This edition of the blogpoll ballot draft has been delayed for two reasons: one is that I've been swamped in weekend work, and the other is that I was finally at the point where I could re-investigate the traditional RTT computerized BlogPoll ballot thingamajig that Joel used in previous years. When he handed me the keys to the ballot, he handed me free reign over numbers. Mmmm, tasty, tasty numbers. So I tinkered around with it, but I was never happy with it - at least not enough to actually use - until now. So we're returning to the computerized ballot for the draft, and I'm ok with listening to overrides for the final. There's a lot of dialing in left to do, so we'll see how she goes. But first, here's the ballot. And remember, this is a 100% resume ballot, and as such, it doesn't care what your team's name is. Performance metrics are everything, and early-season metrics are wonky (due largely to the lack of good, competitive games and the abundance of cupcake diets). So yes, there's a lot of movement.
|24||North Carolina State|
|Last week's ballot|
Ok, first the raw data. Explanations and observations come next.
|24||North Carolina St.||0.278||-1.888||1.026||1.191||1.099||2.138||1.170||1.486||-0.199||1.870||51.500|
WL = Win/Loss; SOS = Strength of Schedule; PED = Pass Efficiency Defense; RD = Run Defense; 3DO = 3rd Down Offense; TD = Total Defense; PEO = Pass Efficiency Offense; OPPG = Opponents' Points Per Game; TO = Total Offense; 3DD = 3rd Down Defense; and W SUM = Weighted Sum (the score that counts)
- What do those numbers mean? Unlike previous editions of this ranking thingy, this one no longer relies on the rank values. For example, the 3rd down defense (3DD) is not assigned as #1, #2, #3, etc., but instead is a scaled value of the actual 3rd down success rate of a team. The technical term is mean centered, unit variance (MCUV). So the most average 3rd down performance will score zero, and a 3rd down performance that is one standard deviation better than average scores one. Subpar performances are negative. This better captures slight and large differences. For example, Florida's 3rd down offense (3DO) is the best in the nation at 2.586. The next best is South Florida at 2.001 - a whopping half standard deviation difference. In the old system, this would have been 1 and 2, respectively.
- How is the weighted sum calculated? The actual score is the weighted sum (W SUM) column. This is done by adding up all the previous columns after having applied a weight to each one. All of the weights will add up to 100, so a W SUM score of 100 means that your weighted average of all the columns is a standard deviation of 1 (note: this is not the actual average). Negative scores mean your team scores below average. This is where things will change over the season. With the new system, I don't yet have a handle on how the weights should be distributed, so I will continue to tinker with them as time goes on.
- Why the change? This gives two advantages. The first was discussed in point 1, where the large differences and small differences aren't forced to the same delta. The second is that it's a lot easier to gauge the relative performance of teams in a given category, as well as across categories. Florida's #D) of 2.586 is staggeringly high, and that value would be eye-popping high in any category. (Note: part of this is due to Florida playing 2 cupcakes. Part is the Tebow extending drives in the second half against UT.)
- Where did the strength of schedule data come from? SOS is borrowed directly from the Colley Matrix. Despite the hatred that some people have for it, I used it because (a) it does a better job of considering the opponent than most other SOS indices, and (b) it's updated every Sunday, so I can data link it. (The NCAA puts theirs in pdf form, which makes it awkward for data mining.) So Colley it is.
Southern Cal? Southern Cal came in at 32 primarily because of their 3rd down offense (3DO). Did you know they were 112th in the nation right now? Neither did I. 3DO is a measure of the ability of a team to extend drives in a must-make situation, and USC isn't getting it done. An average 3DO alone would bump them to around 21-22, but an average 3DO would likely have them beating Washington, bumping them right back up in the top 10ish.
- Ohio State? Ohio State has a slew of good-not-great meh numbers. With the loss, they just don't have anything to bump them up. beyond some other one-loss teams. They are at 40 and are immediately followed by BYU and FSU, who have terrible defensive metrics right now.
- Cincinnati? A good SOS (which is rare at the top) and solid all-around numbers.
- Miami? A ridiculously high SOS atones for pass efficiency defense (PED) problems.
- Iowa? Yeah, nearly-beaten-by-a-1-AA-team Iowa is way up there. Part of that is an artifact of Colley's SoS, which lumps 1-AA teams together for composite scores, but it will settle down as the season goes on.
- Texas? Killed by SOS and a mediocre outing against Wyoming. Teams above them have benefited from the slaughter of innocent women and children - another artifact that will die down as the season gets into conference play.
- Tennessee? Tennessee is at 55, and is the highest 2-loss team in the rankings. I'll include their scores when I present my SEC Power Poll ballot. The bottom line is that their defense is ridiculously good and the offense is not as shabby as you might fear.
- Western Kentucky? They're the bottom of the barrel, with a whopping -178.131 score. It's bad down there.
For a glance at how the scores line up:
It's heavy at the bottom thanks to some truly horrific receiving ends of the cupcake games. But the overall distribution is reasonable. You can see some very slight banding due to the win/loss effect, but it's not overwhelming. More to come on all of this later in the week.