In this week's Varsity Numbers, Bill Connelly revisits some measures and concepts: Adjusted Scores, Covariance, and momentum (or whatever you choose to call it).
17 Oct 2004
Guest Column by Jim Armstrong
In part one of this article, I defined pace as time of possession per play in the first half of games to minimize the effect of teams adjusting their pace in the second half based on the score. We looked at the slowest and fastest offenses in 2003 and mentioned a few teams that have been consistently slow or fast paced over the past several seasons. But we didn't see any obvious relationship of pace to winning. Here we will examine whether employing a slow or fast pace in some game matchups can be a good strategy.
Time of possession has become a major statistic reported in most NFL box scores. Commentators often talk about a team's ability to "control the clock" as being a key to victory. Others advocate a "ball control" offense against a stronger team in order to keep the opposing team's offense off the field. Of course, barring turnovers on punts and kickoffs, football rules require teams to alternate possessions, so generally a team can't give itself more chances to score than its opponent gets. But by slowing the pace a team can reduce the total number of possessions by both teams. As the theory goes, creating a shorter game gives the weaker team a better chance of staying with the stronger team, which would certainly win in a game with an infinite number of possessions.
There have been a couple of works that have examined the strategic aspects of pace in a theoretical manner. Harold and Daniel Sackrowitz, writing in the statistical journal Chance, concluded that because the probability of winning is so sensitive to changes in scoring efficiency, it is misguided to employ a ball control offense. They argue that a team attempting to deviate from its optimal pace will likely reduce its ability to score too much to make the strategy worthwhile. On the other hand, William Krasker's work on footballcommentary.com suggests that not only should weaker teams begin the game in a slowdown, but the stronger team should begin in a hurry-up offense in order to "lengthen" the game. Krasker does acknowledge the work of Sackrowitz and Sackrowitz, and is careful to advocate only pace changes by controlling when the ball is snapped between plays and not those affected by playcalling. As mentioned in Part one, the available data does not allow us to distinguish among all methods of controlling the pace. However, we can still take an empirical look at how pace affects the outcome involving games of unevenly matched teams. If the results show that weaker teams improve its chances of winning by slowing the pace (and vice versa for stronger teams), Krasker's analysis will be confirmed. If not, it will suggest either that such strategies are indeed misguided or that teams haven't been able to figure out how to use these strategies without disturbing their optimal efficiency.
To determine these effects of pace, I obtained point spreads for all the regular season NFL games from 1998-2003, the same six years used in part one. I wasn't too concerned about the exact numbers; instead, I grouped the games into those of evenly matched teams with lines of 3 points or less, those of moderately mismatched teams with lines between 3.5 and 6.5 points, and those of heavily mismatched teams with lines of 7 points or greater. This divided the games into three groups having roughly the same number of games each. Then I labeled each team's first half pace for each game as being Fast, Medium, or Slow such that each of these pace groupings contained exactly one third of the values. Finally, I computed the winning percentage of each combination of these groupings and created a chart showing the results graphically. That chart is shown below, using the convention of negative point spreads to represent the stronger teams and positive point spreads to represent the weaker teams.
Here we see that a fast pace decreases a team's chances of winning pretty much across the entire spectrum of matchups. To understand why, let's ignore the matchups for the moment and look how our three pace groups fare in moving the ball, scoring points, avoiding turnovers, and winning. Recall from part one that there was no obvious or linear relationship of pace to winning, but the breakdown shows that a medium pace gives a team the best chance at winning, describing what is known as a nonlinear relationship. As I did in the chart, the pace breakdowns in the table below are based on the average pace an offense maintains over the first half of a game, not a breakdown of individual drives by pace.
In general, a fast pace seems to be the least effective, but perhaps a fast pace can show up as a result rather than a cause. For example, drives ending in a turnover in the first few plays will look fast simply because there weren't many opportunities in between plays for a team to slow its pace. Shorter drives (in terms of numbers of plays) are more likely to have been ended by a turnover than longer drives. In fact, one could argue that it's difficult to make any conclusions about a team's intended pace after most drives of few plays. To account for this, I repeated the process above after discarding all drives of fewer than five plays, which cut the sample size roughly in half. The resulting chart and table are shown below.
Obviously, these drives are going to be more successful, but the conclusions are similar. The gaps involving turnovers have shrunk so that the relationship with pace is no longer significant. The charts show no evidence that weaker teams have improved their chances of winning by running a slowdown ball control offense, nor do they indicate that stronger teams have been helped by running a hurry-up offense. A slow paced offense will generally only reduce the total number of possessions by one or two per team per game, probably not enough to make much of a difference against a much stronger team. Rather, the numbers suggest that overall, a medium paced offense gives a team its best chance of winning the game. It is somewhat curious that a slow paced offense doesn't doesn't do as well in terms of winning, given how favorably it performs in the drive efficiency numbers. That may be an interesting topic for further study.
Of course, those are just general conclusions. We've already seen in part one teams that have had recent success running slow paced and fast paced offenses, namely Tennessee and Philadelphia, respectively. Perhaps just as important is for a team to stick consistently with whatever pace seems to work best since, as Sackrowitz and Sackrowitz show, even small deviations from a team's optimal pace can lead to reduced efficiency. To test this theory, I measured the variance of pace of each teams' drives for the six seasons and then divided those team-seasons in half, those with high variance and those with low variance. I then looked at how many wins each group averaged.
|High pace variance||7.48|
|Low pace variance||8.52|
This shows that teams can expect an average of one additional win per season just by maintaining a relatively consistent pace from game to game. And it also supports the suggestion inferred from the charts that trying to adjust an offense's pace substantially based on the game matchup is not a particularly promising strategy.
So in summary, what can conclusions can we make from all this?
Teams can win with fast paced or slow paced offenses, but there are only a few teams that tend to maintain a very slow or very fast pace year after year.
It may be counterproductive for a weaker team to employ a slowdown offense or for a stronger team to employ a hurry-up offense right off the bat. Rather, recent evidence shows that in general, a medium paced offense yields the best chance of winning, regardless of the relative strengths of the two teams. An alternate theory is that for the most part, teams just haven't figured out how to adjust their pace efficiently to the matchup.
So rather than trying to adjust the pace for the matchup, it appears to be better for a team to maintain a relatively consistent first half pace throughout the season. This supports the assumption that teams have an optimal pace at which they operate most proficiently. They should figure out what that pace is and stay with it, at least until game situations dictate a change.
Jim Armstrong wrote about consistency in turnovers back in April, and is now contributing his drive stats to our site each week. If you are interested in writing a guest column, something that takes a new angle on the NFL, please email us your idea at firstname.lastname@example.org.
4 comments, Last at 27 Mar 2007, 12:47am by online new car buying