# More from the New York Times

Two weeks ago Martin Schmidt penned The Keeping Score column for The New York Times. This was Marty’s second Keeping Score column. In today’s New York Times you can see my first. The piece, entitled “To Get a Grip on Turnovers, Follow the Bouncing Ball”, expands upon a point made in The Wages of Wins. Specifically, turnovers in the NFL are very hard to predict.

Malcolm Gladwell often utilizes his blog to expand upon what he writes for The New Yorker. I thought I would follow Gladwell’s lead and expand upon my New York Times column in this forum.

In the column I began by noting that turnovers are quite important. This is certainly part of the conventional wisdom in football, and the data supports this view. Specifically, in 2005 the teams in the top ten in turnover differential all won at least ten games. Of the teams in the bottom ten in turnovers, none managed to achieve double digits in victories.

Looking at a larger sample, we see that 68 teams since 1995 finished the season with a turnover differential that was +10 or higher. These teams averaged nearly 11 wins. In contrast, 54 teams had a turnover differential of -10 or lower. These teams only won, on average, six contests.

Coaches, the media, and fans know that turnovers are important. What may not be fully understood, though, is that turnovers are very unpredictable. Again, consider teams that finished with a +10 turnover differential from 1995 to 2004. Of these teams, 24% were able to repeat a +10 differential or better the next season. Another 35%, though, posted a negative turnover differential the next season.

Okay, as we do in The Wages of Wins, let’s turn to regression analysis. If one regresses a team’s turnover differential this season on the team’s differential last season one can see clearly the inconsistency in turnovers. Specifically, the simple regression indicates that less than 2% of current turnover differential is explained by what a team did last season. The correlation coefficient between current and past turnover differential is less than 0.15.

The same story is true at the player level. For both quarterbacks and running backs the rate of turnovers per play is not very correlated with what the same player did last season.

Here is the punch-line: Turnovers impact outcomes. Teams will rise and fall in the standings this year because of interceptions and fumbles. But who will throw these interceptions and who will fumble the ball is hard to know at the onset of the season.

And this is one reason football is such a great sport. We will be surprised by the 2006 season. A team we expect to win will lose more than we thought. A team we expect to lose will win more than we would have predicted. All we know for certain is that there is uncertainty.

Except, of course, for the Detroit Lions. As a fan of this team, I am pretty certain I will end the season disappointed. Of course for fans of Detroit the really important part of the year is the April Draft. Already we wonder who will be available when we are on the clock in 2007.

- DJ