'Tis the season to make predictions.
With the NFL regular season about to begin, pundits are producing an abundance of predictions. Perhaps the boldest forecast came from Tiki Barber, who predicted that the Detroit Lions -- a team that has won only two games across the past two seasons -- will win nine games in 2010.
Barber's prediction was met with some scorn. But for fans of the Lions -- and since I was born in Detroit I have spent a lifetime in this group -- Barber's bold declaration might lead to some optimism for 2010.
Of course, such optimism depends on Barber's ability to see the future. And on this subject there may be some doubt. It's not that Barber doesn't have a great track record. Actually, I have no idea how well he has forecasted in the past. What I do know is that making accurate predictions in the NFL is difficult.
To illustrate, let's consider a recent comment by Brian Burke at Advanced NFL Statistics. Burke argued that preseason predictions are worthless. How did he reach this conclusion?
He began with the predictions for the 2009 season offered by Football Outsiders, a website devoted to statistics and football. The predictions of Football Outsiders were then compared to a pair of less sophisticated forecasts. The first -- which Burke labeled as the Constant Median Approximation (CoMA) -- assumed each team would win eight games in 2009. As Burke notes, such a forecast is what you would offer if you knew nothing about the NFL.
A second forecast was also considered. This second attempt simply assumed that a team would win six games plus ¼ of a team's victories from the previous season. In other words, if a team won twelve games last year, this second method -- which Burke labeled "Koko" (after the monkey) -- would predict nine victories this year.
So Burke had three models. As Burke notes, the forecast from Football Outsiders was not substantially better than the CoMA approach. And the Koko model -- yes, the one named after the monkey -- actually did the best job of predicting the 2009 season. Burke argues, "This is a notable result because it suggests that Football Outsiders' predictions are so bad, they are literally worse than having no football knowledge at all. It's like negative information, draining us of insight."
Burke emphasizes -- and I would concur -- that it's not entirely fair to pick on Football Outsiders. One suspects you would see similar results from anyone trying forecast the future in football. And this is because there are aspects of football that are very difficult to predict.
For example, turnovers clearly have an impact on outcomes in the NFL. Analysts consistently argue that the team that wins the turnover battle has a good chance of winning games. Turnovers, though, are very hard to predict. For example, as we note in Stumbling on Wins, less than one percent of a quarterback's interceptions per pass attempt in the current season can be explained by a quarterback's interceptions per pass attempt the previous season. In other words, Jay Cutler's sudden propensity to throw interceptions in Chicago was not something the Bears -- or fans of the Bears -- could have predicted prior to the 2009 season. And Cutler's problem with interceptions last year does not tells us much about what will happen in 2010.
Interceptions are not the only aspect of a quarterback's performance that defies prediction. A similar result can be seen with respect to a quarterback's fumbles. And fumbles for running backs are also very hard to forecast.
Beyond turnovers is the issue of player injury. Football is the only game where an athlete is almost guaranteed to suffer an injury at some point in his career. The timing of these injuries, though, can't be predicted. And if a key player on a team goes down, the team's ability to win might change dramatically.
Let's say, though, that a person found a way to forecast turnovers and injuries. Well, there is still another obstacle a forecaster must overcome. Rosters in the NFL are constantly changing. For example, let's consider the Detroit Lions. The Lions only won two games last year. But when we look at this same team in 2010 we see a number of different faces. On defense, only Cliff Avril, DeAndre Levy, Julian Peterson, and Louis Delmas are scheduled to start again for the Lions. In other words, as many as seven of the eleven starters on opening day in 2010 did not start for this team in 2009. When we look at the offense, the team has also made significant additions at running back, wide receiver, tight end, and offensive guard.
Again, the Lions have historically been awful. But once again, many of the players who will wear Honolulu Blue and Silver in 2010 did not wear this uniform in the past. And since the players in the clothes have changed, isn't it possible the outcomes we will observe will also change? In other words, isn't it possible that Tiki Barber is right?
Okay, it may be possible. But if you are a fan of the Lions (did I mention I am in this group?) it still seems hard to believe. We certainly suspect that once the games start, fans of the Lions are once again going to start thinking about the 2011 NFL draft. Before the games start, though, there is nothing wrong with thinking this might be the year the Lions finally win a few games. Once again, football is hard to predict. And the Lions have to suprise us one of these years (really, don't they?)
Update: For more on why the Lions will be good in 2010, please see this post at The Wages of Wins Journal.