart (sorry Chris).
Ben Roethlisberger +11 EV+ and Larry Fitzgerald +9 have the best claims to the most valuable season. Fitzgerald’s EV+ was the best wideout rating in the 9 years of data I have accumulated.
If you don’t like mid-majors or wideouts, Philip Rivers was the best BCS QB or RB at +8 for NC State. If you think the winner needs to be a “winner” and have a team in title contention, then Leinart’s your man. At +7 he is well ahead of the other title contenders, eventual winner Jason White from OU (+4) and Matt Mauck from LSU (+0).
My vote: Larry Fitzgerald, a once in a decade receiving season and no dominant BCS QBs makes this my clear choice
My Post-Season Top 25
Rank | Team | EV+ | AP Rank |
1 | USC | 19 | 1 |
2 | LSU | 16 | 2 |
3 | Miami (OH) | 14 | 10 |
4 | Oklahoma | 14 | 3 |
5 | Michigan | 14 | 6 |
6 | Kansas St | 14 | 14 |
7 | Georgia | 13 | 7 |
8 | Texas | 12 | 12 |
9 | Oregon St | 10 | NR |
10 | Miami (FL) | 10 | 5 |
11 | Minnesota | 9 | 20 |
12 | Washington St | 9 | 9 |
13 | Maryland | 9 | 17 |
14 | Tennessee | 9 | 15 |
15 | Florida St | 9 | 11 |
16 | Utah | 9 | 21 |
17 | Arkansas | 8 | NR |
18 | Auburn | 8 | NR |
19 | Iowa | 7 | 8 |
20 | Nebraska | 7 | 19 |
21 | Ohio St | 7 | 4 |
22 | NC St | 6 | NR |
23 | Clemson | 6 | 22 |
24 | New Mexico | 6 | NR |
25 | Texas Tech | 6 | NR |
26 | Boise St | 6 | 16 |
27 | Bowling Green | 6 | 23 |
31 | Purdue | 5 | 18 |
36 | TCU | 3 | 25 |
37 | Mississippi | 3 | 13 |
38 | Florida | 3 | 24 |
The top five compared pretty closely between the two. The notable exceptions were the two teams from Ohio. Roethlisberger’s Red Hawks check at #3 in my final rankings where that other team from Ohio was highly overrated based on a number of close wins. Michigan gets a slight bump up to #5.
Glossary
EV+: Opponent adjusted expected value. How many points a player/team/unit was worth relative to average performance against the same opponent. Excludes all garbage time plays and fumbles (considered random).
WPA: Win Percent Added. How much each player/unit contributed to winning games. Each team starts a game with a 50% chance. For the first three quarters in a competitive game this will track directly alongside EV. In games with larger margins almost all plays have zero value. Plays in the fourth quarter/OT of tight games will have a higher leverage based on the situation. This is largely were WPA deviates from EV+. WPA for offensive players will be higher if the teams defense is bad and vice versa. If your defense is giving you good field position and stopping the opponent, there is less opportunity for the offense to “win” the game. If your defense is terrible your WPA numbers will be higher because there is constant pressure for the offense to succeed. Includes all plays.
Ranks: The ()’d numbers after ratings are the national rank first followed by the conference rank. If there is a team/name after the rank, that is the leader of the category.
Talent Rank: Total team talent estimate based on recruiting rankings of all players on that year’s roster based on all available major scouting services. Only counts plays currently on the roster and weights players with more years on the team higher than new recruits.
We had less talent than OSU? I don't know if I buy that one.
According to the ESPN NFL Draft Tracker:
2004 Draft OSU 14 players drafted; 3 in first round
UM 4 players drafted; 1 in first round
2005 Draft OSU 3 players drafted 0 in first round
UM 3 players drafted; 2 in first round
2006 Draft OSU 9 players drafted; 5 in first round
UM 3 players drafted; 0 in first round
2007 Draft OSU 8 players drafted; 2 in first round
UM 7 players drafted; 1 in first round
Total 04-07 drafts OSU 34 players drafted; 10 in first round
UM 17 players drafted; 4 in first round
I can accept that there might be measures of talent other than being drafted by the NFL, but the professionals seemed to have the opinion that the 2003 OSU team, from its seniors down to its freshmen, had more draft-worthy talent than the 2003 UM team.
For 2003 its a bit rough and you might be right. I only have recruiting rankings back to 2002 so most of the upperclassmen weren't counted in that total.
It's funny how those WPA charts also could represent our collective happiness over the course of those games.
It's sort of an MGoZeitgeist. There were a couple of times last year when reviewing these charts that I felt they were either more or less optimistic than I was at the time, but for the most part they matched up well.
It's cool to me that we, as fans, have sort of an internal WPA calculator going on at all times during games.
March 27th, 2012 at 12:09 PM ^
I don't remember these charts very well. Is there a way to incorporate the spread to get a more realistic starting point (more realistic than 0.5 for every game)?
March 27th, 2012 at 12:24 PM ^
Yes, the Mathlete has provided spread adjusted WPA charts in the past: http://mgoblog.com/diaries/mid-week-metrics-honors-fallen-coach
I took a Math class at UofM a couple of years back (Math 462) on the topic of mathmatical modeling. For my final project of the semester I took a stab at making my own ranking system along the same vein, and looked back at some historical seasons to see how my model would have set the national championship game compared to how it actually played out.
Seeing this post reminded of my old project, so I pulled it up and looked at what my model gave for the 2003 season, it's interesting in the fact that it is in no way close to your rankings at all.
March 27th, 2012 at 10:02 PM ^
No mention of John Navarre is complete without THIS LINK.
Comments