The Overtake and Feedback (OAF) algorithm "home page"

The page that lists the final rankings/ratings when the Overtake and Feedback algorithm (OAF) is applied to previous college football season's can be found here. This specifics of this algorithm are spelled out in the 2004 February issue of Dr. Dobb's Journal. (A quick summary of the mechanics of this can be found here.) How well OAF has performed, in comparison to the seven computer rating systems (that were part of the Bowl Championship Series (BCS) in 2003) when predicting bowl games, can be found below. (Here are the most recent Ratings using OAF.)

This algorithm was "created" in July and August of 2002, and seems to capture a team's level of success during the year more than measuring its relative strength against the competition, which is probably why it correlates fairly well with the AP and coaches' year end polls. It was designed with that effect in mind in an attempt to quantitatively evaluate who had the best season versus attempting to determine which team is the best. An undefeated team that only defeats weaker competition will not be ranked too highly at the end of the season, e.g. Tulane was 12-0 in 1998, but only finished ranked #25 by the OAF algorithm. In some respect, strength of schedule factors in naturally during the application of the OAF algorithm, even though there is no value directly computed and used in the algorithm that measures how tough a team's opponents were that year...

OAF - Overtake and Feedback                WC - Wes Colley                JS - Jeff Sagarin

A&H - Jeff Anderson and Chris Hester    KM - Kenneth Massey        PW - Peter Wolfe

RB - Richard Billingsley                         LV - Las Vegas Line

  OAF A&H RB WC KM JS PW LV * *
Wins in 2002 (15 games all agreed, a 10-5 record; 28 games total) 18 15 14 16 14 14 15 15 All
Picks
All
Ranks
Wins in 2003 (10 games all agreed, an 8-2 record; 28 games total) 18 14 20 14 19 19 17 19 All
Picks
All
 Ranks
Wins in 2004 (12 games all agreed, a 9-3 record; 28 games total) 18 16 17 18 15 15 12 15 All
Picks
All
Ranks
Wins in 2005 (12 games all agreed, a 7-5 record; 28 games total) 15 14 17 16 15 15 16 17 All
Picks
All
Ranks
Wins in 2006(15 games all agreed, a 12-3 record; 32 games total) 22 21 20 21 22 20 21 23 All
Picks
All
Ranks
Wins in 2007(13 games all agreed, an 8-5 record; 32 games total) 19 21 16 21 18 21 18 24 All
Picks
All
Ranks
Wins in 2008 (20 games all agreed, an 8-12 record; 34 games total) 20 16 15 16 16 17 14 22 All
Picks
All
Ranks
Overall (210 bowl games)
2002 - 2008
130 117 119 122 119 121 103 135    

* Ranks and Picks for the rating system published in the NY Times are included in 2002-2003, since that system was part of the BCS formula through 2004, and it also appeared in Kenneth Massey's Comparison page (where more systems can be found), and the final, pre-bowl top 25 from that paper provided the partial selections in 2004. (It is no longer included in those two sets of pages.) I have also compiled an analysis of all such systems posted on Massey's page, with regards to the bowl games played after the 2002 season to the present.  Those results can be found here. A small summary from those results is posted below. (More about each system might be found at the web page linked to each system's full name that appears near the top of Kenneth's page.) The system developed by Kambour has a 4 game lead over PerformanZ and Coffey's systems. (As seen above, the Las Vegas line also selected 135 winners, and the next highest total after 128 is 125, appearing twice.)

System name as listed

Acronym

Total

2002

2003

2004

2005 2006 2007 2008

on compare.htm

     

Kambour

KAM

135

13

21

16

15 19 28 23

PerformanZ

PFZ

131

15

18

15

19 20 23 21

Coffey

COF

131

14

20

18

16 22 21 18

Dolphin

DOL

130

15

21

17

17 22 21 17

McCormick

MCK

130

16

21

22

16 20 20 15

Overtake & Feedback

OAF

130

18

18

18

15 22 19 20

Dwiggins

DWI

129

15

18

19

21 21 18 17
Claassen CLA 128 16 18 16 17 22 20 19

Bassett

BAS 128 11 18 20 18 18 19 24

Born

BRN

128

14

20

14

17 21 20 22

(In 2003, I incorrectly copied several ratings from Massey's comparison page, affecting one prediction for each of three systems, so I have placed those corrections on a separate error page to not disturb the time stamp on the "All Ranks" pages above, for validation purposes. The "All Picks" pages have the corrected selections.)

Back to Prof. Trono's Home page. (This page last modified January 16, 2009 .)