Saturday, July 11, 2009

Pieces of Steele | by Jay

Phil Steele always appends his jam-packed preview guide with a few pages of feature articles concerning various football metrics. This year is no exception. I thought I'd highlight two of his metrics and see what they say about the Irish and their opponents this year. Both of these metrics try to predict improvement (or regression) based on some simple counting items.

"For Better or Worse: Close Wins, Close Losses." This metric posits that a team that suffers a lot of close losses (defined by Steele as a touchdown or less) will probably improve the following year; likewise, a team that benefits from a lot of close wins stands to regress.

Steele runs the numbers for the past seven years and finds that teams that net more than 2 close losses improve their record 68% of the time the following year. On the flip side, a team that nets 2 or more close wins will see their record worsen 64% of the time; with three or more net close wins, that number climbs to 71%.

The Irish had 2 close wins (Stanford, Navy) and 3 close losses (North Carolina, Pittsburgh and Syracuse) for a net of -1, so the indicator probably doesn't indicate much for us. (Steele mentions that he sometimes throws out late scores that skew the results; perhaps you could treat the Navy game with its late touchdown in such a fashion, which would lower the ND score to net -2 and provide an indicator of improvement).

The other teams on the ND opponent slate look like this:

2009 Schedule
Close
Wins
Close
Losses
Net
Nevada (7-5) 1 3 -2
Michigan (3-9) 1 4 -3
Michigan State (9-3) 2 1 +1
Purdue (4-8) 2 2 0
Washington (0-12) 0 3 -3
Southern Cal (11-1) 1 1 0
Boston College (9-4) 4 2 +2
Washington St. (2-11) 1 0 +1
Navy (8-4) 4 1 +3
Pittsburgh (9-3) 4 1 +3
Connecticut (7-5) 3 2 +1
Stanford (5-7) 2 3 -1

So Nevada, Michigan and Washington would have a high chance to improve according to Steele, while Boston College, Navy and Pitt would stand to regress.

"Turnovers = Turnaround." Similar to the "Close Wins, Close Losses" metric is Steele's assessment of net turnovers as a predictor of improvement/regression the following year (drawing a line at net double-digit turnovers). In this study he looks over the past 16 years and finds 235 teams that had a net plus 10 turnovers or more. 154 of those then had weaker records the following year (65.5%). He also finds 191 teams minus 10 turnovers or worse, with 130 of them having better records the following year (68%). Again, these are pretty good indicators.

Turnovers Gained & Lost in 2008
2009 Schedule
Gained Lost
Net
Nevada (7-5) 25
24
+1
Michigan (3-9) 20
30
-10
Michigan State (9-3) 24
22
+2
Purdue (4-8) 20
23
-3
Washington (0-12) 10
27
-17
Southern Cal (11-1) 29
22
+7
Boston College (9-4) 36
33
+3
Washington St. (2-11) 13
38
-25
Navy (8-4) 30
15
+15
Pittsburgh (9-3) 24
27
-3
Connecticut (7-5) 25
29
-4
Stanford (5-7) 21
25
-4

Only four teams make Steele's net-10 cutoff: Michigan, Washington, and Washingston State all stand to improve, while Navy should expect a dropoff. Washington State, by the way, led the field in futility; their net -25 was the worst in college football last year.

For the Irish's part, they gained 25 turnovers but gave away 28, for a net loss of -3.