Correlations of NFL statistics to team regular season wins

Yes, here at TwoMinuteWarning we are basically NFL number crunchers. We have created a slew of ratings and statistics designed to better reflect and measure team abilities and capabilities.

One of the best tests for any stat in terms of its predictive value is how effective it is in projecting the winner against the point spread. This is the basis for a large part of our research. On the other hand, a more detailed and arguably precise view of a number's worth can be demonstrated by looking at the correlation of the number to various team performance stats.

With correlation you get a result from -1 to +1 in range. A +1 correlation means that the two areas you are comparing are perfectly related -- a better figure in one stat will show a corresponding gain in the other stat. A perfect correlation is not to be expected, except for an obvious test (i.e. a team's score versus its opponents is directly related to the outcome of the game). All of the stats we are looking at initially would be expected to have some level of positive correlation.

There are three main classes of data we examine below:

• Drive Chart stats: a look at how NFL teams perform on a 'drive by drive' basis
• Play-by-Play ratings: a breakdown of team performance at the play-by-play level of detail
• Down-by-Down stats: primarily the PBP numbers taken in context of the specific down

Let's go to the numbers at hand: this reflects final regular season statistics from 1998 to 2002, with the stat in question being correlated with team wins during the sixteen game campaign. The higher the correlation value, the more pronounced that stat's effect on a team's regular season wins.

Comments: Yes, that's lot of factors, sixty in all to consider. We have not yet got to the nets (eg an offensive number minus the defensive number) where the correlation will in many cases be higher, but that will be the subject for a follow-on piece.

At the head of the class is a team's basic offensive efficiency on a drive level -- the average points they score for each time they get the ball. It's no shock this would come out near the top, although some people would debate whether offense or defense would be the king, but the scorers edged the defenders in this sample.

The #2 statistics is how often a team forces its opponents to pass on first down, which again isn't a suprise in that when teams are behind they are much more likely to be passing out the gate. Of course, what would be interesting would be to retest these numhers, and this stat in particular, looking at say only the first half of games. Maybe then the passing on first down number would change in importance.

The #4, #7 and #8 stats are all passing play-by-play data numbers. This is significant and suggests that teams which are not efficient in passing the ball on offense are at a huge disadvantage. Indeed, moreover, since the passing 'Play Success Rate' comes in so high, it's also implying what the 'West Coast offense' advocates have been saying for years: a short, high percentage passing game is stronger than a deep-throwing, occasional big gain one.

There is of course a fair amount of dependency in some of these stats. For instance, the offensive average time of possession comes in as highly correlated with a team's W-L record. This though may not be so much a factor of the time of possession in a pure sense, but more a statement that teams with leads run the ball and look to run the clock, so consequently their average time per drive will be longer than a team behind that is looking to move the ball in a hurry. In other words, there's a little of "a team that wins, wins" at work here.

Nonetheless, the numbers tell a story and a look at both the nets, and the first half alone data seems worthy of our attention. For those looking to examine the numbers by category of statistics, the table that follows is easier to work with.