What are the keys to winning football games? I have done a statistical analysis of the 2009 season to determine which metrics (yards, sacks, etc) have the most effect on winning percentage. The answers may surprise and dismay long-time Steelers fans. (For Johnny and other math guys, I applied a linear regression between each metric and the winning pct. for all 32 teams, then calculated the correlation coefficient R, which I have expressed as a percentage for simplicity.) I examined more than 50 different metrics and summarized the highlights.

Let's start with a truism: the team that scores the most points wins. Scoring points has a 88% correlation with winning. Gaining yards is good, but only if it leads to scoring. I've also included net points for the offense, which subtracts return TDs; this metric is slightly better than points scored.

Now here's something interesting: running the football successfully has zero (0%) correlation to winning. None. Passing, on the other hand, has a very high correlation, particularly passing efficiency (yards per pass attempt). Not giving up sacks is important, as they reduce passing efficiency. Limiting interceptions is even more important. Fumbles appear to be random events that affect bad teams only slightly more than teams.

On defense, once again, points are more important than yards. (Net points again excludes return TDs and adds TDs that the defense scores.) The key thing here, however, is that keeping the other team from scoring is less important that scoring. In other words, a good offense is more valuable than a good defense. Stopping the pass is more important than stopping the run, but run defense is still useful. I've included CHFF's vaunted Defensive Hog Index here, but statistically, it has a modest correlation to winning, not much better than sacks or interceptions. Suprisingly, third-down defense and red-zone defense have a low correlation to winning.

This data shows why most coaches pay little attention to special teams. Perhaps because special teams are not on the field very much, there is little correlation between special-teams play and winning. In fact, bad teams have better punting statistics than good teams, because they punt often and typically punt into a longer field. Even field-goal accuracy has little correlation with winning.

A good offense can lose games because of a bad defense, and vice versa. Therefore, statistics that combine offensive and defense performance have a higher correlation to winning. Point differential has the highest correlation (93%) of any metric I analyzed. Because of the importance of the passing game, passer rating differential also has a strong correlation, as does the often-discussed turnover ratio. Curiously, good teams and bad teams seem to have a similar number of penalties.

In case the above analysis seems like statistical mumbo-jumbo, let's take a look at the performance of the top teams in the NFL, those that recorded 10 or more wins in 2009. This table shows where these teams rank in the NFL (1-32) on several important metrics.

As you can see, all of the top 7 offenses in net points scored appear on this list, but only 4 of the top 7 defenses. On yards per pass, 7 of the top 8 teams made this list, but on yards per rush, two of the worst teams in the league won 13 or more games. For stopping the pass and stopping the run, the record is mixed, but the top teams rank a bit better in stopping the pass.

In summary, the 2009 data show:

• Passing is more important to winning than running.
• Offense is more important than defense.
• The critical measure of an offense or defense is points (preferably net points), not yards.
• Interceptions and sacks are important.
• Fumbles and penalties are random events that are statistically unimportant.
• Special-teams play is largely irrelevant to winning.

In Part Two of this post, I will use the above metrics to analyze the Steelers 2009 performance and determine what improvements will most likely result in an improved winning percentage.