Thought today I would get some information out on football statistics that should be known to anybody interested in doing basic football analysis:
1. Passing statistics are crucial to NFL teams. The #1 passing statistic that correlates best to wins is touchdown passes (which is not really surprising). The NFL's passer rating system correlates very well to wins.
2. Offenses of teams and defenses of teams exist independently of each other for the most part. You will frequently hear announcers say that team y's offense is so explosive that it leaves their defense on the field far too long. In fact, it is well-established that there is a correlation between most points scored and *fewest* points allowed, so that if your favorite team is scoring lots of points and allowing lots of points, you can bet its because your defense sucks. However, the correlation this season is only 28%, which says that most of the difference in points allowed is actual difference on defense, not an artifact of the offense's performance. And that is a high number compared to recent history--the correlation in 2002 (which is after all the last complete year) was only 14% and in 17% in 2001.
3. Football statistics can be adjusted for era and when you do this many of the acknowledged best players from history come to the fore despite seasons that might seem poor by today's standards. For example, quarterback passer ratings jumped in 1978 and 1979 to the point where all the announcers were commenting on how vastly different the game had become. They were right. League-wide passer ratings had hit a cyclical low of about 60 in 1977; if that sounds amazing, consider that Tampa Bay's alleged quarterbacks combined for a total of something like 3 TDs and 36 INTs that season. The next year they jumped to like 68; they had never been so high before; they have never been so low since
Wednesday, November 12, 2003
I've been enjoying visiting the Football Ousiders'
website. They are the new home of Tuesday Morning Quarterback, Gregg Easterbrook's excellent weekly column of observations on the football weekend.
They rate the teams via a method that involves analyzing play by play data. I happen to think this is a great method, for reasons that I will get into later. However, for whatever reason, their rating system seems to be currently flawed, as best as I can tell with a limited amount of data. It may be that time will prove their system right, and my system wrong.
Let me start by explaining my system. Each game for every team is adjusted to an equivalent score against an average NFL team on a neutral field. For example, the week before last San Francisco beat the Rams at home by a score of 30-10. The spreadsheet notes that St. Louis has scored 1.281 times as many points in all its games as an average NFL team. Therefore, San Francisco's giving up 10 points to the Rams is like giving up 7.8 points to an average NFL team (10/1.281). Since the game was in SF, we add an additional point for the home field advantage and say that San Francisco's defense performed equivalent last week to giving up 8.8 points to an average opponent on a neutral field. St. Louis’s defense has given up .956 times as many points this year as an average NFL team. Therefore San Francisco's 30 points scored is equivalent to scoring 31.4 (30/.956) points against a normal team, adjusted to 30.4 points on a neutral field... to be continued...