The Rocky Top Talk Computer Rankings: What the Numbers Say (and Don't Say)

Rankings are a funny thing.  As complex as football is, as large as the D-1 gene pool is, and as few games are played, the most honest answer would be to not rank at all (or perhaps to rank in batches).  But that just wouldn't be any fun.  We're obsessed with ranking things - with finding out who's "best".  Heck, we even rank the ranking systems in our minds.  That's why, even though I'm highly partisan toward resume rankings, I don't blow a gasket over other preferences.  It's all part of the college football experience and, whether we like to admit it or not, it really does add to the enjoyment of the game.

One thing I appreciate about the system we use is that, rudimentary as it is, it's very easy to sift through the numbers and find out why teams get ranked the way they do.  Additionally, if the rankings seem odd, it's easy to find corrections and figure out what went wrong.  And, once in a while, the numbers stubbornly point out that we're ranking teams out of name brand recognition rather than actually paying attention.  So, with all of that in mind, here's a look at where the numbers are today, and what they tell us about some of the teams.  After the jump will be all of the numbers for all 120 teams, which makes for a rather large table.  But I'll put the commentary ahead of the table so it's easier to find.

Florida vs. Alabama

I mentioned in my SEC Power Poll ballot that Florida was ahead of Alabama for statistical reasons, and the chart shows what I was looking at.  Both teams have the same record and Alabama has played a tougher schedule to date, but the difference in performance within schedules has given the edge to Florida.  The most notable difference is Third Down Offense (3DO), where Florida is off-the-charts good.  Without trying to beat that dead horse (too late), it's a nice luxury to have a shotgun linebacker who is nearly guaranteed three yards on third and short, even when everybody in the stadium knows it's coming.  The other surprise is just how efficient Florida is at passing the ball.  For all we make of their lack of a deep threat, they're one of the most efficient passing teams in the country - far more so than Alabama.  (Look at the PEO column.)  Passing efficiency basically takes the NCAA qb rating system and applies it to the season as a whole, and the one observation is that Florida quite simply doesn't make mistakes passing the ball.  Those are the biggest reasons Florida edged out Alabama in my ballot; the rest of the metrics are within margin of error after accounting for strength of schedule..

Texas

If Texas were to lose a game, they'd plummet pretty hard in the rankings.  As it is, Virginia Tech has nearly eclipsed them despite having a loss.  Why?  First is strength of schedule - Texas just hasn't played anybody good yet.  Second is that their numbers aren't as dominating as they should be for having played such a soft schedule, and it's mostly their first halves.  They don't measure out well on passing efficiency defense and third down offense (relatively speaking) because they let lesser teams hang around for half the game until their opponents wear down.  For all of Oklahoma's problems so far, Texas has their share as well and can't head into the Red River Rivalry too cocky.  Texas is still rightly the favorite, but it's not really that big of a gap; Oklahoma is basically what Texas would be against a decent schedule and a loss or two.

The Value of Defense

One thing you'll observe about our system is that defense gets significantly more weight than offense.  The reason for being defense-heavy is that the weights are primarily derived from the correlations that Doc Saturday (then SMQ) derived when figuring out which numbers most strongly correlated to wins.  Now, the reason that defense correlates more strongly than offense is that defense is more consistent from game to game.  As we saw with Auburn last week, a high-powered offense can end up taking a nap on occasion.  In that game, Auburn lost 3 out of their 4 fumbles, for instance, and did not pick up Arkansas's lone fumble.  The 8 penalties and Chris Todd's inability to hit wide open receivers did the rest.

As a general rule, if I am asked which of two indistinguishable teams is better, I will generally pick the one with a stronger defense.  Offense is just as important for a team; it's just not as reliable.

I Swear I Don't Mean to Favor Tennessee!

Based primarily on the previous point, you'll notice that Tennessee (#33; score of 41.2) is significantly ahead of the next .500 team (Washington, #42, score of 26.6).  Again, it all goes back to defense and strength of schedule.  Tennessee has struggled on the offensive side of the ball, but the defense has held firm against some very tough competition so far.  But the offensive numbers aren't abysmal either; even passing efficiency currently grades out at the NCAA average, thanks to four meh and two Crompton4Heisman!!! passing games.

But this is why I tend to speak so optimistically about Tennessee for the rest of the year.  We're not going to be going into Alabama as even odds in Vegas or anything, but you can know that the defense will show up.  Even with the offense at its worst, the team has been in games against very good opponents, so there's no reason to think that UT can't compete in every game on the rest of their schedule.

Why Arkansas is Hated

Arkansas seems to be a popular choice for the 5th best team in the SEC right now (typically behind LSU and South Carolina).  The problem is that they're surviving on tremendous offensive play and absolutely horrific defensive play.  Take a look at #57 on the chart, where Arkansas resides.  Their defense is below average (usually decidedly so) in all categories but third down defense.  And third down defense shows the weakest correlation to wins of all categories in the rankings.  Meanwhile, Arkansas is explicitly bad in third down offense, which oddly is the strongest offensive correlator to success.  So while the passing game is fun to watch, they're not good at all in the areas that you most need in order to win games consistently.  That's why the computers think people are overrating Arkansas.

Ok, the Table

Here are the weights used, with a caveat.  If you use these weights directly, you will get the right order, but the numbers will be too small.  If you divide all the weights by 0.81 first, then apply them, you'll get the right numbers.  This is because I force the weights to sum to 100 for consistency from week to week, but I figure out the balances in integers.  It's easier that way.

W/L = 25; SOS = 15; Pass Efficiency Defense (PED) = 9; Run Defense (RD) = 8; Third Down Offense (3DO) = 7; Total Defense (TD) = 6; Pass Efficiency Offense (PEO) = 5; Opponents' Points Per Game (OPPG) = 3; Total Offense = 2; Third Down Defense (3DD) = 1

To calculate, take each weight and multiply it by the value in the appropriate column for a team.  Do this for all weights and add the results together.  For example, Florida's W/L of 1.721 multiplied by 25/0.81 is 53.12.  For SOS, it's 0.262 multiplied by 15/0.81, which is 4.85.  Add the two together, and the sum is currently 57.97.  Continue with the remaining columns and there you go.  (It's easiest if you copy the whole table into Excel and go from there.)