Does running the football cause the Notre Dame Fighting Irish to win? Or does the Irish leading cause them to run the football?
My colleagues note that the Irish were 20-0 under Brian Kelly when 60 percent of their plays during a game were runs.
I remained skeptical that running causes the winning, so I asked Kevin Rudy to weigh in. Rudy is a Penn State graduate who works for Minitab, a statistical software company. He’s been applying his expertise to sports for more than 15 years. In other words, Rudy is not the sort of fellow you’d invite to your fantasy football league — unless you enjoy losing.
I supplied the expert with a lot of raw data: the number of runs and passes executed during Brian Kelly’s tenure based on the score of the game. It’s right here.
He took over from there. Forewarned: There is a lot of #SportsMath here.
THE 60 PERCENT THRESHOLD
First, he plotted the running percentage of each individual game in the Kelly era and divided it between wins and losses.
Rudy offered a reference line at 60 percent, because that was considered by my colleagues to be a significant threshold. (Incidentally, there are three losses where the percentage of run plays was near - but not above - 60 percent: the 2015 Stanford game and last year’s losses to Texas and North Carolina State.)
Rudy noted that about a third of Notre Dame’s wins come when they run the ball more than 60 percent of the time. Even when they don’t hit that threshold, they still win more (39 games) than they lose (31 games).
HOW SCORE FACTORS IN THE DECISION MAKING
Rudy then produced a chart that shows how often Notre Dame ran the ball depending on the score.
This conclusion won’t surprise anyone who watches football: The Irish tend to pass more when they are trailing and run more when they’re leading. As Rudy noted, Notre Dame doesn’t hit a run/pass mix of 60 percent until they lead by more than two scores.
Remember, we’re looking at three specific types of games:
- Notre Dame loses
- Notre Dame wins, but less than 60 percent of their plays are runs
- Notre Dame wins and more than 60 percent of their plays are runs
In the graph above, the blue bars are games Notre Dame lost. In those specific games, the Irish offense ran 63 percent of its total plays when trailing in the contest.
Now move your eyes to red bars, which represents Irish plays during games where their running percentage was more than 60 percent. In those specific games - all games the Irish won, remember - Notre Dame ran just 12 percent of their total plays when trailing and ZERO plays when they trailed by more than one score in any of those games.
Bring your eyes back to the blue bars momentarily. Again, these are Notre Dame losses. There’s not a single play they ran when leading by 17 or more points in their 31 losses.
Back to the red bars (60+ percent run games). The Irish ran 23 percent of their plays with a lead of 17 points or more. Now check out the yellow bars. These are games in which the Irish won, but ran the ball less than 60 percent of the time. In this instance, 18 percent of their plays were run with a lead of 17 points or greater.
“It’s easy to see why the Irish have such a good record when their high run/ratio is above 60 percent,” wrote Rudy. “It’s because they get to run a big chunk of their plays with a large lead.”
ONE LAST ARGUMENT
If you’ve followed the data this far - and you’re still not convinced - you may have one final rebuttal: Notre Dame, you argue, ran the ball when the game was close - and that’s why they got to the point where they could be run-heavy with a two score or better lead.
Rudy looked at that too, by specifically focusing on instances where the game was within one score. He ran a “binary logistic regression,” which is a very complicated way of saying: “What’s the relationship between the run/pass ratio when the score is close and the outcome of the game?”
Notre Dame loses games with a wide range of run/pass ratios when the score was close. On one end of the spectrum, you have the BCS National Championship game - which consisted of Alabama’s first two scores sandwiching three unproductive Irish pass plays. On the other end, you have a game where 77 percent of the Irish’s plays were runs when the game was within one score - and they lost.
“Only 3.9 percent of the probability of whether Notre Dame wins is explained with how often they run the ball when the score is close,” Rudy explained. “Another way to think of it is that 96 percent of the reason Notre Dame wins or loses is explained by something else.”
GLIMMER OF HOPE FOR #RTDB ENTHUSIASTS
Rudy said the probability of Notre Dame winning the ballgame does increase as the run/pass ratio increases, which seems like great news for those who just want Kelly to run the dang ball.
Don’t get too excited, though.
In losses, Notre Dame ran the ball 47 percent of the time when the score was close. In wins, it was 53 percent.
“With a run/pass ratio of 0.47, Notre Dame has a probability of winning of 63 percent,” said Rudy. “With a run/pass ratio of 0.53, Notre Dame has a probability of winning of 68 percent. In other words, how much Notre Dame runs and passes when the score is close has very little effect on whether they win the game.”
Thank you to Kevin Rudy for his time and expertise. Follow him on Twitter and read his writings over at MiniTab. For more on this topic - including binary logic regression - read Rudy’s “Correlation is Not Causation: Why Running the Football Doesn’t Cause You to Win Games in the NFL.”