Tuesday, August 28, 2007

What Makes a Heisman?

So, earlier this summer, while wondering the chances of our boy Matt Ryan, I started thinking about what the Heisman trophy really has become to mean in the past few years. It turned into me playing with statistics and trying to look at a numerical method to really try to understand the Heisman race.

Our editor here is totally with me on this, because we've had this discussion before: the Heisman has become less of the "best player in college football" and more the most key valuable part of the best team in the country. At the end of the day, if you play for a team that is not in a major conference and does not have a real shot at the championship, or if you play defense, you have to put up some absolutely ridiculous numbers to get noticed.

Just ask Colt Brennan.

I went back in time, thanks to some archived statistics and the College Football Encyclopedia that sits on my desk. I really wanted to try to break down what parts seem to matter the most to make a player really reflect the "most valuable part of the offense of the best part of the country." So, before I give the results, I want to talk about what I used to build these numbers.

I really broke it down into four things:

1) How good was the team: you have to go back to 1999 to find a Heisman winner who did not at least have a shot at the national championship. That was when Ron Dayne won the statue from Wisconsin, who, at the end of the day, still won the Big Ten and played for the Roses. You don't win the Heisman, no matter how good your numbers are, without a good supporting cast and just an overall good team. This is assumption number one: at the end of the day, these are the best teams in the land, and the idea is to find the player who seems to lead that crop of individuals.

2) How good your opponents are: racking up ridiculous stats against weak competition will not help you win the Heisman. That's why you need to really have an outstanding season outside of a BCS conference to get noticed. Even in a good conference, if you are playing an SEC schedule, the stats will get a small boost because of the quality of opponents. For these considerations in the formula, there is a small bonus given to teams from major conferences (and Notre Dame, just because all of their opponents are from the big leagues, for the most part).

3) Individual production statistics: of course they matter. I'm not that creative to try to dance with numbers to make intangibles more important than the actual facts. To discount the amount of production that an individual contributes is not possible. It's how that plays into the rest of the factors that matter.

4) How much do those stats matter to the overall impact of the offense: here's where I get to play with things. Good numbers on a team where everyone is producing at a high rate will not matter as much as one player really taking charge. The statistic I played with the most was one I created, the Offensive Impact Factor (OIF). This says of a teams points on offense, how many of them is the individual responsible for. As I mentioned above, these teams are all good. They all have fairly decent offensive lines and receivers and running backs, therefore it is fair to give the role players credit for TDs and Yards they accumulate.

Ultimately, the formula comes down to a winning percentage, strength-of-schedule factor, the production factor, the OIF and finally some small bonuses that most voters will take into account (BCS conference, type of bowl game invited). These are small but they can't be ignored.

So, I started going back and testing my ideas of what makes up the Heisman numbers. I started with last year and took the finalists for the trophy based on votes received. I also was interested in a few other people I thought were candidates. So, here is how 2005 worked out with the numbers.

Troy Smith, OSU, QB, OIF=2.208, Score=8.958
Brady Quinn, ND, QB, 2.765, 8.752
Colt Brennan, Hawaii, QB, 3.543, 5.653
Mike Hart, Michigan, RB, 1.640, 7.649
Darren McFadden, Arkansas, RB, 1.953, 6.549
Calvin Johnson, GTech, WR, 2.824, 7.004

Here's why I found this interesting. A few of my notions were right. Even though Brennan's numbers were absolutely ridiculous last year, he got hurt by not being in a main conference and, in addition, not even winning the one he was in. Brady Quinn gets hurt by his strength of schedule and Troy Smith got rewarded for beating Michigan and getting into the National Championship Game.

So, think about this a little bit, I will be back later in the week to post some more of the results. I really would appreciate feedback on the thinking behind this. For the most part, barring one exception in 2003, the formula was pretty accurate. I have a list of 11 candidates across different positions that I will be doing this for all season long. That list will be unveiled tomorrow, along with the results from 2005, 2004 and 2003.

Finally, ACC preview on Thursday. Then kickoff on Saturday. Smell that? It's football season.

No comments:

Post a Comment