Lindy's Five Essential Websites (Non-Major Media) for 2013
[+] Team Summaries

Saturday, June 1, 2013

Measuring the Golson Effect

With Everett Golson out at Notre Dame, what impact will his poor academic judgement have on the Domers in 2013? That's a very reasonable question, but it's not the question I'm going to answer. There are too many variables and too little information. Instead, I'm going to try to answer a different question: What would Notre Dame 2012 have looked like without Golson? Then, I will use that answer to speculate on 2013.

To answer my question, I make a few assumptions/simplifications. First, I assume that if Golson were not playing Tommy Rees would have played in his place. In the case of the Navy game that is obviously not true. I don't care. Notre Dame would have won that game with me at quarterback, so it doesn't really matter.

Second, I measure passing performance using EPA: expected points added. EPA is a statistic that converts a player's performance into points by estimating each play's impact on the expected points a team will score on that drive. It has a long track record of predicting Heisman winners, and I find it to be the best single metric for measuring performance. Using the EPA I find that Rees was .053 points worse per pass attempt than Golson (.204-.152 plus rounding error). I assume that we can estimate Rees' performance in place of Golson by subtracting .053 per play from Golson's performance in that particular game. (An EPA/per pass of .204 is not good, but above average; the elite quarterbacks hover around .4.)

Third, I assume that Rees would run at his own rate at his own average level of success, not at the same rate and at the same level of success as Golson. I assume that the rest of the running plays would be picked up by the running backs, and that their performance on those plays would have been average (EPA=0).

For the season as a whole, I find that Golson made Notre Dame 20 points better than they would have been with Rees at quarterback. Most of that difference comes from the Pitt game (12.6 points). Against Purdue, Michigan and Stanford, the model predicts that Notre Dame would have been better off with Rees in at quaterback for the entire game, and because Rees played the entire game against BYU, those results are unaffected by the substitution. In four games, the difference is less than a field goal, and in seven games the difference between Rees and Golson at quarterback is less than five points. Only against Pitt would the hypothetical performance gap between Rees and Golson been enough to change the outcome of the game.

What does this mean for 2013? First, Golson is clearly better than Rees. Higher upside for Golson means the gap between the two would have grown. This difference will probably be more important in 2013 as Notre Dame will depend more on the quarterback as a play maker. But the quarterback is not the only player on the field. I'd put the conservative estimate of the cost of losing Golson in 2013 at 2 points per game, and the unconservative estimate at a field goal per game. That figure balloons if Rees goes down early with injury. But in the end, if Notre Dame can keep six opponents under seven points and all but two under 20, they'll still have a puncher's chance at 10 wins.

No comments:

Post a Comment