Welcome to the 5th NFL PT Awards.  The purpose of these
awards is to help raise the publicity for those systems that
have been superior in various qualities of interest.
Most awards will be based entirely on the numbers I've
monitored with my prediction tracker web pages.  In 2004 I
followed the weekly performance of 50 computer systems rating
the NFL.


  BEST STRAIGHT UP WINNERS (Entire Season)

Winner: Least Absolute Value Regression

  For the second year in a row one of the systems that I operated had the
most correct picks for the season.  This year it was Least Absolute Value
Regression.  The credit for this one goes to Gilbert Bassett, a professor
at the University of Illinois-Chicago, who wrote about this system in 1997.
Least Absolute Value Regression had a record of 180-87, 67.42%.
  The Pigskin Index finished one game behind.  The line had a poor season,
finishing in 11th place.  Least Absolute Regression finishes just a fraction
off the record of 67.92% set by Beck Elo last season.


 2004 Winner:  Least Absolute Value Regression
 2003 Winner:  Beck Elo
 2002 Winner:  Herman Matthews
 2001 Winner:  JFM Power Ratings, ARGH Power Ratings
 2000 Winner:  CPA Rankings
 1999 Winner:  Kenneth Massey Ratings




  BEST AGAINST THE SPREAD (Entire Season)

Winner: Least Absolute Value Regression


  This might not be the most useful measure of the quality of a football
rankings system but betting on the NFL is big business.  And yes I know it
would be more meaningfull if I looked at a subset of games like when a system
was more than three points away from the line but all I have is every game results.
This season the computer ratings did a very good job overall against the spread.
The lines seemed to be off for about the first two thirds of the season.  Only 5 out
of 50 systems did worse than 50%.
 The best system picking against the spread was Least absolute value regression (L1)
L1 had a record of 149-110 (57.53%).  This sets a new record in this category.
L1 had a great season, setting this record and barely missing the straight up record.


 Winner 2004:  Least Absolute Value Regression
 winner 2003:  CPA Retro Rankings
 winner 2002:  FreeSportsPlay.com
 winner 2001:  CPA Rankings
 winner 2000:  Yourlinx
 winner 1999:  PerformanZ Ratings




  SMALLEST MEAN ABSOLUTE ERROR (Entire Season) 

Winner: Vegas Line

  The line remains on top in this categoy this season.  The line had an
average error of 10.4213 for all games.  The closest competitors this year,
other than the opening line, was Grid Iron Gold at 10.49.  The record 10.28,
obtained by the line in 2001.

 Winner 2004:  Vegas Line
 winner 2003:  Vegas Line
 winner 2002:  FreeSportsPlay.com
 winner 2001:  Vegas Line
 winner 2000:  Vegas Line
 winner 1999:  Vegas Line




  SMALLEST BIAS (Entire Season)  

Winner: System Average

 Thes lowest system bias this year was found in the average of all the
systems combined, with an average bias of 0.0167.  I would say the
typical system this year gave slightly too many points to the home team.
Jeff Bihl had a mark that beat the system average but since he only
had data for the second half it is hard to compare.


 Winner 2004:  System Average
 winner 2003:  CPA Rankings
 winner 2002:  Sagarin Pure Points
 winner 2001:  ARGH Power Ratings
 winner 2000:  Pigskin Index
 winner 1999:  Flyman Ratings




  MOST ACCURATE SYSTEM - SMALLEST MEAN SQUARE ERROR (Entire Season)

Winner:  Vegas Line

 Despite the fact that a couple of times about I have talked about the
line have a down year it actually comes out winning the most accurate
system award.  Interestingly it wasn't even that close.  The line had
an average mean square error of 166.808.  Which is a big improvement
over recent years and even sets the 5 year record in this category.
The closest competitor was Hank Trexler at 172.



 Winner 2004   Vegas Line
 winner 2003:  Vegas Line  (Opening)
 winner 2002:  FreeSportsPlay.com
 winner 2001:  Vegas Line
 winner 2000:  Pigskin Index





                        SECOND HALF AWARDS 

 I like to look at the results over the second half of the season.  The
systems have had time to become 'burned in' to the season's data and
there is less reliance on starting values or previous seasons.
Second half data consists of all games from week 10 through the superbowl.


  BEST STRAIGHT UP WINNERS (Second Half)

Winner:  Pigskin Index

  For the first time ever we  have a single winner here.  It had been a three
way tie for the last three seasons.  But five teams did tie for second place
only one game behind Pigskin Index.  Pigskin finished the second half with
a record of 99-38, 72.26%.  This also sets a record for this category.
It seems the 2004 season was an extremely good season for NFL systems,
setting new records in almost every category.


 Winner 2004: Pigskin Index
 Winner 2003: JFM Power Ratings, Kasulis Enhanced Spread, Beck Elo
 Winner 2002: Least absolute value regression, JFM Power Ratings, CPA Rankings
 Winner 2001: Pythagorean Ratings, Elo Ratings, Ken Massey Ratings
 Winner 2000: PerformanZ Ratings, Scoring Effeciency
 Winner 1999: Ken Massey, Ed Kambor, Pythagorean Ratings




  BEST AGAINST THE SPREAD (Second Half)

Winner:  Stewart Huckaby

  In what was a good overall season against the spread the record for
best against the spread in the second half we also saw unusually high marks
during the second half.  The best was ARGH Power Ratings, with a record of 74-51, 59.2%.
Several others were high as well, Matthews Grid finished just a fraction behind ARGH.


 Winner 2004: ARGH Power Ratings
 Winner 2003: Logistic Regression, Viacheslav Ugolnikov
 Winner 2002: Least absolute value regression
 Winner 2001: Stat Fox
 Winner 2000: Pythagorean Ratings
 Winner 1999: Least squares using team specific home field advantages




  SMALLEST MEAN ABSOLUTE ERROR (Second Half)


Winner: Least Squares Regression

  Historically, the Vegas line dominates the mean absolute error category everywhere
except for here, the second half of the NFL.  This year the best mean error for
the second half of the season was 10.367, Least Squares Regression (L2).  The Regression
systems had a very good year.  L2 was actually fairly dominant here, The closest compitor
was my computer adjusted line at 10.51 and the line at 10.52.
L2's mark is just barely off the record of 10.3588 set by the line in 2001.


 Winner 2004: Least Squares Regression
 Winner 2003: Vegas Line
 Winner 2002: FreeSportsPlays.com
 Winner 2001: Vegas Line
 Winner 2000: Pythagorean Ratings



  SMALLEST BIAS (Second Half)    

Winner: Tom Benson Gridmarks

  This category was not nearly as good as the other categories this year. Although
not as bad as last year, systems were still off, giving a  half point or so to
the home team.  The best mark this year was by Tom Benson, 0.0216.  I have
to be a little disappointed as my PerformanZ Ratings finished just .002 points behind
in second place.
  For the past few seasons the home field has been strongle over estimated.  Maybe there was
a little over compensation as this year we were slighted underestimating the home team.
Or maybe things just moved back to being approximately typical.


 Winner 2004: Tom Benson Gridmarks
 Winner 2003: CPA Rankings
 Winner 2002: StatFox.com
 Winner 2001: Monte Carlo Markov Chain
 Winner 2000: Flyman Performance Ratings
 Winner 1999: Flyman Performance Ratings



  MOST ACCURATE SYSTEM - SMALLEST MEAN SQUARE ERROR (Second Half)


Winner:  Computer Adjusted Line

  The most accurate predictor for the second half was my computer adjusted line.
To create this line I start with the Vegas line and nudge the line up or down
depending on what percentage of the systems would place a bet on a given side.
My hope was to come up with something that would improve on the line's already
good accuracy.  I am fairly happy with the results.  It did pick up this award
for second half accuracy, and I only started it mid season.
The computer adjusted line had a mean square error of 166.638 which was just
a fraction higher than the line, 166.835.  Least Squares was the closest system.
This was the best mark seen since Pythagorean set the record of 165.954 in 2000.


 Winner 2004:  Computer Adjusted Line
 Winner 2003:  Viacheslav Ugolnikov
 Winner 2002:  FreeSportsPlay.com
 Winner 2001:  Hank's Power Ratings
 Winner 2000:  Pythagorean Ratings



   BEST PREDICTIVE SYSTEM (Entire Season)

Winner: Least Absolute Value Regression

  For best overall awards I gave points to a system in each
category based on how well they do.  I then sum the points.  Hopefully this
gives an accurate read on who is the best system across the board.
  This year there was a fairly dominant winner in Least Absolute Valuue Regression.
The closest competitor was Pigskin Index and then the system median.
This makes five different winners in five seasons.


 Winner 2004:  Least Absolute Value Regression
 Winner 2003:  Beck Elo
 Winner 2002:  Matthews Grid
 Winner 2001:  CPA Rankings
 Winner 2000:  Ed Kambour



  BEST PREDICTIVE SYSTEM (Second Half)

Winner:  Stewart Huckaby

 There always has to be one surprise.  ARGH didn't have the greatest win/loss record this
season but being very strong in one category and above average in the rest of the categories
was enough to win this award by 1 point over the system median and 2 points over Mike Greenfield
This also makes five winners in five seasons for the second half as well.


 Winner 2004:  ARGH Power Ratings
 Winner 2003:  CPA Rankings
 Winner 2002:  Kenneth Massey
 Winner 2001:  Monte Carlo Markov Chain
 WInner 2000:  Pythagorean Ratings





                RETRO AWARDS 

  The retrodictive categories are based on each systems final
standings and measure things by applying those final ratings
to the entire season in retrospect.

As I did in the college awards I am going to break this down
into big systems (estimate team specific home field advantage)
and simple systems (do not estimate team specific home field advantage).




  MOST RETRODICTIVE WINS - Large System 

Winner:  The Sports Report

 The best retrodiction record goes to The Sports Report SLOTS system with
a record of 236-31 (88.396)  This  beats the previous record of 83.15%
set by SLOTS last season.  The closest competitor finished 31 games behind
the SLOTS system.


 Winner 2004: The Sports Report SLOTS
 Winner 2003: The Sports Report SLOTS
 Winner 2002: The Sports Report SLOTS
 Winner 2001: The system average
 Winner 2000: CPA Rankings



  MOST RETRODICTIVE WINS - Simple Ssystem 

Winner:  Brian Gabrielle

 The best retrodiction record for those not fitting a team home field
advantage came from Brian Gabrielle, 197-70, 73.28%  This is 41 games worse than
the large system winner.  This is down Alder's 76% last season.


 Winner 2004: Brian Gabrielle
 Winner 2003: Frank Alder





  SMALLEST RETRODICTIVE ABSOLUTE ERROR - Large System

Winner: Least Squares with team home field advantages

  Least squares using team specific home field advantage wins this award
for the fifth straight year.  It has a winning average error of 8.51 points.
This sets a new record.  The previous record was 8.76 set in 2002.



 Winner 2004: Least squares with team specific home field advantage
 Winner 2003: Least squares with team specific home field advantage
 Winner 2002: Least squares with team specific home field advantage
 Winner 2001: Least squares with team specific home field advantage
 Winner 2000: Least squares with team specific home field advantage





  SMALLEST RETRODICTIVE ABSOLUTE ERROR - Simple System

Winner: Sagarin Predictive/Points


Sagarin Points wins this category for the second consecutive season. Sagarin's error
of 9.04 was good for 5th place overall and is an improvement over his markof 9.72 last
season.


 Winner 2004: Sagarin Points
 Winner 2003: Sagarin Points





  SMALLEST RETRODICTIVE BIAS

Winner: CPA Rankings, Steve Wrathell

  For the bias category I have not been splitting into large or small
systems.  The smallest retro bias this year was observed in Steve Wrathell's
CPA Rankings, with a bias of 0.006.   The numbers are not terrible
here this year.  On average the bias tended to be about 0.4 points to the
home team.  The retrodiction awards are looking a lot like last season.

 Winner 2004: CPA Rankings
 Winner 2003: CPA Rankings
 Winner 2002: PerformanZ Ratings
 Winner 2001: Herman Matthews
 Winner 2000: Massey Ratings




 MOST ACCURATE SYSTEM - SMALLEST MEAN SQUARE ERROR - Large system


Winner: Least Squares with team home field advantages

  As with absolute error, Least squares with team specific home field advantage
takes this award for the firsth straight season.  The winning mean square
error was 112.865.  This break the previous record of 126.085 set in 2002.
CPA Rankings was a reasonably close runner up at 138.824.
Least squares with team specific HFA dominated the mean error and mean square
error categories this season.



 Winner 2004: Least squares with team specific home field advantage
 Winner 2003: Least squares with team specific home field advantage
 Winner 2002: Least squares with team specific home field advantage
 Winner 2001: Least squares with team specific home field advantage
 Winner 2000: Least squares with team specific home field advantage



 MOST ACCURATE SYSTEM - SMALLEST MEAN SQUARE ERROR - Simple system

Winner: Least Squares Regression


The top simple system was Least Squares Regression (L2) which was good for 5th
place overall.  L2's mean square error was 132.830.  Which better the mark of
Sagarin Points from last year.  Sagarin Points was only about a point behind
this year.


 Winner 2004: Least Square Regression
 Winner 2003: Sagarin points





      BEST OVERALL RETRODICTIVE SYSTEM IN 2004  -  Large System

Winner: Least Squares with team home field advantages

 As with the other overall awards, points are
assigned to systems in each category then summed up.  The system with the
most points wins.  And since this is a "of the year" award I do think
participation needs to be considered.
  Based on total points TSR SLOTS would have the most points, but with
predictions in only 25% of the games I don't feel it right to be considered
the best of the year.  So the retrodiction system of the year goes to
Least Square Regression with team HFA.



 Winner 2004: Least Squares with team home field advantages
 Winner 2003: TSR SLOTS
 Winner 2002: Least Squares with team home field advantages
 Winner 2001: CPA Rankings
 Winner 2000: CPA Rankings



  BEST OVERALL RETRODICTIVE SYSTEM IN 2003  -  Small System

Winner: Sagarin Predictive/Points

  Among the simple systems I give the overall title to Sagarin Points.  His system was
good for 5th place overall but it was the highest among the simple systems.  The system
closest behind was the combined Sagarin rating.

Winner 2004:  Sagarin Points
Winner 2003:  Frank Alder