Moving The (Stati)Sticks: Utah

Moving The (Stati)Sticks: Utah

Submitted by Adam Schnepp on September 11th, 2015 at 1:53 PM



Brian recently wrote a great Picture Pages that wrapped the “football is a game of inches” trope in a box and tied a nice, neat bow around it. Reading it is a visceral experience, a reminder of the miniscule events that can swing a game. If that’s looking at football through a microscope, what happens when we climb up to the photo deck and pull out a wide-angle lens?

We’re lucky enough to have people like Bill Connelly and MGoBlog’s own The Mathlete wondering the same thing and working toward achieving focus with a zoomed-out view of the game. Connelly wrote an influential piece about the five factors that he found to influence the outcome of a game, while The Mathlete’s research has come up with four. The purpose of this new weekly post is to pick through those factors and find what influenced last weekend’s outcome, as well as whether it was expected (turnovers are not great, Bob) or unexpected (I’d give you an example but then it’d fall into the “obvious” category, wouldn’t it).

I can hear my boss down the hall give me a quick summary of what I need to know about advanced stats before I have to switch to a different tab.

That’s oddly specific. It’s almost like the author of this piece has some prior experience with a scenario very similar to this one in a prior job. Anyway, Bill Connelly’s five factors are:

  • Explosiveness: If you’re averaging more yards per play than your opponent you’re in good shape. Win probability swings wildly here; averaging just 0.1 yards per play more than your opponent raises win probability from 50% to 55%. Connelly also has a more advanced Equivalent Points Per Play metric that also accounts for the equivalent point value of the yard line from which a play is run.
  • Efficiency: Last season the coaches often said that they “fell behind the sticks.” In effect, they were saying that the offense wasn’t very efficient. Success Rate is basically a measure of how well your offense stays on track; to be counted as a success a first-down play needs to get 50% of necessary yardage, a second-down play needs 70%, and third- and fourth-down plays need 100%.
  • Field Position: Turns out starting closer to the endzone than your opponent is kind of a big deal. I’ll let Connelly explain it numerically.
  • Finishing Drives: Connelly looks at how successful a team was inside the opponents’ 40-yard line. His argument is that there isn’t that big a difference in how teams perform inside the 20, but you do see some difference if you expand the range you’re looking at by 20 yards. It’s a pretty intuitive thing to look at; if you’re getting into opponent territory and not coming away with points you’re probably not going to win.
  • Turnovers: Just don’t do them. Giving up the ball cuts off your opportunity to score while handing the opponent an extra one. We’ve already talked about field position, and with that being a component of a turnover you can understand how costly they can be from the standpoint of win expectancy.
    The Mathlete’s four factors are similar in principle to Connelly’s, but are calculated differently. In his words, they are:

Conversion rate = [1st Downs gained]/[1st Down plays (including first play of drive)]. A three and out is 0/1. A one play touchdown is 1/1. Two first downs and then a stop is 2/3, etc.

Bonus Yards = [Yards gained beyond the first down line]/[Total plays from scrimmage]

Field Position = The expected point difference per game for where a team’s offense starts and where a team’s defense starts. Each drive is given an expected value based on the start of scrimmage, all of the drives for the offense and defense are totaled and compared. This accounts for all elements of field position: turnovers, special teams, drive penetration etc.

The fourth one is points per trip inside the red zone, which is self explanatory.

I’m back. Can you start writing about numbers so it looks like I’m doing something important and work related?

I mean, I guess? I’d recommend opening a sheet in Excel but that’s just me. Let’s look at The Mathlete’s factors first.

Team Field Pos Rank Conv Rate Rank Bonus YPP Rank Red Zone Rank
Michigan 21.0 60 73 30 1.52 59 5.6 27
Utah 27.9 28 73 30 1.64 55 5.6 27

That’s about as close as it gets with the exception of field position. Mathlete has Michigan’s average start at their 21 while Bill Connelly’s advanced stats box score has Michigan’s average start at their 30.9. I did some charting myself and also came up with 30.9, so the 21 is likely just a typo. The national average starting position is 29.6. With both schools a little over a yard away from that mark, field position turns into another category that’s essentially a wash.

Switching from where they started to where they finished, Michigan and Utah both made six trips inside the 40 and 3 inside the red zone, and both came away with 17 points from those trips. As Mathlete has noted, that results in both teams averaging a fairly good 5.6 points per red zone trip.

From this we can conclude that field position data doesn’t do a good job of explaining what happened. Looking at efficiency doesn’t unmuddy the waters much either. Bill Connelly uses a stat called Leverage Rate as a part of efficiency. Leverage Rate basically tells you what percentage of plays the offense was on track (i.e. all first downs, second and seven or less, and third or fourth and four or less). Michigan’s leverage rate was 70.8%, while Utah’s was 72.9%. Both topped the national average of 68.3%.

Digging through more of Connelly’s advanced box score tells us what the eye test already has: the run game was a little below average, while the passing game was actually really good if you exclude the three interceptions.

I’m hiding in the bathroom please just get to the point.

The first place there’s any real separation is passing downs success rate, where Michigan was successful on 33.3% of throws while Utah was successful on only 10.5%. The separation in passing games is echoed by Michigan’s 25.8 to 19.1 advantage in equivalent points from passing (I highly recommend checking out the definition of equivalent points here).

That doesn’t mean that Utah’s passing game was a total flop; their IsoPPP, which is basically a measure of explosiveness, was 3.37 while Michigan’s was just 1.95. (The national average is 1.84). As mentioned above, I did some rudimentary charting where my criteria for explosiveness was a 10+ yard run or a 20+ yard pass. I had Utah with two such passes in the game while Michigan had three. I’m not sure I’m using the best explosiveness proxy, as Utah did have a number of throws in the range of 15 yards that didn’t meet my criteria for explosiveness but certainly could have had an effect on IsoPPP.

Hey thanks I had to go back to my desk and you still haven’t given me a good explanation of what happened.

If I had to pin it on one thing it’d be turnovers. Utah was +2 in turnover margin, while their turnover points margin was +13.8. In a game where all other factors were relatively close this is massive. In fact, Connelly’s projected scoring margin had Michigan losing by 13.0, but they managed to lose by only seven.

That, along with a good statistical performance from the passing game, was counterintuitively positive. The running game, however, was a fairly clear negative. The offensive line’s struggles resulted in 2.38 line yards per carry, well below the 2.82 LY/carry national average. Even Utah’s line generated 2.74 LY/carry despite the outward perception that Devontae Booker was bottled up. Michigan’s rushing success rate was 31.0%, which isn’t even in the neighborhood of the 42.9% national average. Not having an explosive run game is tenable if your success rate is high, which Michigan’s wasn’t. Improvement there and some regression to the mean in the turnover department should lead to success for Michigan.