Last season when I interviewed Tulane head coach Willie Fritz, he told me that he didn’t believe in luck. I brought up the concept of turnover luck and he was happy to offer his insights, citing that in his 25 years of coaching his team has always had a positive turnover margin. Fritz doesn’t see anything on the field as random and my off-season analysis on turnovers shows that turnover margin is predictable in college football.
Turnovers have a huge influence on game outcomes and therefore the predictability of turnover margin is of great interest when attempting to forecast the upcoming season. I collected turnover data from the previous 14 seasons. Looking at the data some things look repeatable while others do not.
Best Average TO Margin: Alabama +8.4 per season
Worst Average TO Margin: New Mexico State -7.5 per season
Who’s better, Alabama or New Mexico State? So we have to ask ourselves, is it coincidence that Alabama has the best historical turnover margin while New Mexico State has the worst? Boise State is close behind Alabama for the best turnover margin while Idaho is second in the cellar. These results do not at all seem random.
Best TO Margins: 2013 Houston +25, 2017 Wyoming +24
Worst TO Margins: 2017 SJSU -26, 2008 WSU -25
Now let’s look at the extremes, as last year featured two of them. Wyoming was a whopping +24 in their 8-5 season with a bowl win, but San Jose State took the cake in 2017 posting a -26 turnover margin in their 2-11 campaign. Extremes of 20+ or -20 are rare and represent less than two percent of the data. In fact, no team has ever posted back-to-back seasons of 20 plus. This provides some evidence that an element of turnovers is random.
To formalize our analysis we can run some statistical tests. A correlation coefficient measures the strength of a relationship and a value of one indicates a perfect positive relationship. If we compare a team’s turnover margin this year to the one from the previous year, the value is .17, indicating that last year’s turnover margin says something about this year’s turnover margin. We can get a little fancier and doing so shows us that looking at the previous two seasons gives us the best indication of a team’s turnover margin this season. I’ve provided a simple formula to approximate a team’s predicted turnover margin for the upcoming season
TO Margin = (TO Margin Last Year + TO Margin 2 years ago) / 6
For example, Boston College has a TO margin of +5 in 2017 and +7 in 2016. So our forecast for Boston College would be +2. The model results indicate that teams who historically win the turnover battle will continue to win the turnover battle. However, teams do tend to trend towards average and extreme results are often not repeated.
Follow Jordan on Twitter @JordanSundheim