Funky Analytics: Don't be surprised if the Offense Struggles

So a cougarboard post http://www.cougarboard.com/board/message.html?id=12842724

using data from Football Outsidershttp://www.footballoutsiders.com/varsity-numbers/2014/varsity-numbers-updated-2014-projections

got me thinking.



Using Football Outsiders rankings of each team's offense and defense, I try to see how BYU would fare against each opponent's opposing unit then combine those scores to get a sense of the winner overall.



So, for example, BYU's defense is ranked 20th (column C). I assign the D points based on the number of teams against whom it has a higher ranking: 128 teams total, so BYU's D gets 120 points (Column D). From that I subtract the number of similarly-calculated points for the opponent (Columns A and B) to get the number of points (or ranking spots, if you like) the D earns the the team (col E). I do the same thing with BYU's Offense: 128-H=F and opponents (128-F=G) to get the "points" earned by the Offense: I-G=J. I then add the Defensive points and offensive points to get the total rank differential.



Any number less than 0 is a big trouble spot, and the more negative, the more trouble. I figure that you should spot the home team about 10 rank points per side (Offense and Defense) for a 20 rank point home field advantage--but that number's kind of arbitrary.


BYUanalytics2014.png

Here's what I see:

BYU's D should look great most of the season, with the possible exception of Nevada and UCF and maybe Boise.

BYU's O will look terrible, simply terrible throughout the first half of the season, but should pull it together for the second half.

All combined, BYU's 2 toughest games look to be UCF and Boise State! All the more so when you figure that those are both away games. Texas and Houston should be the same level of difficulty, except that Texas is away, making it dangerous, and Houston is at home.



A few challenges/caveats/assumptions:

This treats ordinal data (ranks) as though they were cardinal.

This weights the O the same as the D in terms of contribution to success

This takes the rankings from Football Outsiders as accurate--they're probably the best around, but they're still basing things off of what they observed in last year's data and projecting it to this year. Really can't get around that. With that, I think our O is ranked a bit too low and our D is ranked a bit too high; I also think that Texas is ranked too low and Houston is ranked too high.

Not to mention UCONN's new head coach didn't have much trouble slowing down BYU when he was OC at ND....

CrimsonCoug wrote:
A few challenges/caveats/assumptions:

This treats ordinal data (ranks) as though they were cardinal.

This weights the O the same as the D in terms of contribution to success

This takes the rankings from Football Outsiders as accurate--they're probably the best around, but they're still basing things off of what they observed in last year's data and projecting it to this year. Really can't get around that. With that, I think our O is ranked a bit too low and our D is ranked a bit too high; I also think that Texas is ranked too low and Houston is ranked too high.

The rankings give some good overview of trends that may loosely correlate to show how BYU's offense and defense might matchup against the 2014 opponents, but there are some pretty significant caveats and challenges. For instance, the #25 offense and the #40 offense may be very similar. Whereas, the #5 offense and the #20 offense may be miles apart - or the #49 defense may be light years ahead of the #64 defense.



Also, I didn't see any indication what the rankings were based upon. Is it yardage? Scoring? Combination? For instance, BYU offense was #15 in yards per game; however, it was #56 in scoring. Both are important metrics. Also, do the rankings account for '13 versus '14 schedules?



Also, I question the (even ballpark) soundness of some rankings. UCF is ranked #25 for offense. In 2013, UCF was #31 in scoring and #41 in yards per game. They lost their #3 NFL draft-pick QB and starting running back - and yet, they are predicted to be BETTER in 2014? That does not compute.



I also note that Utah St.'s offense is predicted to rank #124. That means there will only be 4 worse FBS offenses in 2014 and MTSU, UNLV, Nevada, Cal, and Virginia will all be better. Utah St. finished 2013 at #61 in yards/game and #51 in scoring - mostly without Chuckie. How are they the 5th worst FBS offense in 2014? Most analysts predict Utah St. to contend for the MWC championship and some have anointed Chuckie as a dark-horse Heisman candidate. Does not compute.



Conversely, Utah St. finished 2013 ranked #12 in defensive yards/game and #7 in defensive scoring. The Aggies lost 3/4 of the secondary and most backups, 2/3 of the D-line, and an all-star linebacker. The defense will still be good - but how can you justify to give USU the #7 ranked defense in the country? You can't.

snoscythe wrote:
Not to mention UCONN's new head coach didn't have much trouble slowing down BYU when he was OC at ND....

I have to agree here. UConn's defense is supposed to be their strength. Tomorrow could be a very boring game with a lot of punts.

I'm expecting lots of kinks in the offense. I know the offense is supposed to be on the rise, but until I actually see it performing better than last year I am assuming it will be the same.



UConn's coach knows what he's up against. Given Anae's inability to scheme, that doesn't bode well for us. We'll likely score a couple TDs in the first half and then cross our fingers the defense holds up.

YNot wrote:

The rankings give some good overview of trends that may loosely correlate to show how BYU's offense and defense might matchup against the 2014 opponents, but there are some pretty significant caveats and challenges. For instance, the #25 offense and the #40 offense may be very similar. Whereas, the #5 offense and the #20 offense may be miles apart - or the #49 defense may be light years ahead of the #64 defense.

Yup. That's the ordinal/cardinal problem I mentioned. With cardinal data the space between 0.01 and 0.02 is the same as between 453.04 and 453.05; Not so with ordinal data. Still, I think that while there will be more space separating the ends of the distribution, the middle of the distribution is probably fairly even. With a few exceptions, most of the rankings are not at the tails.

YNot wrote:
Also, I didn't see any indication what the rankings were based upon. Is it yardage? Scoring? Combination? For instance, BYU offense was #15 in yards per game; however, it was #56 in scoring. Both are important metrics. Also, do the rankings account for '13 versus '14 schedules?

Here's from the Football Outsiders website: "Football Outsiders official college football ratings, combining play-by-play and drive-based efficiency systems developed independently by Bill Connelly and Brian Fremeau. F/+ ratings represent the best measure of overall college team performance we've measured in terms of predictive quality and retrodictive analysis." Elsewhere they say that F/+ also accounts for strength of the opponent.



Frustratingly, I can't find anything on FO about what they base their predictions on in weighting last years observed data & next years opponents.

YNot wrote:
Also, I question the (even ballpark) soundness of some rankings. UCF blah blah blah... Utah State blah blah blah... Chuckie Keaton's friggin awesome blah bleeh blo blah...That does not compute...how can you justify ... You can't.

I'm not justifying anything, brother. Tain't my data. The F.O. guys have come up with an answer for that, though:



"FO Angry Troll Mad Lib: Template created by reader zlionsfan to be used when fans who are unfamiliar with FO's statistical methods wish to complain about the placement of their favorite team in that week's FOXSports.com power rankings (based on DVOA). Reads as follows:

<team> is clearly ranked <too high/too low> because <reason unrelated to DVOA>. <subjective ranking system> is way better than this. <unrelated team-supporting or -denigrating comment, preferably with poor spelling and/or chat-acceptable spelling>"



The FO guys tend to do a pretty good job on average, but they're going to be more wrong about some teams than others.



Look, I agree with you; there are some big problems, especially with the Utah State data--so much so that I wonder if the USU entry was a typo. That's why I didn't make a big deal about the USU result--I think they're off on both O and D, but probably about right summing up the two of them.



All models are wrong; some models (or parts of models) are useful. I'm just using it to try to ground my expectations on more than subjective talking heads. And, acknowledging the flaws of the data, I find my expectations tempered, particularly on offense. So I'm glad Bronco has spent more time with the Offense in the off-season, and I'm not going to break out the torches and pitchforks if we struggle a bit in the beginning, or stumble against UCF. I'm also not ready to call the Defense world beaters if we just hold the opposing O to less points than ours--we need to go beyond that to merit high praise. Glad that our new D-Coord gets to cut his play-calling teeth on less threatening opposition.



Now...On To Game Day!