Fiu, Cirminiello, Mitchell on TV - Campus Insiders | Buy College Football Tickets

Compu-Picks Analysis: Injury Starts Lost

Mr Pac Ten
Posted Jul 22, 2012


Offseason Compu-Picks Analysis Part 6: Starts lost to injury

Any of you who regularly read Phil Steele's annual preview magazine (or regularly visit his website and saw the 2009, 2010, or 2011 article) probably noticed that one key point of emphasis is injury starts lost. Specifically, Phil Steele suggests that teams who enjoy especially good injury fortune in one year tend to have a worse record the next, because injury luck tends to revert to normal (and the opposite for those who suffer an especially bad slate of injuries in a year).

As part of creating a preseason compu-picks model, I looked at a number of potentially relevant predictive variables. Among these variables was injury starts lost. Was Phil Steele right? Do injuries really have a legitimate predictive value for the next year?

The answer appears to be yes (I say appears to be due to the limited data set I have available; with a larger data set I could potentially come to a stronger conclusion). Injury luck does appear to has meaningful predictive value. The reason for this is twofold: first, that injury improvement correlates to record improvement; and second, that injuries from one year appear to have a very small correlation with injuries for the next year. Let's look at those two points in more detail:

1) Injury improvement correlates to record improvement.
As a test of this, I took at look at the nine teams who improved their injury luck by at least 30 starts either between 2009 and 2010 or 2010 and 2011, and looked at how their W/L records changed in the same time frame. They were as follows:

2009 - 2010:

Washington St
2009: 1-11 (67)
2010: 2-10 (20)

Oklahoma
2009: 8-5 (42)
2010: 12-2 (4)

Oklahoma St
2009: 9-4 (40)
2010: 11-2 (2)

Syracuse
2009: 4-8 (41)
2010: 8-5 (8)

Arkansas St
2009: 5-7 (37)
2010: 4-8 (5)

2010 - 2011:

North Carolina
2010: 8-5 (89)
2011: 7-6 (26)

San Jose St
2010: 1-12 (78)
2011: 5-7 (19)

UL-Lafayette
2010: 3-9 (48)
2011: 9-4 (15)

Virginia
2010: 4-8 (34)
2011: 8-5 (4)

Seven of the nine improved their record, and two dropped, each by just one win. Of those that improved their record, four added four wins and one added six wins. In addition, Oklahoma St, a team who added two wins, was one of most stunning results of the 2010 season; everyone expected them to be 5th or 6th in the Big 12 South, and instead tied for the division title and won the Alamo Bowl. So six of the nine either improved their record by FOUR OR MORE wins, or defied a consensus of collapse and actually got substantially better. And one of the two who got worse, UNC, had most of those starts lost as season-long suspensions rather than standard injuries (which suggests that they may just have been an outlier).

And on the other side, here are the teams whose injury luck got worse by 30 or more starts:

2010 - 2011:

North Carolina:
2009: 8-5 (19)
2010: 8-5 (89)

UCLA:
2009: 7-6 (14)
2010: 4-8 (59)

SJ St
2009: 2-10 (35)
2010: 1-12 (78)

2010 - 2011:

Maryland:
2010: 9-4 (10)
2011: 2-10 (59)

Ohio St:
2010: 12-1 (7)
2011: 6-7 (55)

Toledo
2010: 8-5 (7)
2011: 9-4 (53)

Texas Tech
2010: 8-5 (3)
2011: 5-7 (38)

Oregon St
2010: 5-7 (21)
2011: 3-9 (54)

Akron
2010: 1-11 (5)
2011: 1-11 (37)

Army
2010: 7-6 (21)
2011: 3-9 (51)

Florida St
2010: 10-4 (16)
2011: 9-4 (46)

This is a similar story to the top improvers, only in reverse. Eight of the eleven worsened their records, two of the ten had the exact same record, and only one got better (and then just by a single win). Of those who worsened their records, three were by 4+ wins (Maryland by 7, Ohio St by 6, Army 4 by), and another two were by 3 wins (UCLA, Texas Tech).

Overall, there's a very clear pattern to these results. Teams that get a lot better in injury luck tend to get better in record, sometimes by a lot. And teams that get a lot worse in injury luck tend to get worse in record, sometimes by a lot. And this leads us to the second important point.

2) Injuries from one year appear to have a very small correlation with injuries for the next year.

As a test of this, I decided to run a correlation between one year's turnover margin and the next year's. Since I only have full data for 2009 - 2011 seasons, I only have 240 data points; enough to establish something of a conclusion, but not enough to really get into the kind of multi-year analysis I would like in order to be truly confident in the apparent conclusion.

So running the correlation between 2009's injury starts lost and 2010's, I found that the correlation was a mere 3.6%. That's essentially nothing. When I ran a correlation between 2010 and 2011, it came up 11.1%, greater than the 3.6%, but still suggesting that the correlation is very light. Combining the two, I got an 8.5% correlation. Now, it's certainly true that we're just looking at 240 data points, so it's premature to come up with too strong of a conclusion, but a 8.5% correlation is still an 8.5% correlation. While it's certainly possible that there are other patterns at work (perhaps over multiple years, or due to some other predictable factors), the injury losses from one year strongly appear to have virtually no correlation, positive or negative, with the injury losses from the next.

And examining the data in more detail (here I just looked at 2009-2010 for convenience), this conclusion appears to be borne out. In 2010, the average number of injury starts lost was 19.50. The ten teams with four or fewer starts lost to injury in 2009 averaged 21.4 injury starts lost in 2010, with two teams (Purdue and NM St) above 30, and only one (Rutgers) below 10. And the nine teams with 35 or more starts lost to injury in 2009 averaged 21.3 injury starts lost in 2010, with four teams (Arkansas St, OK St, Syracuse, Oklahoma) below 10 and two (Tennessee and SJ St with 78) above 30.

Both sets ended up about average, which is what the 0 correlation hypothesis would expect. It is interesting to note that the teams who had the highest injury losses in 2009 were far more variable in terms of injury losses in 2010 than those with the lowest injury losses. It is unclear whether this is a meaningful result or simply random noise in a data set that contains substantial variation. This is an issue I may re-examine when I obtain more data in the future.

Among other things, this suggests that some of the programs who've been hardest hit by injuries (UNC, SJ St, Tennessee, BC, Wazzu, Colorado St, UCLA, Vandy, Navy, and South Carolina have averaged 30+ starts lost to injury for 2009-2011) may be better, perhaps much better, than they've appeared. And those who've enjoyed the most injury luck (Rutgers, Temple, Kansas, Clemson, Baylor, and Cal have averaged under 10 starts lost to injury for 2009-2011) may be in for rough sailing if and when their luck turns.

And for those of you who like to play along at home, here is the data I used (source: Phil Steele website).

Starts Lost to Injury - by team by year

Team Average 2011 2010 2009
North Carolina 44.7 26 89 19
San Jose State 44 19 78 35
Tennessee 38 36 34 44
Boston College 37.7 55 26 32
Washington State 37.3 25 20 67
Colorado State 37.3 52 22 38
UCLA 36.7 37 59 14
Vanderbilt 34.3 42 19 42
Navy 33 25 45 29
South Carolina 30.3 35 34 22
Maryland 29.7 59 10 20
Florida State 29.3 46 16 26
Rice 29 38 35 14
Brigham Young 28.3 24 43 18
Louisiana-Lafayette 28 15 48 21
Ohio 28 22 35 27
Air Force 27.3 36 25 21
Ohio State 27.3 55 7 20
Alabama-Birmingham 26.7 30 29 21
Fresno State 26.3 31 36 12
Miami (Ohio) 26.3 29 22 28
North Texas 26 25 37 16
Tulsa 26 31 26 21
Southern Methodist 25.7 21 38 18
Oregon State 25.7 54 21 2
Utah State 25.3 16 43 17
Ball State 25.3 25 28 23
Missouri 25.3 36 31 9
Memphis 25.3 44 15 17
Oklahoma 25 29 4 42
Army 25 51 21 3
Colorado 24.7 40 17 17
New Mexico 24.3 31 27 15
Hawaii 24 26 13 33
Middle Tennessee State 24 35 20 17
Texas-El Paso 23.3 14 39 17
Southern Mississippi 23.3 28 25 17
Arizona State 23 34 13 22
Oklahoma State 23 27 2 40
Toledo 23 53 7 9
New Mexico State 22.3 32 31 4
Miami (Florida) 22 42 13 11
Purdue 21.7 30 32 3
Arkansas 21.7 21 19 25
Akron 21.7 37 5 23
Virginia Tech 21.3 25 20 19
Notre Dame 21 20 30 13
Eastern Michigan 20.7 12 16 34
Buffalo 20.7 16 15 31
Idaho 20.7 33 25 4
Kent 20.3 17 26 18
Syracuse 20.3 12 8 41
Kentucky 20.3 17 10 34
Louisville 20.3 24 17 20
Marshall 20 22 22 16
Louisiana State 20 31 16 13
Boise State 20 37 15 8
East Carolina 19.7 35 20 4
North Carolina State 19.7 31 7 21
Minnesota 19.3 9 20 29
Wisconsin 19.3 18 28 12
Central Michigan 19.3 28 20 10
Western Kentucky 19 13 21 23
Nevada 19 19 25 13
Indiana 18.3 22 24 9
Texas Christian 18.3 22 21 12
San Diego State 18.3 20 16 19
Pittsburgh 18.3 30 18 7
Stanford 17.7 17 11 25
Washington 17.7 19 10 24
Michigan 17.3 11 22 19
Florida 17.3 29 12 11
Oregon 17.3 24 6 22
Houston 17 6 21 24
Alabama 17 9 13 29
Auburn 17 15 17 19
Nebraska 16.7 21 26 3
Arizona 16.7 23 5 22
Texas Tech 16.3 38 3 8
Penn State 16 15 22 11
South Florida 16 8 14 26
Georgia 16 22 12 14
Connecticut 16 20 9 19
Florida Atlantic 16 22 7 19
Mississippi 15.7 21 16 10
Wake Forest 15.7 22 11 14
Iowa 15.3 11 18 17
Troy 15.3 17 23 6
Bowling Green 15 11 23 11
Louisiana Tech 15 14 17 14
Kansas State 15 25 10 10
Virginia 14.7 4 34 6
Arkansas State 14.7 2 5 37
Utah 14.7 16 14 14
Georgia Tech 14.3 9 13 21
Wyoming 14.3 27 10 6
Iowa State 14 21 7 14
Florida International 13.7 4 9 28
Western Michigan 13.7 13 12 16
Central Florida 13.3 10 18 12
Illinois 13.3 11 10 19
Duke 13.3 21 15 4
Southern California 13 11 10 18
Nevada-Las Vegas 13 18 16 5
Texas A&M 13 27 4 8
Northwestern 12.3 15 7 15
Tulane 11.7 10 11 14
West Virginia 11.7 14 11 10
Michigan State 11.7 18 7 10
Cincinnati 11.3 15 10 9
Mississippi State 10.7 11 14 7
Louisiana-Monroe 10.7 13 12 7
Northern Illinois 10.3 7 8 16
Texas 10.3 15 8 8
California 9.67 3 9 17
Baylor 9.67 10 1 18
Clemson 9.33 12 11 5
Kansas 9 8 17 2
Temple 5.67 7 4 6
Rutgers 5 5 6 4

2012 Compu-Picks Blog

Questions, comments or suggestions? Email me at cfn_ms@hotmail.com

Follow cfn_ms on Twitter