Improved Linear Model
This model was trained to match/predict the teams that are selected by the College Football Playoff (CFP) committee to compete for the NCAA National Championship in football. Each of the links for the weeks below as well as the links for pre-bowl related CFP standings (and post-bowl, poll-related comparisons) contain the quantities determined by the power rating system (as described in the book The Hidden Game of Football, by Carroll, Palmer and Thorn, 1988) that are used to make these predictions (by this model). The OD column represents the average difference between how many points a team’s offense has scored and how many points their defense has surrendered. The SOS column represents that strength of schedule that said team earned against its opponents that year. These quantities are determined using the full margin of victory (MOV) with all games played against Football Bowl Subdivision (FBS) opponents that year. The two columns ending in Z are these same quantities where the largest allowed MOV is 1 point, essentially capturing only the Win-Loss behavior of that particular season’s games. (Games against non-FBS opponents are included when MOV is ‘ignored’; all of those opponents are grouped together under one generic, team name.)
The multiplying weights used to derive the quantity in the Rating column, in all the files linked to below, are as follows: 0.30912775 * OD, 0.83784781 * SOS, 85.99451009 * ODZ, 49.28798644 * SOSZ and 0.44385664 * # of losses that year (where this last quantity is subtracted from the sum of the four other, aforementioned products). You can read more about this model here.
CFP Era: 2014 to the present …
The Improved Linear Model (ILM) had correctly selected 22 of the 24 top four teams chosen by the CFP committee during the first six years of the CFP: 2014-2019. (The power rating system, when ignoring MOV, correctly selected 20, and when using the full MOV, the same system correctly selected 16.) During 2020, with the pandemic impacting team’s schedules, etc., the ILP was still able to correctly identify 3 of the 4 teams that were invited to the CFP. In 2021 and 2022, the ILM agreed with the committee on all four teams that were invited to compete for the National Championship, including FSU in 2023 – so only three out of the four invited teams were correctly matched that year, raising its performance to correctly selecting 36 of the 40 teams chosen to compete in the College Football Playoff. (The power rating system, working with the full MOV, had 2 CFP selected teams appearing in its top four in 2020, 3 in 2021, and 2 in 2022, and just 1 in 2023 raising its correct predictions to 24. With regards to the power rating system when it ignores MOV, it correctly chose 3 of the 4 teams in 2020 and 2021; however, it had all four invited teams in 2022 AND in 2023 (excluding undefeated FSU!), raising its total to 34 – out of the selected 40 over the last 10 years.)
2023: Weeks #9, #10, #11, #12, #13, Pre-Bowls, Post-Bowls
2022: Weeks #9, #10, #11, #12, #13, Pre-Bowls, Post-Bowls
2021: Weeks #9, #10, #11, #12, #13 , Pre-Bowls, Post-Bowls
2020: Weeks #13, #14, #15, #16, Pre-Bowls, Post-Bowls (Caveats about 2020.)
2019: Weeks #10, #11, #12, #13, #14, Pre-Bowls, Post-Bowls
2018: Weeks #9, #10, #11, #12, #13, Pre-Bowls, Post-Bowls
2017: Weeks #9, #10, #11, #12, #13, Pre-Bowls, Post-Bowls
2016: Weeks #9, #10, #11, #12, #13, Pre-Bowls, Post-Bowls
2015: Weeks #9, #10, #11, #12, #13, Pre-Bowls, Post-Bowls
2014: Weeks #10, #11, #12, #13, #14, #15, Pre-Bowls, Post-Bowls
BCS Era: 1998-2013 (16 years)
The Improved Linear Model matched 26 of the 32 top two teams chosen to compete for the NCAA National Championship (in football) during the 16 BCS years; the power rating system matched (a different set of) 19 teams (both with and without MOV).
2013: Pre-Bowls, Post-Bowls
2012: Pre-Bowls, Post-Bowls
2011: Pre-Bowls, Post-Bowls
2010: Pre-Bowls, Post-Bowls (35 post season bowl games in all this year)
2009: Pre-Bowls, Post-Bowls
2008: Pre-Bowls, Post-Bowls
2007: Pre-Bowls, Post-Bowls
2006: Pre-Bowls, Post-Bowls
2005: Pre-Bowls, Post-Bowls
2004: Pre-Bowls, Post-Bowls
2003: Pre-Bowls, Post-Bowls
2002: Pre-Bowls, Post-Bowls
2001: Pre-Bowls, Post-Bowls
2000: Pre-Bowls, Post-Bowls (25 bowl games this year)
1999: Pre-Bowls, Post-Bowls
1998: Pre-Bowls, Post-Bowls
Poll-Based Years: 1965 up to 1997
1997: Pre-Bowls, Post-Bowls
1996: Pre-Bowls, Post-Bowls
1995: Pre-Bowls, Post-Bowls
1994: Pre-Bowls, Post-Bowls
1993: Pre-Bowls, Post-Bowls
1992: Pre-Bowls, Post-Bowls
1991: Pre-Bowls, Post-Bowls
1990: Pre-Bowls, Post-Bowls (19 bowl games)
1989: Pre-Bowls, Post-Bowls
1988: Pre-Bowls, Post-Bowls
1987: Pre-Bowls, Post-Bowls
1986: Pre-Bowls, Post-Bowls
1985: Pre-Bowls, Post-Bowls
1984: Pre-Bowls, Post-Bowls
1983: Pre-Bowls, Post-Bowls
1982: Pre-Bowls, Post-Bowls
A major restructuring of which teams were designated as Division 1 occurred in 1982. Before that, the 8 college teams in the Ivy League, and another 17 colleges, were considered Division 1. Various teams were re-categorized between 1965 and 1981 (roughly another 16 teams), but that was usually only a few of teams in certain years.
1981: Pre-Bowls, Post-Bowls
1980: Pre-Bowls, Post-Bowls (15 bowl games)
1979: Pre-Bowls, Post-Bowls
1978: Pre-Bowls, Post-Bowls
1977: Pre-Bowls, Post-Bowls
1976: Pre-Bowls, Post-Bowls
1975: Pre-Bowls, Post-Bowls
1974: Pre-Bowls, Post-Bowls
1973: Pre-Bowls, Post-Bowls
1972: Pre-Bowls, Post-Bowls
1971: Pre-Bowls, Post-Bowls
1970: Pre-Bowls, Post-Bowls (11 bowl games)
1969: Pre-Bowls, Post-Bowls
1968: Pre-Bowls, Post-Bowls
1967: Pre-Bowls, Post-Bowls
1966: Pre-Bowls, Post-Bowls
1965: Pre-Bowls, Post-Bowls
1964: Pre-Bowls, Post-Bowls
1963: Pre-Bowls, Post-Bowls
1962: Pre-Bowls, Post-Bowls
1961: Pre-Bowls, Post-Bowls
1960: Pre-Bowls, Post-Bowls (9 bowl games)
1959: Pre-Bowls, Post-Bowls
1958: Pre-Bowls, Post-Bowls
1957: Pre-Bowls, Post-Bowls
1956: Pre-Bowls, Post-Bowls
1955: Pre-Bowls, Post-Bowls
1954: Pre-Bowls, Post-Bowls
1953: Pre-Bowls, Post-Bowls
1952: Pre-Bowls, Post-Bowls
1951: Pre-Bowls, Post-Bowls
1950: Pre-Bowls, Post-Bowls (8 bowl games)