Wednesday, January 13, 2010

Revisiting the 2009 NFL Season: An EXTREMELY Observational Analysis

In a past post, Guy mentioned his discontent with using RSD as a measure of balance at the 9-game point in the season for the NFL. If you think back to the post-Patriots crushing of the Titans coverage, announcers were claiming that the NFL was out of whack and the balance in the game was decreasing. I suggested that probably wasn't the case to any significant extent, and it seems like the end of the season played out pretty evenly (I can't even count how many teams were in the AFC Wild Card hunt). I decided to put together some random facts about this season vs. past seasons in NFL in order to see if there really is a balance change going on.

I started with the average points per game (per team) from 2005 to 2009. Why? Just to get an idea of the scoring environment in each season:

2005: 21.48
2006: 20.66
2007: 21.69
2008: 22.03
2009: 21.47

Of course, this doesn't tell us anything about balance. I was actually surprised to see the very little difference between these seasons. What about point differentials. First, I took the standard deviation of the difference in and Points For and Points Against across teams for each season (the mean is, of course, ZERO):

2005: 91.39
2006: 92.53
2007: 116.32
2008: 107.21
2009: 119.31

2005 and 2006 definitely stick out here. As for 2009 compared to the previous 2 seasons, I don't see anything special. To add to this, I calculated the mean and standard deviation of point differentails using the individual game scores, rather than team totals (Means with SD is in parentheses):

2005: 11.6 (9.61)
2006: 11.43 (8.81)
2007: 12.47 (9.50)
2008: 12.22 (9.54)
2009: 12.97 (10.37)

I didn't run any statistical tests on this small sample; however, 2009 definitely does stick out somewhat here not only in the mean score differential, but also the variability we saw across the season. I'm not sure how to do this myself, but I'm curious if this is a product of uneven competition, or style of play in football. Could it be that the pass first mentality is increasing, while running is used only after a lead? If this is the case, I think we would expect teams with early large leads to maintain those leads through the end of the game, then run out the clock but not score later, leaving us with large gaps in scoring, but not much difference in the average score per team per season. I'm just thinking aloud here, and I'm not really sure how things would play out in the results for the season. Next, I counted the number of games that were decided by two touchdowns or more over the time period:

2005: 75
2006: 76
2007: 86
2008: 88
2009: 87

I followed this up by simply counting the number of games where the winning team scored more than 28 points (or more than 4 touchdowns+XP), and then the number of games in which the losing team scored less than 14 points (or less than 2 touchdowns+XP).

Greater Than 28 Points for Winners (losers)

2005: 89 (9)
2006: 92 (12)
2007: 116 (14)
2008: 123 (17)
2009: 112 (13)

Less Than 14 Points for Losers (winners):

2005: 117 (30)
2006: 117 (28)
2007: 109 (29)
2008: 105 (26)
2009: 115 (22)

I see a couple possible anecdotal things from this. First, it seems to have been increasingly difficult to win if you do not score 14 points or more. At the same time, many more winning teams are scoring more than 4 touchdowns than in 2005/2006. Here's an aggregate of the two, where I counted the number of games that had both a >28 point scoring winner and a less than 14 point scoring loser:

2005: 30
2006: 28
2007: 36
2008: 44
2009: 45

So, as expected, we see the increasing gap that we saw from before. This is all just some observational data, and I'm not trying to make any conclusions from it. But right now, it doesn't seem like there should be significant concern for this season, but that some sort of structural change occurred after the 2006 season. Below is the Standard Deviation of Win Percents for the NFL:

2005: 0.193
2006: 0.181
2007: 0.208
2008: 0.207
2009: 0.201
Again, 2007 seems to be the first season in which we see a significant change in the apparent balance of the league. So if announcers wanted to blow their whistles, perhaps that would have been a better time to do so. I'm still not all that worried, though. Perhaps it's simply a change in the tails of our team talent distribution. We could evaluate this using the expected tail probabilities if all teams were of equal strength...or the probability that we would see teams outside 2 Standard Deviations (in this case, below 0.250 and above 0.750) from the mean win percent (0.500, of course). Below I list the percentage of teams we see outside the expected if the league were perfectly balanced:

2005: 12.5%
2006: 15.6%
2007: 18.8%
2008: 12.5%
2009: 18.8%

It's tough to decipher anything here, but it looks like we have more in the tails than we should. For perspective, we would expect about 4.6% of the teams to be in the tails based on random chance alone in a perfectly balanced league. Of course, these numbers are affected by the number of games played head-to-head, but in an relative sense, I think they are interesting to look at.

The way I see it, one of three things could be happening:

1) The overall parity in the league decreased in 2007 and has been similar since. 2007 was the Patriots' big year, so we would probably expect to see something going on that year.

2) The structure of the game has changed and when teams are way ahead, they are much more able to keep that spread (and this may not matter which team is ahead). In addition, given the high variance, teams that are in close games are more likely to stay close. Perhaps if the Titans were up by 28 on the Patriots in the first quarter, they would have continued scoring, while the Patriots would not, even given the same talent. Why would this happen? I don't know. But it is a possibility.

3) Nothing has happened. Given the small sample I show here, the changes could be simply random in the scheme of a larger sample of NFL history.

In the end, this little look is not nearly enough data here to really make any conclusions. I'm short on time, but anyone willing to look further back would be welcome to do so (I invite anyone interested to look, and possibly add more comparisons of different measures). However, I think it's important to note that if announcers wanted to raise significant concern over the balance in NFL, they should have done it when the Patriots went 16-0 in 2007.
For reference, below are the end-season RSD's for the years in question. These really don't raise much concern with me over the general dispersion of winning for 2009. Again, we see a slight change in 2007, but over the last 3 years, RSD is lowest (more balanced W%) than the previous 2 seasons. Perhaps some of it is a result of the Colts losing their last 2 games. But it's tough to really make a conclusion about whether or not they definitely would have won them, so I'll leave it as is.
2005: 1.540 (different from what I had before, but this seems to be correct)
2006: 1.447
2007: 1.661
2008: 1.658
2009: 1.612

No comments:

Post a Comment