Tom and Mike have climbed to the top of the Hill of Nonsense and unleash another wave of insane wagering on the world.
11 Oct 2004
by Aaron Schatz
(Hey, are you allergic to numbers? Do you have only 90 seconds to read this article before your boss comes over to your desk and asks what you are doing? If so, we've got a convenient summary if you click here.)
Football analysts have always been searching for one magic number that can predict wins and losses. We know which teams score the most points or gain the most yards, but we want a way to explain which teams will win despite underwhelming statistics (and, more importantly to some people, underdog status on the point spreads). One of the more popular efficiency statistics, particularly with gambling touts, is yards per point.
The goal of yards per point is to determine offenses that make the most of their scoring opportunities, and defenses that might allow lots of yardage but improve in the red zone and prevent the other team from scoring (the cliched "bend but don't break" philosophy). It's as simple as an efficiency rating can get. Divide yards gained by points. That's it, you don't even need a calculator. The fewer yards a team needs to get a point, the more efficient the offense. On the other side of the ball, divide yards allowed by points allowed, and the more yards that a team gives up for each point, the more efficient the defense. Sometimes the statistic is given as points per yard instead, but that's the same idea, just reversing the order of what numbers mean good teams or bad teams.
I honestly had not thought much about yards per point since I started analyzing football statistics a year and a half ago. I had run into the term maybe once or twice. I never took any time to look into it. And then, a couple weeks ago, it suddenly showed up twice in the same week. First, when I did an appearance on WEEI Sports Radio in Boston, the caller before my interview expounded upon the virtues of yards per point as the only number anyone needed to know which teams were the best in the NFL. "Well, we're having Aaron Schatz from Football Outsiders on next, so we'll ask him about that," said Dale Arnold. Uh-oh. Suddenly, I had to scramble to figure 2003 yards per point numbers and decide over one commercial break whether I thought this was a good statistic or not.
A week later, I finally posted the 2001 DVOA ratings and commentary on Football Outsiders, and guess what came up again in the discussion thread. It was bound to happen, because the 2001 Patriots are the poster children for those who believe that yards per point explains why teams with mediocre numbers win more games. On offense, the 2001 Patriots rated second behind Seattle. On defense, they came in third behind Chicago and Philadelphia. In the official NFL team standings, based on yards gained and allowed, they ranked 19th on offense and 24th on defense. That first set of rankings sure sounds more like a Super Bowl champion than the second set. Here's what reader Brian Sheppard wrote:
The Pats lead the league that year in two dimensions that might not show up in DVOA, but made a big difference in wins: offensive points per yard, and defensive points per yard. If I recall, the Pats scored a point for about every 11 yards that their offense gained, whereas the defense allowed a point for about every 20 yards gained by the opponent. With such an edge in efficiency, it is hard to lose.
Now I had to think. It's true that for the four years for which I have DVOA statistics (our innovative ratings explained here), the team whose ratings are most inaccurate is probably the 2001 Patriots. But I couldn't think of what yards per point could capture that DVOA did not, except perhaps for luck. Red zone plays are already counted as 20 percent more important in the latest DVOA upgrade, which made "bend but don't break" defenses rank more in line with actual performance. Touchdowns get extra credit, Adam Vinatieri's field goal success was tracked, what could I be missing? This warranted further investigation.
Since four games aren't really enough to tell us about this season, let's look at last season to get a better idea of how teams do in yards per point. Here are the leaders from 2003 in offensive yards per point (fewer yards being better):
OK, Kansas City being the most efficient offense in the league, that makes a lot of sense. St. Louis being second makes sense also, as long as you are willing to forget about turnovers. And then, third, there is the offensive juggernaut known as the Baltimore Ravens. I mean offensive juggernaut in the same way that I mean "William F. Buckley enjoys the refined comedy of Crank Yankers."
How about the defenses? These are the leaders from 2003 in defensive yards per point (more yards being better):
Alright then, the Patriots had the most efficient defense in the league. Makes total sense. Philadelphia is second. DVOA says that the Eagles did not actually have a very good defense last year (they rank 23rd), but the conventional wisdom says that they were still primarily a defensive team, so we'll allow it. Miami third, definitely. (They are first if you only count intersquad practices.) But proclaiming that the New York Jets and Kansas City Chiefs had two of the five most efficient defenses in the league last year is sort of like nominating Catwoman and Baby Geniuses 2 for next year's Best Picture Oscar.
It isn't just in 2003 that weird teams show up at the top of these statistics. In 2002, the top offense in yards per point was New Orleans. The top defense was Tampa Bay, but ranked number three was that defensive juggernaut, the Giants. They weren't even in the top half of the league in defense according to DVOA.
And there's one more problem with yards per point. The goal is to pick out teams that are better than raw yardage would otherwise indicate, but what works for the 2001 Patriots doesn't work for that other team from the past few years that won despite unimpressive numbers, the 2003 Panthers. The Panthers were 17th in offensive yards per point (15.7) and 14th in defensive yards per point (15.5).
So what's the problem? While yards per point does a good job of indicating teams that can turn yardage into points (or prevent yardage from turning into points), it misses, or misattributes, three very important reasons why teams win football games: field position, special teams, and turnovers.
The appearance of the Baltimore offense in the 2003 top five makes the field position problem fairly apparent. I've discussed this numerous times: field position is fluid, and often the offense has a better chance to score because the defense gives them a better opportunity. This is, in fact, the entire point of the Baltimore Ravens. As long as their defense can stop people, and their special teams can provide good returns and long punts and kickoffs, it doesn't matter that their offense doesn't gain that many yards at a time (except when running an unexpected flea flicker). Each time they get the ball, they'll be closer and closer to the goal line, until at some point even a mediocre set of downs will result in a score. The Ravens' average drive in 2003 started at their 35.5-yard line, the best in the league and a yard-and-a-half closer to scoring than the second place team (surprisingly, the San Francisco 49ers, whose average start came on the 34-yard line).
So if the Ravens defense stops the opponent, and the Ravens get the punt back at midfield, now they need only 50 yards to score six points. Even better, let's say that Ray Lewis whacks the opposing receiver into the next county, and the ball comes loose, and now the offense gets to start from the opposing 25-yard line. Hey, the Ravens' offense is efficient -- they need only 25 yards to get six points!
Special teams create the same dilemma. Oakland, whose offense completely broke down last year and was eventually stuck with Rick Mirer at quarterback, ends up ranked ninth in the league in offensive yards per point. Efficiency? No, good special teams. The Raiders rank fourth overall in our special teams ratings for 2003, including first in the league in punt returns. So they were getting a lot of good field position that had nothing to do with offense or defense. They also have a very good kicker, which means that more of those drives turn into three points instead of zero points and a line of scrimmage change eight yards back. That may be a good indicator of overall team quality, but it isn't a good indicator that the offensive unit per se is efficient.
As for turnovers, there is a reason for the cliched statement that "the team that wins the turnover battle wins the game." While it is true that turnovers have an element of randomness -- a team just happens to hop on more fumbles than another team, one cornerback gambles and gets a pick while the other gambles and lets his receiver sprint for the touchdown -- avoiding turnovers on offense (and getting them on defense) is a real skill and one of great importance. The St. Louis Rams finished second in the NFL in offensive yards per point last season, but they were far from the most efficient offense. Every time they turned the ball over, it kept them off the scoreboard and, in many instances, gave the opposing offense easy field position. (There is also the issue of defensive touchdowns; why should an offense be rated more efficient when it gets six points with no yards, thanks to a defender returning a turnover for a score?)
Another good test of yards per point would be to look at how well it correlates with wins, compared to VOA or other statistics. (Correlation is explained here.) If you look just at offense and defense separately, yards per point correlates with wins just about as well as VOA, and much better than plain ordinary yards. Combine offense with defense, however, and VOA is a much better indicator of wins. Here are the correlation coefficients from 2000-2003 for VOA, yards per point, and yards. Special teams are not included, and this is VOA not adjusted for opponent:
|Yards per point||-.632||.615||.771|
(The negative numbers don't mean that correlations are less strong; that just recognizes that it is better to have fewer yards per point on offense and a lower VOA and yards allowed on defense.)
The point here is not that yards per point is not a useful statistic. It is a better measure of a team's ability than raw yardage. Sometimes it incorrectly lists a unit as "efficient" due to issues with field position or turnovers, but other times it does capture the idea that certain teams do a better job of converting their drives into points (or preventing other teams from doing the same). It's very useful for comparing two teams in college football, where there isn't any kind of standardized play-by-play data that would allow for complicated statistical breakdowns like DVOA. But when you are looking at the NFL, there's nothing in yards per point that isn't included in VOA/DVOA as well, and it doesn't really provide an explanation as to why a team that wins games doesn't perform well by Football Outsiders measures.
The other issue with yards per point that is worth mentioning is that it doesn't carryover from season to season. The NFL by its very nature has a lot of turnover from year to year, but yards per point is a lousy method of predicting how well teams will play the next year. In fact, raw yardage correlates much better from season to season than yards per point.
|Yards per point||.339||.193||.229|
(Once again, that's VOA, which is not adjusted for opponent, and no special teams are included. Year-to-year correlation for DVOA, adjusted for schedule strength each season, is even better, particularly if special teams are included.)