The Challenge of Bowl Game Analysis
Usually, teams that square off in bowl games have at most one or two common opponents. Riding your analysis on one game when college football teams have such a variable level of performance is dangerous, and that's made obvious here. Ute fans will point to the Michigan game for evidence that the Utes are significantly better, and BYU fans will note that their margin of victory in the Utah State game was vast compared to how the Utes did.
In this case, margins of victory overall give a slight advantage to the Cougars. Overall, they went 126-93 while the Utes went 102-72. There's a ton of noise inherent in final scores, like special teams scores, return touchdowns, and injuries which change the face of a team temporarily. We can use sophisticated statistics to give a much more complete impression of how these teams performed against their four common opponents.
A Better Way to Keep Score
Measuring teams by how many points they score limits your data. The Utes should get some amount of credit, for example, for the drives against UCLA that went well until they stalled. BYU's defense gave up a handful of long plays, but they had some great defensive drives that were lost when a player went the distance.
We can do better than yards per game or yards per play, though. By measuring the performance of BYU and Utah against how each of their opponents performed against the rest of their schedule, we can form a neutral statistic I'm calling Yards+. A score of 100 means Utah performed as average on the play. Scores higher than 100 mean the yards were more than average, while scores less than 100 mean there were less. This has different implications on offense and defense.
These charts are histograms, showing how Utah performed in Yards+ on relevant plays. The Utes' strong defense means that significantly more than 50% of the Utes' plays fall into the first three buckets, which run from the best defensive play of the sample with a score of -129 (the tackle for a five yard loss against Fresno State where 'Seni and Lowell blew up the line) to the worst play with a score of 1544 (the 82 yard touchdown pass near the end of the Fresno State game). In all, 66% of the 270 plays scored fall into the first three buckets, which you would call successful plays for the Ute defense.
The Utes offense, meanwhile, is clustered a little lower than you would like, with plenty of below-average plays. Like the defense, there's a tail at the end of the chart where the Utes happened to blow the doors off the opposing defense.
These stats are good, but there aren't really enough of them. With only four games and a substantial number of garbage plays for both teams, extra measures are needed to create a true impression of how you can expect the Utes and Cougars to perform moving forward. A sampling distribution* smooths out the curves, takes away the overpowering influence of a handful of big plays, and creates a more accurate picture.
These statistics paint a different picture than the 'offense vs. defense' narrative that's been developing. BYU's offense is capable of being explosive, but not at a rate that makes up for a surprising number of plays where they were just below average, in terms of yards per play. Utah makes its bread and butter on offense not by being hugely successful, but by not failing hardly at all, at least compared to the Cougars.
The defenses have a similar profile; BYU has a fairly low number of average performances, trying to make up for it with a high number of successful defensive plays. In other stats, we see this in BYU's high sack numbers but modest efficiency scores. The Utes have played a conservative defense this year, limiting their opponents to slightly below average performances on most plays and giving up comparatively few long plays. Utah's low havoc totals (as compared to last year) have been replaced with a dominant ability to stop long plays.
There's something else we can do with these 'yards+' numbers. We can break them down into down, distance, and pass/run components, figure out the probabilities of particular results, and model an entire game from start to finish. We can make adjustments for things like home field advantage, recruiting rankings, and other things that are difficult to measure but we feel confident impact the outcome of a game. If we wrote a computer program that ran 10,000 simulations of this game, it would give us percentage chances of victory for each team an a likely final score. This is precisely the approach that statistics like FPI and F/+ use to determine their outcomes.
Advanced Stats Overview
Utah gets a significant bump from the sample we used above, particularly on offense. The most likely reason for the precipitous decline in Utah offense over the end of the season is the buildup of injuries at skill positions. BYU isn't an exceptional defensive team, but Utah is going to struggle to score if they don't either recover major pieces like Covey and Scott or figure out how to replace their production.
BYU isn't spectacular in any particular area, ranking in the mid-thirties to forties in both offense and defense. The Utes are what they've always been on defense, but the offense has slipped to 2014 levels. If the Utes haven't recovered their offensive mojo, these stats project a close, hard fought, and low scoring game like we're so used to seeing from the Holy War. If Covey and Scott are back at full strength and the Utes are putting a low 50s offense on the field, it might turn into a blowout.
Vegas and the Polls
This game opened as a close one, with Utah getting a three point edge. It's slid a hair closer since, down to the Utes being 2.5 point favorites. Massey has a wider gap between these two teams, with the ratings aggregator putting the Utes at 24th and the Cougars at 38th.
I don't expect Utah's offensive decline to completely reverse itself; Covey's high ankle sprain is an injury that limits effectiveness even after the player has returned to the field, and although Joe Williams is a good player and is giving everything he's got, Devontae Booker's absence is a significant hole in the scheme. On the other hand, I was surprised to learn that BYU's offense was relatively weak this season and has struggled in a number of critical ways. The media impression of them as a firepower team was bolstered by a few hail mary passes to begin the season. I am projecting a defensive slugfest with the Utes winning the turnover battle and severely limiting the run game, putting Mangum into some really bad spots.
Utah 20, BYU 13
*A sample distribution is formed by taking a large number of small samples from your population and measure each sample's average (in this case, I took 1,000 samples from each team's offensive and defensive performance) and then plot the sample means on a histogram. Since you know that sampling creates a normally distributed set of numbers, you get a nice smooth distribution that doesn't have as many holes in it.