Wednesday, October 18, 2006

Dropping in on the Conversation | by Jay

With the first BCS standings finally out, it's probably a good time to jump back into the Conversation and see where we are. Remember, the chief codicil of the Conversation is that we judge a team simply by what they have proven on the field this year. This is not a Power Ranking. This is not a discussion of "nine times out of ten, team A would beat team B." There is no speculation on the future allowed, and no lingering, hangover reputation from last year permitted. We don't rank teams by perceived strength, how they played, "style points", or how many scores they ran up against an inferior opponent. Generally speaking, we're trying emulate the other major American sports, where subjectivity, whimsy, and irritable bowels by grumpy human pollsters have no bearing on a team's fortunes. We're striving for simplicity here -- a common sense snapshot of which teams have accomplished the most so far.

This is best measured by an arcane statistical metric called Wins. (Radical, huh?)

(By the way, the fellas over at MarkMayBeWrong have been keeping tabs on the Conversation week-to-week with posts like this. Recently they posted a kickass, interactive table of all the data on wins and losses to date in the college football season; feel free to use it as a reference through the rest of this post. They've also taken the initial idea behind the Conversation and added a few wrinkles of their own, including an emphasis on which teams have done better than .500 against BCS competition and a tier system that divides teams into three groups. Great stuff.)

So the other night I sat down, poured myself a Jameson, pulled up the game results, and took a look. Here's my methodology. See if this makes sense.

The first thing I wanted to do was pare down the list of 117 Division 1-A teams to a reasonable number of teams that have actually "proved something". I know we got a little pushback on using BCS competition as a baseline from some commenters in the inaugural post, but you have to start somewhere. The vast majority of top college football talent flows each year from the high schools and into the 65 BCS conference teams (plus Notre Dame), so generally speaking, beating a BCS team would qualify as a "quality win" for our purposes. (A good non-BCS team can still get into the conversation, but only if it has proved something against BCS competition).

So the first thing was to look up how many wins each team in Division 1 has against BCS competition, and some interesting details immediately began to bubble up. As we know, almost every team has played six or seven games so far. And yet out of all those teams, only 42 have 2+ BCS wins so far. That seemed like a pretty good natural cutoff.

The next step was to try and get a relative measure of strength of schedule for each of these teams. Keeping with the basic precept that Wins are uniquely important, I looked at opponents wins and losses for each team and came up with this simple tally.

for Team X...
Give one point for every BCS win.
Subtract a point for every loss (BCS or otherwise).
Give one point for every BCS win by a team that team X has beaten.
Subtract one point for any loss by a team that team X has lost to.
This rewards a team for beating good opponents, while penalizing a team for losing to poorer ones. I added it up, then sorted the teams by point total. There were 21 teams that had at least a +4 in the points column; again, that seemed like a good natural cutoff.


Any Loss
Opp W
Opp L
Southern Cal 6 0 +13 0 19
Michigan 6 0 +12 0 18
Auburn 5 1 +12 -1 15
Ohio State 5 0 +10 0 15
California 5 1 +9 -1 12
Notre Dame 5 1 +8 0 12
Arkansas 3 1 +9 0 11
Florida 4 1 +9 -1 11
Clemson 3 1 +7 -1 8
Tennessee 2
Wake Forest 5 1 +5 -1 8
Louisville 4 0 +3 0 7
Oregon 4 1 +5 -1 7
Georgia Tech 3 1 +5 -1 6
Rutgers 3 0 +3 0 6
Texas 3 1 +4 0 6
West Virginia 3 0 +2 0 5
Boston College
Missouri 3 1 +3 -1 4
Nebraska 3 1 +2 0 4
Wisconsin 3 1 +2 0 4

Of course this is just the initial, rudimentary index, and doesn't take into account the best indication of relative value: head-to-head matchups. Each one of these 21 teams have proven enough to get into the Conversation, but there a bunch of games within the top 21 where teams have already played each other. These actual, battle-tested wins should always reign supreme when comparing teams -- we don't want a situation where, say Arkansas has beaten Auburn yet ends up below them in the ranking. That would be a travesty. Let's look at the win chains:
Southern Cal > Arkansas > Auburn > Florida > Tennessee > Cal > Oregon
Southern Cal > Nebraska
Michigan > Notre Dame > Georgia Tech
Michigan > Wisconsin
Ohio State > Texas
Boston College > Clemson > Wake Forest
This is where it gets fun. This is where the jigsawing starts. We need to splice those chains in amongst each other, but keep the win order intact. Luckily we don't have any circular series yet (where team A beat team B who beat team C, who in turn beat team A) so it's a little easier. (How are we going to handle those conundrums in the future? I guess we'll cross that bridge when we come to it.) We also have four teams who have not played anyone else in the Conversation yet: Louisville, Rutgers, West Virginia, and Missouri. Where to slot them?

I'm still moving around teams like so many Scrabble pieces, so I'll throw it out to y'all. Your mission, should you choose to accept it: come up with a sensible way to rank these 21 teams, holding true to the relative strengths of schedules, and keeping those head-to-head win chains intact.