Well, as I see it, the best team in the NFL is N’ahleans.  You all may remember that last year I did weekly NFL Power Rankings.    I hadn’t gotten around to doing them yet this year, and somehow the front page launched the NFL’s hottest teams as one of our sign-on screens, which was pretty embarrassing because the data was a year old.   I think it was an internal plot to get me back to work.

            OK, you remember how these ratings are figured?    Each team starts with a score of 100.00.    In the first week of the season San Francisco played Arizona, at Arizona, and the 49ers won 20-16.    If we assume that the home field advantage is 2.5 points, that suggests that San Francisco is 6.5 points better than Arizona—the four points by which they won, plus 2.5 for the fact that they were on the road.   If the total value of the two teams is 200.00 points and San Francisco is 6.50 points better than Arizona, that means that San Francisco is at a level of 103.25, and Arizona is at 96.75.  

            We figure every game like that.   San Francisco in the first round of calculations has output scores of 103.25 for the Arizona game, 105.25 for their second game, 99.75 for their third game) and 116.25 for the fourth game—an average of 106.125.   Arizona has first-round scores of 96.75, 108.25 and 88.25—an average of 97.75.   Thus, in the second round of calculations, the total value of these two teams is 203.875 (106.125 plus 97.75).   That makes a second-round output score, for this San Francisco/Arizona game, of 105.1875 for San Francisco, and 98.6875 for Arizona.

            We repeat this process many times, putting the data through many rounds of calculations, until the data stops moving.    We reach a point, eventually, where the output data is exactly the same, whether you start with initial values of 100.00 each for San Francisco and Arizona, or of 200 for Arizona but zero for San Francisco.   The “initial assumption” data entirely disappears, and the end results are shaped entirely by the results of the contests.  

            By this process, the best team in the NFL is the Saints.   The Saints are 4-0; they have beaten Detroit, Philadelphia, Buffalo and the Jets by a total of 78 points.   The quality of their competition hasn’t been great, but neither has it been bad; it’s middle-of-the pack.   Philadelphia and the Jets are good teams.   The Jets, then, have been 19.5 points a game better than near-average competition—thus, they rank at 119.1.

            These are the power rankings for the 30 NFL teams:

            They’ve only played four games each—three games for some teams—and you can’t read too much into it.   One touchdown is seven points—still about 2 points in a team’s ranking.    The rankings will change a lot by the end of the season.

            While we are doing this, however, there is a lot of other information that we can gather.   We can figure each team’s strength of schedule so far.    Arizona has played the toughest schedule so far; the New York Giants have played the easiest.  

            Arizona has played San Francisco, Jacksonville and Indianapolis—all good teams.   The Giants have played three of the league’s worst teams—Tampa Bay, Washington and Kansas City—plus one team (Dallas) that appears to be, based on what we can tell so far, pretty average.

            We can figure each team’s “temperature”, although, at this time of the year, the order of the teams by temperature is nearly the same as the order of the teams by the power ranking.    These are the hottest and coldest teams:

            Later on in the season, there will be teams which have a high temperature but a low power ranking.  At this point of the season, though, they’re all the same, just on a different scale. 

            One of the most striking things about this season’s data, as opposed to last season’s, is that the “scoring and allowing” tendencies of teams are much more pronounced than they were last season.    The most wide-open teams in the league are Philadelphia and Detroit.   We estimate at this point that if Philadelphia were to play Detroit, about 70 points would be scored.   The most tight-to-the-vest teams so far are Denver and Washington.   We estimate that if Denver were to play Washington, only 12 points would likely be scored.   These numbers are much more extreme than the numbers we had last season—perhaps because it is early in the year, and the ups and downs of scoring haven’t evened out yet.   But when we get to the predictions for this week, we’re going to have very different point projections for several of the games than Las Vegas does.

            Another thing we can figure is each team’s consistency.   The most consistent team in the league, so far, has been San Francisco.    We have “output scores”, for San Francisco’s four games this year, of 112.2, 111.7, 109.7 and 112.5—all within a three-point range, and a standard deviation of 1.4.     The most inconsistent team has been Arizona, with scores for their three games ranging from 113.4 to 97.4.

            Looking ahead to week five, there are four teams that have bye weeks—the Packers, Saints, Bears and Chargers.   That leaves 14 games.    Four of those games, we see as likely blow-outs:

Minnesota at St. Louis

            Vikings 31, Rams 7

            Tampa Bay at Philadelphia

            Eagles 39, Buccaneers 14

Atlanta at San Francisco

            49ers 29, Falcons 10

            Indianapolis at Tennessee

            Colts 33, Titans 18

            I don’t know if 33-18 is really a “blowout”, but anyway, it’s clear who should win, which doesn’t mean they will.   Tennessee is not a bad team; they’re just a team that can’t find a win.   Doesn’t mean that they won’t next week.  

            All of those games, of course, we are picking the same way as the oddsmakers are picking them, and in all cases we have a larger margin of victory than the oddsmakers do.   You probably should bet on them to be right, rather than us; they’ve been doing this a lot longer than we have.   But we don’t see the underdogs in any of these games as being likely to cover the spread.

            On the other ten games, we agree with the Las Vegas oddsmakers on eight, disagree on two.   Working our way down the margin of victory scale, we come to Oakland at the Giants.  We have the game as Giants 24, Raiders 9.   The oddsmakers have it, essentially, as Giants 27, Raiders 12.   They see 39 points; we see 33.   Other than that, we basically agree with the oddsmakers.

            Arizona at Houston, we have it as Cardinals 31, Texans 21.   The oddsmakers have the Cardinals as 5 and a half point favorites with 49 points to be scored, which means 27-22.  Again, we’re close.

            Pittsburgh at Detroit.  We see it as Steelers 29, Lions 20.     The oddsmakers have it, in essence, at Steelers 27, Lions 17.    We think more than 44 points might be scored; otherwise, we agree.   A similar game is Dallas at Kansas City—similar, because in both cases you have a team which was good last year and has been so-so this year, on the road, playing a team that probably isn’t quite as bad this year as they were last year.  We see the Cowboys as winning 28-20; Las Vegas has it 26-17.   No real difference. 

            Cincinnati at Baltimore, we have it Ravens 27, Bengals 21.    Las Vegas has it Ravens 25, Bengals 17.   We think that more than 42 points might be scored in the game; otherwise, we’re in line.

            Cleveland at Buffalo. . .when two teams this bad get together, anything can happen.  We get the same score here as we did in the other game—Buffalo 27, Cleveland 21.   The oddsmakers have it Buffalo 23, Cleveland 17. . ..so, same thing.   We think that more than 40 points will be scored, otherwise we’re OK with it. 

            The Monday night game is the Jets at the Dolphins.    Las Vegas has it 19-17, Jets, or let’s say 20-17 because 20-17 is a better score for a football game than 19-17.   We have it as Jets 13, Dolphins 7.   So they see 36, 37 points being scored in the game, we see 20.

            This gets back to the wisdom of crowds.    We’re projecting 20 points—which is a very low number—based on the scores of the games these teams have played this year, in which it appears that both teams are dragging the average downward.   What Las Vegas knows by experience is that there aren’t very many games in which only 20 points are scored, so you shouldn’t bet on that happening.    About which they are certainly correct, but. ..that’s what our model predicts, so we’re going with it until we refine the model.

            Jacksonville goes to Seattle.   Las Vegas has the Jaguars winning by 3; we have it 4, 21 to 17.

New England at Denver.  Here, finally, we’re going to split from Las Vegas.   The oddsmakers are favoring New England by 3 points.   We have it Broncos 15, Patriots 9—eight field goals.    They have 41, 42 points being scored; we have 24.  

            Washington at Carolina. . .Las Vegas has Carolina winning this by 3 or 4 points, about 20-17.  We have Washington winning it by one—Washington 10, Carolina 9.

            So. . .not a betting man, not urging you to bet.    But if you’re betting, in those two games we would bet the underdog, and we would bet the under in the over-under.

Bill James

Lawrence, Kansas

October 7, 2009