So far, we’ve been mostly about team rankings. Based on a statistical physics Ph.D. from Stanford, some guy developed an algorithm that nicely accounts for margin of victory and strength of schedule in ranking teams. And the rankings give predictions that are more accurate than the Vegas line in picking the winner of college football bowl games.
But analytics on the team level doesn’t provide much insight into American tackle football. It has become clear to us that people want a deeper application of analytics in football. How well does a team pass the ball? Run the ball? Who should win the Heisman trophy when the contenders play schedules of such varying strength? This is our first post towards applying our algorithm to answering these questions.
The Power Rank algorithm in Passing and Rushing Analytics
We’ll start with passing and rushing analytics. Traditionally, the media uses yards per game to evaluate both defense and offense in these two categories. However, this is deeply flawed. Some teams throw the ball much more than others. Using passing yards per game to compare Mike Leach’s Air Raid offense which rarely runs the ball with Air Force’s triple option offense that rarely throws the ball is meaningless.
The first step in passing and rushing analytics is dividing total yards by the number of attempts. This normalization is analogous to tempo free basketball statistics in which points, assists and just about everything else are divided by the number of possessions for a team. It allows for a fair comparison between fast break teams like North Carolina and half court teams like Northern Iowa. Dean Oliver pioneered this analysis with his book Basketball on Paper, and Ken Pomeroy uses it in his college basketball analytics.
However, yards per attempt still does not account for strength of schedule in evaluating passing and rushing. We apply our algorithm to adjust yards per attempt for schedule strength. This is particularly important in college football because of the wide range of team strength. Oregon might rush for almost 10 yards a carry against Missouri State. These rushes all count in their raw yards per rush attempt. Here, we adjust this number to account for a lower subdivision defense.
Let’s look at some rather surprising insights from these passing and rushing analytics in college football.
1. How good was the Texas pass defense?
Don’t let all those words per day fool you. Matt Hinton of CBS Sports is a numbers guy. When he served as Dr. Saturday over at Yahoo Sports, he did a very complete analysis showing the predictive power of recruiting rankings in college football. So when he recently wrote an article about how Texas’s defense struggled against top passing offenses in the Big 12, we took notice. He comments how poorly the pass defense performed against Baylor, Oklahoma and Oklahoma State last year.
Well, Baylor certainly ripped apart Texas through the air. Robert Griffin III threw for an astounding 13.7 yards per attempt against Texas, earning Griffin the Heisman trophy. But Baylor tore up everyone, topping our pass offense rankings. Our analytics predicted they would throw for 9.12 yards per attempt against an average bowl subdivision pass defense.
However, the Texas defense actually performed well against Oklahoma and Oklahoma State, holding both below their season average in yards per attempt. Overall, Texas had the 6th best pass defense last year. This is much better than the 13th they rank in raw yards per pass attempt, since they faced some very good pass offenses in the Big 12.
2. How good was Alabama’s passing defense and offense?
It’s not that surprising that Alabama had the best pass defense in the nation last year. They finished first across the board in yards per game, yards per attempt and our adjusted yards per attempt. Our analytics predict they would allow a microscopic 4.02 yards per attempt against an average subdivision pass offense. The secondary had 3 players drafted in the NFL draft, including safety Mark Barron and cornerback Dre Kirkpatrick in the first round.
However, it may be surprising that Alabama had the 10th best passing offense in the country. Quarterback AJ McCarron and crew racked up 7.22 yards per pass attempt, good for 27th in the country. When adjusting for schedule strength, they get upgraded due to 2 strong performances against LSU. Against this 2nd best pass defense, Alabama threw for about 6.2 yards per attempt in both meetings, well above the 4.53 that LSU gave up on average.
Alabama’s strength in both pass offense and defense played a large role in their National Championship run last year. Passing correlates to winning more than running. Remind us to write an article about that.
3. Which back up running back should I draft on my fantasy team?
Utah State really stands out in our rushing offense rankings. They finished 3rd, behind only the up tempo style of Oregon and the offensive line tradition of Wisconsin. Even after adjusting for schedule strength, Utah State’s rushing attack ranked higher than the pro style offenses at Alabama (11) and Stanford (21). Robert Turbin and Michael Smith, the two running backs that powered this attack, were both taken in the NFL draft and could make for interesting additions to anyone’s fantasy team.
Seattle drafted Robert Turbin, which most likely makes him Marshawn Lynch’s back up this season. While Turbin was Utah State’s leading rusher by total yards, Michael Smith actually had a better yards per carry. His 7.63 yards per rush attempt over a significant 114 carries was quite a bit better than the 6.09 of Turbin. Tampa Bay drafted Smith, and according to Athlon, he has LeGarrette Blount and fellow draft pick Doug Martin of Boise State to compete with for carries.
Here’s where analytics makes a rather bold prediction. Boise State ended last season with the 80th best rushing offense. Yes, the run first pro style offense led by quarterback Kellen Moore finished just above the bottom third of teams. They didn’t earn enough yards per attempt against rather poor rush defenses. Of course, rushing offense strongly depends on the offensive line. However, left tackle Nate Potter of Boise State got drafted, while no linemen from Utah State cracked the 7 rounds of the NFL draft. This analysis suggests that Michael Smith is a better running back than Doug Martin.
You probably don’t even need to draft Michael Smith on your fantasy team. Just take him off the scrap heap mid season.
With these passing and rushing rankings for both offense and defense, we have 480 new numbers to sort through from last season. We could go on for days writing about the passing and rushing analytics. But it would probably be better just to post last year’s rankings and let you figure it out for yourself. To see the rankings, click here.
Thanks for reading.