Well, at least that's what some "fun with numbers" says.
I've been enamored with looking at yards per play differential as an indication of a team's strength. Basically, if two teams play and they both average abour 5 yds/play, they are equal. If one averages 6 yds/play and the other averages 5, the one with 6 is better, and the yds per play differential is 1 yppd.
So, I decided to look at our schedule's yppd. And not only the teams we play, but they teams they play. So, for example, ND's average yppd is currently 1.0. Michigan's is 2.4. If I average our opponents (only the one's we've played), our opponents are at 1.4 yppd. Or, on average, they are 1.4 yppd better than the teams they've played.
Then, I thought looking just at our opponents wasn't enough - what if someone's played a very easy schedule. So, I looked at our opponent's opponents yppd (confused yet). Not too surprisingly, our opponents have played relatively easy schedules and have a -0.5 yppd. In another words, the average of USF, UM, MSU, and Pitt's opponents typically gain 0.5 yds less per play than their opponents.
Taking these three numbers - our yppd + our opponent's yppd + our opponent's opponent's yppd... and added them to create a number - of how good ND (and other teams) are. You could think of this as - how good doe ND win the line of scrimage relative to our opponent's - and then adjusting for strength of schedule.
Based on this, here's what the numbers say - ND is better than any team on our schedule:
Notre Dame 1.9
Stanford 1.7
Michigan 1.5
Michigan St. 1.2
Southern California 1.1
Navy 0.9
South Fla. 0.6
Pittsburgh 0.5
Air Force 0.2
Maryland (0.0)
Wake Forest (0.1)
Boston College (0.4)
Purdue (1.3)
And Purdue is the worst team we'll play. Granted, this is on a limited set of data... and should become more relevant as the season goes on. But if this is any indication, we have played as well as any team we're going to play (save the turnovers and boneheaded mistakes).
For fun, here are the top teams according to this logic for this year:
Alabama 2.8
Georgia Tech 2.7
Nebraska 2.1
Virginia Tech 2.0
Florida 1.9
Wisconsin 1.9
Notre Dame 1.9
Texas A&M 1.8
Stanford 1.7
South Carolina 1.7
UCF 1.7
Baylor 1.5
LSU 1.5
Michigan 1.5
Penn St. 1.4
North Carolina 1.4
Illinois 1.2
Michigan St. 1.2
Tennessee 1.2
San Diego St. 1.2
Washington 1.1
Texas 1.1
Southern California 1.1
Georgia 0.9
Navy 0.9
Looking at the list and then some of the underlying data, I think this analysis would be better served with more game data (for example - Air Force only has one data point, the game against TCU; Tenn has only 2 games of data, Baylor only 2, etc... b/c I excluded all non- DIA games from the analysis).
But, this would suggest we should feel very good about Purdue (although Purdue only has two data points...). And all be said about a night game, them having a week off, etc - we should have the ability to beat them pretty badly. It will be interesting to see what happens.
Originally, I had thought of forcing a 50/50 mix of the two to have more balance. I combined this with the weightings of 1.3 & .7 (so they add up to 1) and then weighted the yppd too, so: a team gets 100% of their yppd, 67% of their opponents, and 23% of their opponent's opponents. Here's the new top 25:
ReplyDeleteGeorgia Tech 5.0
Alabama 3.5
Virginia Tech 2.9
Wisconsin 2.6
Florida 2.6
Stanford 2.4
Nebraska 2.3
Notre Dame 2.1
North Carolina 1.9
Texas A&M 1.9
Navy 1.9
LSU 1.8
South Carolina 1.7
Illinois 1.6
Baylor 1.6
Michigan 1.6
Texas 1.6
Temple 1.4
Penn St. 1.3
UCF 1.3
South Fla. 1.2
Michigan St. 1.2
Southern California 1.2
Utah 1.1
Oregon 1.1
And then ND opponents:
Stanford 2.4
Notre Dame 2.1
Navy 1.9
Michigan 1.6
South Fla. 1.2
Michigan St. 1.2
Southern California 1.2
Pittsburgh 0.7
Air Force 0.5
Maryland (0.6)
Purdue (0.6)
Wake Forest (0.8)
Boston College (1.0)
Navy is suffering from a lack of datapoints... but it is interesting to see that by this measure, our schedule is pretty tough. Purdue is still towards the bottom of the list. And BC is the 88th best team by this analysis - which speaks to a pretty strong schedule overall.
I expect this analysis will get more interesting as the season progresses. Right now the top 25 teams have a bunch of teams that play a very easy schedule. For example, Wisconsin is at +3.8 based on head-to-head, but the schedule adjustment is -1.2, which leads to the 2.6 overall. ND is at +1.3 based on head to head with a schedule adjustment of +0.8 (I should note that I've excluded I-AA games). When ND plays easier teams, it may get even better numbers and when Wisconsin plays some real games, it may bring their production down. Also interesting that pretty much every team at the top of the list has a schedule adjustment implying an easy schedule - GT -0.9, AL -0.7, VT -0.3, WI -1.2, FL -1.3, Standford -1.2.
When i ran the numbers for Purdue, I threw out the I-AA game.
ReplyDeletePurdue - in head to head with Rice & Midd TN, is +0.4 yppd (using the adjustments for run/pass detailed in my response to OD above).
I then used an adjustment for strength of schedule (based on Purdue's opponents - and their opponent's opponents) that came to -1.0.
So, all together, they came out at -0.6.
ND is at +1.3 yppd and with a SOS adjustment, add +0.8 to get to +2.1.
This would mean that ND should be +2.1 - -0.6 = +2.7 yppd for the game.
Last weekend, here are the games that were at +2.5 yppd or more:
Alabama 38 Arkansas 14
Baylor 56 Rice 31
Fresno St. 48 Idaho 24
East Carolina 28 UAB 23
Florida 48 Kentucky 10
Michigan St. 45 Central Mich. 7
Nebraska 38 Wyoming 14
Penn St. 34 Eastern Mich. 6
South Carolina 21 Vanderbilt 3
South Fla. 52 UTEP 24
SMU 42 Memphis 0
So, if this data is worth its salts, we should have a good margin of error for winning this weekend.
Assuming Crist gets credit for:
ReplyDelete2010: Boston College, Michigan St., Navy, Pittsburgh, Purdue, Stanford, Western Mich.
And Rees gets:
2010: Army, Miami (FL), Southern California, Tulsa, Utah
2011: Michigan, Michigan St., Pittsburgh
Neither gets credit for (my database is by game, so it's hard to split):
2010: Michigan, South Fla.
Using this, Rees averaged 5.7 yds/play; Crist averaged 5.6 yds/play
Our opponents, on average throughout the season gave up:
5.4 yds/play for Rees’ opponents
5.3 yds/play for Crist opponents
So, both have had a 0.3 yppd. If you factor in the opponents’ & opponents’ opponents’ yppd (as I did above), you get a:
Rees: 0.8
Crist: 0.4
Suggesting that Rees & Crist have performed very similarly, but Rees has played tougher competition overall.
Interestingly, over these games, 50% of plays were runs with Rees, while only 42% with Crist. The yds/rush were much higher with Rees – 4.8 vs 3.8. Crist has a better yds/attempt – 6.5 (rees) vs 6.8 (crist).
For more fun - here's what the top 25 from 2010 using this analysis:
ReplyDeleteteam---yppd---wins
Boise St. 3.4 12
TCU 2.4 13
Ohio St. 2.3 12
Alabama 2.2 10
Hawaii 2.1 10
Oklahoma St. 2.0 11
Auburn 1.9 13
Oregon 1.6 12
Arkansas 1.5 10
Northern Ill. 1.5 11
San Diego St. 1.4 9
Iowa 1.4 8
SMU 1.4 7
Nebraska 1.3 10
Nevada 1.2 13
Stanford 1.2 12
South Carolina 1.1 9
Wisconsin 1.1 11
Texas A&M 1.0 9
Pittsburgh 0.9 8
Miami (FL) 0.9 7
UCF 0.9 11
Georgia 0.9 6
Michigan St. 0.8 11
Arizona 0.8 7
And, here's how ND has been over the last 4 years...
2008 (0.1)
2009 0.4
2010 0.7
2011 1.9
This suggests that we have made some pretty good progress this year as we've played some of the hardest games on our schedule.
I should note that I just built this model and there may be bugs in it... but I think the data is pretty good.