Monday, January 18, 2016

bronco's defense

Greetings and happy new year and happy unbirthday and all that.  It's been a while since I punched anything into this box, but I needed a bit of time to work on the analysis I have today.  If I'd known the basketball team was going to go to hell in a handbasket I might've saved the analysis for later.

What I've been wanting to do is quantify BYU's performance on offense and defense during the Bronco Mendenhall years there and compare that to UVA's performance at the same time.  National rankings are nice, but they don't really satisfy.  We need to have a way to take strength of schedule out of the equation.  So I got to thinking about how to do that.

A simple way for defense would be to take a particular game and see if BYU held that team to fewer yards than they typically muster, and on offense, the opposite.  If BYU were to hold a team to 200 passing yards, when that team averages 250, that's a good performance, right?  We could take each opponents averages and then how they performed against BYU, and produce an answer both on a single-game basis and all season as well.

The one problem with that is: maybe that team beat the shit out of BYU and spent the second half running the ball to bleed clock.  In other words, you still need to bring number of attempts into the equation.  This is why it's so maddening when announcers lazily focus on per-game numbers.  You didn't "hold" them to 200 yards if they passed the ball ten times.

So, to get a number that reflects the quality of a team's performance in a game, say for the run game, we take an opponent and divide (game yards / average yards) as well as (game attempts / average attempts).  For the former, a number below 1 is good, and for the latter, a number above 1 is good.  So we take (1 - first number) and (second number - 1) and add those two together, then multiply by 100 just for readability.  The result is that any number above 0 is an indicator of a good game - you held the opponent to a worse day on the ground than they normally have.  Any number below 0, and you let them have a better day than normal.  This is great because it works equally well for evaluating how you did against Kansas or Alabama.

I went back to 2008, which is only as far back as the eminently excellent site goes.  (Because I used data from that site, I excluded I-AA teams; they don't appear and anyway it's not that useful of a data point.)  I wanted all of Bronco's tenure, but we can't have everything.  Eight years of data points is pretty good.  I also did the same for UVA.  A positive score in this metric is good, but we don't really know how good because I don't have all 128 teams' worth of data.  But we do have a comparison to UVA, which is a start.

An example of how this works: In 2015, both BYU and UVA played UCLA.  (We don't have to limit ourselves to common opponents, but it's just handy for an example.)  UCLA averaged 177.6 yards on 35.3 carries in the run game, 288.3 yards on 39.2 attempts in the passing game, and 465.9 yards on 74.5 plays overall.  All numbers are per game.

Against BYU they ran the ball 38 times for 296 yards.  About average on the carries, way too many yards - that's a terrible effort for BYU, and a score of -59.0.  Against UVA, 152 yards on 34 attempts for a moderately positive score of 10.7 for UVA's defense.

However, BYU held them to 106 yards on 23 attempts - both well below average - for a score of 21.9.  UVA scored -27.4 in the pass game by allowing 351 yards on 37 attempts.

In total, BYU allowed 402 yards.  Great, it's less than their 465.9 average - but on 61 plays vs. their 74.5 average.  That means a negative score of -4.4 overall.  For one game, that's close enough to zero to be a pretty neutral number.  UVA allowed 503 yards on 71 plays, for -12.7 overall.  Unsurprisingly, not only did UCLA beat both teams, but the UVA game was 34-16 and the BYU game was close at 24-23.

Got all that?  Let's present the numbers.  A couple notes: "Total" means the total for the season when you add everything up and treat the season as one game; "average" means each game averaged, so that each game weights the number the same.  Positive and negative games should be self-evident.

Run Defense

Pass defense

Total defense

Takeaways from this:

-- UVA's defense hasn't been bad, mostly.  Most years, UVA has positive scores.  More positive games than negative.

-- But it hasn't been exceptional.  UVA racks up 46 positive games and 42 negative ones, in terms of total defense.  The positive numbers are mostly small ones.

-- There's not a lot of separation between defensive coordinators.  Al Groh was in charge the first two years of this analysis.  Jim Reid was 2010-2012, and Jon Tenuta 2013-2015.  Reid had one really awful year in 2010 and Tenuta one really good one in 2014, and other than that the two DCs are hard to distinguish.  Tenuta was an upgrade over Reid, I think that's fair to conclude, but his 2015 pass defense stunk.  Not as bad as Reid's 2010 run defense.  We don't have a lot of reference points, but the ones we do have say 2010 was just unbelievable in its suckitude.

-- Even so, BYU has been much better.  2008 and 2014 are the only years where UVA had a better defense.  And even then, BYU's pass defense was better than ours in 2014, because it's not like BYU had a bad defense.  Consistently, they've been the better team on defense without a doubt.  They rack up 63 positive games against 35 negative ones - fewer negative games than UVA despite playing 10 more games that counted in this analysis.  (They went to a hell of a lot more bowl games and don't make as much of a habit of playing FCS teams.)

-- BYU tends to win its best defensive performances; UVA tends to waste them.  As you'll see below, the single best defensive game UVA has played in the past eight years was also notorious for some of Mike London's absolute stupidest decisions.

Ergo I think we can safely conclude what we already figured: Bronco Mendenhall is a definite upgrade.  We'll see about offense later - and it should be less time now that I have the technique refined a bit.

Just for fun (and some reference points), here are the five worst and best performances by each team:

Run defense (worst):


-71.1 (2011 vs. Utah - L, 54-10)
-67.4 (2011 vs. Idaho - W, 42-7)**
-59.0 (2015 vs. UCLA - L, 24-23)
-58.5 (2009 vs. Florida State - L, 54-28)
-45.7 (2008 vs. San Diego State - W, 41-12)


-71.0 (2010 vs. Eastern Michigan - W, 48-21)
-68.1 (2010 vs. Duke - L, 55-48)
-67.7 (2011 vs. Florida State - W, 14-13)
-61.5 (2012 vs. Georgia Tech - L, 56-20)
-57.6 (2015 vs. Louisville - L, 38-31)

Pass defense (worst):


-71.6 (2014 vs. Utah State - L, 35-20)
-59.3 (2010 vs. Utah State - L, 31-16)
-57.3 (2015 vs. Missouri - L, 20-16)
-49.9 (2014 vs. Boise State - L, 55-30)
-37.0 (2013 vs. Houston - W, 47-46)


-58.1 (2010 vs. North Carolina - L, 44-10)
-41.0 (2010 vs. Maryland - L, 42-23)
-36.4 (2011 vs. Miami - W, 28-21)
-29.5 (2008 vs. Connecticut - L, 45-10)
-29.0 (2013 vs. Georgia Tech - L, 35-25)

Total defense (worst):


-43.4 (2011 vs. Texas - L, 17-16)
-25.6 (2008 vs. Washington - W, 28-27)
-25.2 (2015 vs. Missouri - L, 20-16)
-24.5 (2014 vs. Boise State - L, 55-30)
-23.7 (2015 vs. East Carolina - W, 45-38)


-48.0 (2012 vs. Georgia Tech - L, 56-20)
-42.2 (2008 vs. Connecticut - L, 45-10)
-35.7 (2010 vs. North Carolina - L, 44-10)
-31.4 (2013 vs. Georgia Tech - L, 35-25)
-29.2 (2011 vs. Auburn - L, 43-24)

Run defense (best):


93.4 (2012 vs. Washington State - W, 30-6)
59.0 (2010 vs. UNLV - W, 55-7)
57.5 (2012 vs. Utah - L, 24-21)
53.6 (2012 vs. New Mexico State - W, 50-14)
52.1 (2013 vs. Houston - W, 47-46)


94.3 (2013 vs. Pittsburgh - L, 14-3)
89.5 (2014 vs. Kent State - W, 45-13)
75.7 (2012 vs. Maryland - L, 27-20)
67.4 (2008 vs. Clemson - L, 13-3)
59.2 (2011 vs. Duke - W, 31-21)

Pass defense (best):


86.9 (2009 vs. Air Force - W, 38-21)
69.5 (2008 vs. Air Force - W, 38-24)
59.6 (2013 vs. Middle Tennessee - W, 37-10)
58.4 (2013 vs. Utah State - W, 31-14)
57.2 (2014 vs. Middle Tennessee - W, 27-7)


57.2 (2012 vs. Virginia Tech - L, 17-14)
50.1 (2014 vs. Louisville - W, 23-21)
50.0 (2009 vs. North Carolina - W, 16-3)
45.6 (2011 vs. Georgia Tech - W, 24-21)
41.5 (2012 vs. NC State - W, 33-6)

Total defense (best):


52.8 (2013 vs. Middle Tennessee - W, 37-10)
42.0 (2014 vs. Middle Tennessee - W, 27-7)
40.4 (2011 vs. New Mexico State - W, 42-7)
39.9 (2012 vs. Utah State - W, 6-3)
39.5 (2010 vs. UNLV - W, 55-7)


50.4 (2012 vs. Virginia Tech - L, 17-14)***
48.1 (2008 vs. Clemson - L, 13-3)
45.6 (2013 vs. Pittsburgh - L, 14-3)
36.2 (2012 vs. NC State - W, 33-6)
35.5 (2013 vs. BYU - W, 19-16)

**Sometimes these scores ain't the best for predicting actual outcomes.  UVA played Idaho that same year and, you'll recall, escaped by the skin of their teeth, 21-20.  This despite having all positive numbers in this scoring metric for that game.  BYU did play excellent pass defense that game, and we haven't looked at the offense yet, but the real explanation, of course, is the special teams and turnovers - you might recall that game as being a particularly nasty example.

*** This was the game where Mike London wisely chose to use his two timeouts to freeze the wheelin' dealin, weed-stealin' Cody Journell instead of to save time for his offense, which was smart because it deflected the criticism from his idiotic decision to try and drive 90 yards for the game-winning score against a howling wind instead of playing for overtime where both teams would've had the same wind conditions.  I mean, it didn't work in that we lost the game, but nobody remembers Mike Rocco trying to throw a 5-yard out pattern across the field which of course got picked off because hurricane.  Brilliant decision-making from start to finish.  One imagines Bronco Mendenhall has more coaching acumen than to do any of that stuff.


Stuart said...

When we first hired Bronco, I was looking for a way to compare his teams to UVA's while accounting for strength of schedule. I tried looking at Strength of Schedule weighted rankings (like those provided by Football Outsiders or ESPN's FPI ratings), but I couldn't find a readable way to compare apples to apples. This seems like a reasonable and fairly balanced metric especially when looking at season long performance.


Anonymous said...

I want to comment about basketball, but I won't here because this football analysis was interesting, unexpected, seems as viable as any analysis I've on fivethirtyeight or kenpom or whatever, and for all those reasons merits its own comments.

But man those basketballs.