Week 8 Picks

Okay, that title is false; I didn't actually make picks.  I was trying to find out the best way to predict a winner, if I could go strictly with Power Rankings, or if momentum comes into play.  I never came up with a good way to approximate momentum, and Power Rankings was the best way to predict.  So....go ahead and put together your own picks...

Power Rankings (Week 8)

No changes to the formula; I did a little playing around with the win rating but didn't see any big improvement.  However, this coming Thursday I'll post again, predicting this week's game.

Rank	Team	Power Ranking
1	Pittsburgh Steelers	1.000
2	New York Jets	0.928
3	New England Patriots	0.856
4	Baltimore Ravens	0.834
5	Tennessee Titans	0.825
6	Houston Texans	0.818
7	Indianapolis Colts	0.788
8	Atlanta Falcons	0.780
9	New York Giants	0.731
10	Seattle Seahawks	0.722
11	Kansas City Chiefs	0.717
12	Green Bay Packers	0.690
13	Washington Redskins	0.690
14	Tampa Bay Buccaneers	0.671
15	Chicago Bears	0.669
16	Philadelphia Eagles	0.667
17	St. Louis Rams	0.647
18	Cleveland Browns	0.633
19	Denver Broncos	0.627
20	Arizona Cardinals	0.625
21	Miami Dolphins	0.608
22	Jacksonville Jaguars	0.604
23	Oakland Raiders	0.591
24	Minnesota Vikings	0.552
25	Cincinnati Bengals	0.552
26	New Orleans Saints	0.535
27	Detroit Lions	0.517
28	Dallas Cowboys	0.500
29	San Diego Chargers	0.498
30	San Francisco 49ers	0.429
31	Carolina Panthers	0.422
32	Buffalo Bills	0.380

 No big shake-ups in the rankings, the top 4 teams are still there.  The NFC is kind of all over the place from week to week, are the top 3 teams in the league Atlanta, NY, and Seattle?

Power Rankings (Week 7)

To test my rankings, I went back to old seasons (2004 through 2009) and calculated a final regular season power ranking.  Next, I compared how these final rankings compared to playoff performance: if someone ranked higher than another team (and subsequently beat them) then hey, a plus; if a lower ranked team beat a higher ranked one, a minus.  Then finally, I tweaked through a whole range of coefficients on the Scoring Efficiency, Yardage Efficiency, and Margin of Victory.  Basically, I tried to find which coefficients best predicted playoff performance.

Fortunately, the behavior was really weird.  For three of the seasons I looked at, Scoring Efficiency was the only value that mattered.  If I set all the other items to zero (and used only Scoring Efficiency), then I best predicted the post-season.  So for this week's power rankings, I used this equation for the win rating:

Win Rating = exp(-1/(3* Scoring Efficiency ))

Thus, the only thing I'm using to rate a team is how they score; if it takes them a few yards to score points, that's good, if it takes alot, it's bad.  It's interesting; bad teams have drives that stall, turnovers, etc.  Good teams score on every drive, and force turnovers.  I'm going to explore what this means in the next couple of weeks, we'll see how it goes. 

And here's this week's power rankings:

Team	Power Ranking
New York Jets	1.000
Pittsburgh Steelers	0.978
New England Patriots	0.900
Baltimore Ravens	0.885
Houston Texans	0.797
Indianapolis Colts	0.797
Tennessee Titans	0.792
Atlanta Falcons	0.792
Philadelphia Eagles	0.767
Seattle Seahawks	0.730
New York Giants	0.725
Green Bay Packers	0.716
Chicago Bears	0.708
Denver Broncos	0.707
St. Louis Rams	0.701
Jacksonville Jaguars	0.695
Arizona Cardinals	0.695
Washington Redskins	0.683
Kansas City Chiefs	0.681
Miami Dolphins	0.645
Minnesota Vikings	0.624
Tampa Bay Buccaneers	0.608
New Orleans Saints	0.594
Cincinnati Bengals	0.578
Cleveland Browns	0.573
San Diego Chargers	0.572
Detroit Lions	0.550
Oakland Raiders	0.544
Dallas Cowboys	0.522
San Francisco 49ers	0.456
Buffalo Bills	0.412
Carolina Panthers	0.388

J-E-T-S baby.  But actually, the Jets, Ravens, Steelers, and Patriots are all pretty equal.  Let them slug it out. 

Power Rankings (Week 6)

Going off last week, we can calculate a win rating for the winning team; for the losing team, I simply subtract the win rating from 1.  The simplest way to calculate a total team ranking is find the average win rating across all games.  However, we wouldn't be taking into account the teams they're playing, and we wouldn't be taking into account the teams those teams were playing, etc.  It's what they call a recursive function.  At some point we would have to stop calculating these team rankings, as (1) it would take forever, and (2) the differences would be so small that it wouldn't even matter.  So, at some point I'll terminate the series, calculating a team ranking on win ratings alone.  Here's the formula:

TeamRanking1(x) = average (WinRating2 * TeamRanking2(x-1) + WinRating3 * TeamRanking3(x-1) +....)

Where x is some value.  When x equals 0:

TeamRanking1 = average (WinRating2 + WingRating3 + ...)

For this week I set x to 4; any longer and the calculations take alot longer.  My final step is I scale everything to 1; divide each team's team ranking by maximum calculated team ranking to get a final power ranking.  And that takes us to this week's rankings:

Team	Power Ranking
Pittsburgh Steelers	1.000
Baltimore Ravens	0.931
Indianapolis Colts	0.809
Tennessee Titans	0.791
New York Jets	0.759
Atlanta Falcons	0.742
Houston Texans	0.715
New York Giants	0.690
Chicago Bears	0.683
Kansas City Chiefs	0.668
Washington Redskins	0.668
Green Bay Packers	0.645
Philadelphia Eagles	0.628
Denver Broncos	0.627
New England Patriots	0.624
Dallas Cowboys	0.615
Tampa Bay Buccaneers	0.596
San Diego Chargers	0.595
Jacksonville Jaguars	0.595
Cincinnati Bengals	0.581
St. Louis Rams	0.572
Oakland Raiders	0.564
Arizona Cardinals	0.528
Cleveland Browns	0.526
Detroit Lions	0.513
Seattle Seahawks	0.508
Minnesota Vikings	0.461
New Orleans Saints	0.459
Miami Dolphins	0.415
San Francisco 49ers	0.317
Carolina Panthers	0.285
Buffalo Bills	0.247

It's interesting, although we have a bunch of one and two loss teams, there's a pretty clear separation; the Steelers and the Ravens are the teams to beat.  And of course, over time, we'll have a more clear picture.  Next week, I'm going to analyze some earlier seasons, and see how I need to tweak the win rating / power ranking.

Win Rating

So this blog is my own weird musings on putting together some kind of football ranking system.  I suppose this all stems from me trying to find something to do while watching football.  It’s not like it’s exactly boring, it’s just I need my mind on something else to stay involved in the game, if that makes sense.  Maybe the best use of my mind, well, I won’t say best use, but a use, is coming up with this ranking system.  In college football, I always considered the BCS computer ranking idea an interesting one.  I mean, think of it, we have arguably the second most popular sport in the country (after the NFL) decided by some system of equations and assumptions.  I like the elegance of the BCS computer over relying purely on polls, it’ll help keep Florida State out of a Fiesta Bowl.  But it seems throughout its history, they’ve been removing variables from ranking system.  For instance, margin of victory is taken out.  Now it’s probably reasonable the BCS has to make changes like that, political pressures and all that.  However, I’m just going do what I feel like here.  I may radically change how I do things at any point, but well, I need somewhere to start from.

In order to come up with any kind of team ranking, the first step is to come up with some way of quantifying a team’s success in a game, think of it as the quarterback rating for the whole team.  This post is purely about putting together that metric.  Here’s the base numbers I’m working with: total offensive yards, points scored, and turnovers.  Points is obvious, and yardage (on both sides of the ball) I feel is a fine indicator of how effectively a team can move the ball, stop an offense, etc.  I’m not considering anything too sexy like pass completion rate, special teams yardage, sacks etc.  For one, it’s hard to get that data in any simple form I can ingest in to a spreadsheet.  And two, I can chug through a lot of seasons by ignoring this information.  So anyway, here’s the three metrics I came up with:

Scoring Efficiency = 100 / (Winner’s Offensive Yards / Total Points))

Yardage Efficiency = (Winner’s Offensive Yards + 38 * Loser’s Turnovers) / (Loser’s Offensive Yards + 38 * Winner’s Turnovers)

Margin of Victory = (Winner’s Points – Loser’s Points) / 7

Now margin of victory is pretty simple, it indicates the number of possessions the winner won by.  For the yardage efficiency, I divide the two yardage values by each other; I feel I can negate any weather / field conditions this way, while still giving a team a way of dominating.  In order to simplify my life, I consider a turnover as a 38 yard play (just quickly, the average net yardage change for a punt in the NFL is 38 yards).  So, if the defense generates two turnovers, I add 76 yards to the total offensive yardage (and if the offense coughs the ball up two times, I add 76 yards to the other teams total offensive yardage).  And then finally, I came up with this metric called scoring efficiency, it finds how many yards it takes to score a touchdown.  It’s my way of approximating stalled drives, failed fourth down conversions, effectiveness of the punting game, and also any special teams touchdowns.  The fewer yards it takes to score a touchdown, the better the team is.

Now to put this all together, I wanted to have a decaying function approaching 1.  I like this idea, I feel it’s a way to take margin of victory into account, where winning by two touchdowns is a bigger deal that winning by three, and is a bigger deal than winning by four, etc.  Also, my goal was to have a bare minimum winning performance to get around 0.5, while any dominating performance will be in the 0.9 range.  So here’s how I combine the values:

Win Rating = exp(-1/(Scoring Efficiency + Yardage Efficiency + Margin of Victory))

For a little example, below are all the games from week 4 of the NFL:

Winner/tie	Loser/tie	Win Rating
St. Louis Rams	Seattle Seahawks	0.808
Jacksonville Jaguars	Indianapolis Colts	0.696
Baltimore Ravens	Pittsburgh Steelers	0.664
San Diego Chargers	Arizona Cardinals	0.896
New Orleans Saints	Carolina Panthers	0.631
Atlanta Falcons	San Francisco 49ers	0.650
Cleveland Browns	Cincinnati Bengals	0.655
Denver Broncos	Tennessee Titans	0.734
New York Giants	Chicago Bears	0.813
Green Bay Packers	Detroit Lions	0.666
New York Jets	Buffalo Bills	0.867
Houston Texans	Oakland Raiders	0.753
Washington Redskins	Philadelphia Eagles	0.669
New England Patriots	Miami Dolphins	0.869

So as you can see, anything in the high .8s is a comfortable victory, while anything in the .6s could have gone either way.  Every week I'll be using this win rating to throw together an overall power ranking - when I put one up I'll tell you how I compute it.