Wednesday, August 15, 2007

Inventing A New Metric: Strength of Schedule

Most of the mainstream statistics, often quoted as they may be, are generally worthless in terms of evaluating the true quality of football teams. And no statistic is more worthless than the one generally used to tabulate strength of schedule.

Here's a list of strength of schedule in 2006
, but don't even waste your time clicking the link, it's not worth the time. When strength of schedule is tabulated, how do they do it? Elementary style, that's how. They just add up the cumulative records of a team's individual opponents, do that for every team, and then rank each team based on opponent's winning percentage. At bottom, it's first grade logic mixed with third grade math.

There are just so many things wrong with that approach.

To begin with, teams with equal records are, in fact, not equal. That's a rather simple insight, but nevertheless it is something that the traditional strength of schedule tabulation does not take into consideration. For example, last year the USC Trojans and the TCU Horned Frogs both went 11-2, but is there really any doubt as to who was the far superior team? Of course not, but even so, both teams count the exact same in strength of schedule.

Beyond that, a few outlier teams can drastically sway a team's strength of schedule ranking when using the traditional tabulation. Take Notre Dame in 2006, for example. Critics argued that the Irish weren't particularly good and mainly won ten games because of a weak schedule. Notre Dame defenders, however, quickly pointed to the fact that the Irish schedule wasn't really that bad, and was in fact the 39th toughest out of the 120 FBS teams when using the traditional tabulation. But that's an incomplete response at best. Even if you use the traditional tabulation, once you break it down even further it's a poor argument. In reality, Notre Dame's ten wins came over teams with a combined record of 56-67 (.459 winning percentage). On the other hand, their three losses came against teams with a combined record of 31-5 (.861 winning percentage). At bottom, Notre Dame beat up on combination of the Sisters of the Poor and some mediocre teams, and then were annihilated when they faced good teams. The underlying reality is that you can have a poor schedule as a whole, but play a couple of really good teams and that will sway your strength of schedule rating into looking pretty solid.

Moreover, what about those other odd-ball things that affect games, like injuries? Obviously, those aren't accounted for, but they should be. Take Tennessee in 2006 for example. In a critical stretch (against LSU and Arkansas), Erik Ainge was injured. Now obviously, Tennessee was a much easier opponent without their star quarterback on the field, but in terms of strength of schedule, they go in the books as a 9-4 team regardless of it you faced Ainge or his back-up Johnathon Crompton. Obviously that's not right.

At bottom, without going into further detail, we really need to invent a new metric for determining strength of schedule simply because the current one is effectively meaningless, and because strength of schedule is highly important.

Nothing is concrete as of this minute, but I'm trying to come up with something for the 2007 season.

At the moment, though, I have a few ideas for how the system would work:
  • Division 1-AA teams would not count in any way whatsoever.
  • Non-BCS teams would count, but to a significantly smaller degree. For example, an 11-2 TCU team would not count as much as an 11-2 USC team.
  • Significant injuries (such as to a quarterback) will make a particular team count less in terms of strength of schedule
  • Expand the data by looking at not just wins and losses, but more advanced statistics such as Pythagorean Wins and the like.
We'll see as the 2007 season progresses.

1 comment:

Unknown said...

I'd also like to credit the "Roll Bama Roll" guys for giving me a new notion related to strength of schedule: bye weeks. Todd and Nico pointed out the Alabama plays 5 teams coming off of bye weeks....the next most is LSU playing 2 teams coming off bye weeks. Thoughts?