Which NFL Stats Best Predict Results?
What NFL stats mean the most in terms of predicting results? Which stats should we ignore? Today on the blog Mark Taylor takes a look at where each team in the NFL ranks in the stats that matter, with a quarter of the 2012 season completed.
The vast majority of the 32 NFL teams have now completed a quarter of their regular season schedule and already opinions are solidifying around which teams are likely to be around for the post season and which teams won't see any post December action.
Seattle's game winning interception Hail Mary against Green Bay in week three, predictably became the tipping point in the refereeing lockout. The replacement refs were finally declared not fit for purpose and quickly found themselves replaced.
Assessing Form Of NFL Teams
Quantifying the quality of refs, especially when there is such a wide ranging difference in ability and experience between the replacements and the locked out seniors, is relatively easy. But what about rating the teams in a sport with far fewer regular season games than football and a squad turnover that also provides less season on season continuity. Are the current high flying NFL teams really the current cream of the crop or have they played against easy teams and will they fall back to earth as the season progresses.
The NFL is bursting at the seams with statistics that can be used to evaluate the current strengths of sides, but which set of numbers should we really be taking notice of and which ones are mere statistical padding?
To be useful, rather than simply descriptive of past events, a statistic must also predict future performance. In tournaments with limited games, it's also desirable that the stat stabilizes relatively quickly. We don't want to wait until week 14 before we can start to make educated match predictions. And lastly, we need to be able to recognize when teams have benefitted from a relatively weak schedule.
Which NFL Stats Matter?
Some recorded numbers tick some, but not all of these boxes. Turnovers for example in the form of unrecovered fumbles and interceptions have a huge impact on a single game result. Less than 10% of teams can overcome a turnover differential of three or more to win either straight up or against the spread, so turnovers are powerful in game indicators. Unfortunately, the partly random nature of a turnover and the relative rarity of the occurrence leads to past ability in creating or allowing turnovers being a poor predictor of future performance in this area.
Instead we need to look at the so called efficiency statistics, such as yards per run and yards per pass on both sides of the ball. Combined, they not only do an excellent job at predicting the most likely future match outcome they also rapidly become stable and crude, but effective adjustments for strength of schedule make them an essential tool to use through the winter months.
Seattle's dramatic last second touchdown (it was actually an interception in the end zone by Green Bay) did illustrate one point that has become universal in the current NFL. It is primarily a sport where the offense gets the benefit of the doubt, but more importantly from a statistical standpoint, it is a passing league. Increasingly and in general, the better the team passes the ball then the more success they can expect to enjoy.
So to be able to better understand how teams have won in the past and how they are likely to win in the future we must firstly calculate our efficiency statistics. Rate figures are always preferable compared to accumulated totals, where persistence rather than skill is rewarded. Therefore rushing yards per attempt and passing yards, minus sack yards per passing attempt provide an excellent starting point to analyze NFL matchups.
Incorporating Strength Of Schedule
These fundamental NFL stats are useful, but they become more potent if we now correct for strength of schedule. A team may have impressive raw yards per rush number because they really are a good running side or because they have so far faced poor running defenses, or more likely a combination of the two.
As of Monday all bar Indianapolis and Pittsburgh had played four games. Each of a team's four opponents would most probably also have played four games. So the opponents of a team's opponents will number 16. We can therefore see how a team's opponents fared against this wider and hopefully balanced batch of opponents and compare these efficiency figures to those recorded by our target team.
For example, Denver as a team have rushed for 3.86 yards per carry so far this season, good enough for them to be ranked around midway in this category. However, this raw stat has been achieved in games against four below average run defenses. Denver's opponents so far have been Pittsburgh, Oakland, Atlanta and Houston and they have allowed almost 4.6 yards per carry in their 15 combined games.
Suddenly Denver's 3.86 yards per carry they gained against a group of teams which are allowing 4.6 yards per carry looks a lot less impressive. The Broncos are only gaining 84% of the yardage per carry of their opponent's opponents.
Team By Team Ratings So Far In 2012
Below I've listed the passing and running stats expressed in this type of percentage terms for each team so far this season. Figures for the offense in excess of 100% indicate a team playing above average and figures below 100% for defenses similarly indicate a defense playing above average football. Teams are listed in decreasing likelihood that they would defeat a perfectly average opponent. Therefore, the Ravens are currently the best side in the NFL.
|Team||Rushing Offense %||Passing Offense %||Rushing Defense %||Passing Defense %|
These figures provide a variety of uses. Firstly, they stabilize very quickly. Around three quarters of the teams that are either above or below average in their various categories by week four will remain so for the rest of the season. The inclusion of strength of schedule components can highlight teams who have feasted on the weak. We have already seen that Denver tumble from 17th in raw offensive running stats to 28th once schedule is partly incorporated into the mix and they are by no means the exception.
Most importantly this type of number, whether by eyeballing such things as passing mismatches or by more formally using historical data regressed against actual results, can be used in selecting game outcome.
We can also illustrate the importance of passing in the modern NFL. Regression analysis using these passing and running efficiencies indicates that in a game between two entirely average sides, the home team will win about 57% of the time. Well in line with what actually occurs. If the home side was then able to improve their passing efficiency by 20% keeping everything else constant, they would go on to win 68% of such games. But a similar increase in just their running efficiency would only lift their long term win percentage to 63%. The NFL truly is a passing league.
I’ll return to this topic at the end of the regular season, where we can use this approach to sift the Superbowl contenders from the post season teams who may owe their presence to a weak division and an easy inter conference schedule.
Read more of Mark's work on his The Power Of Goals blog
And follow Mark of Twitter: @MarkTaylor0