Bassett Football Model

The Bassett football model is a system for ranking the relative strengths of football teams. A key feature of this model is that it recognizes that if two teams could play more than once the same team would not always win (there is an element of chance involved). The system is based on a simple model of a football game. For a given match-up of two teams, a number of model games are simulated by a computer and the average scores and the percentage of times each team wins is computed. Each week during the football season the best fit of offensive and defensive strengths is found for each team consistent with the actual scores. I do include considerations for home field advantage and for the possibility that teams may score less than expected in a blow-out if they choose to run out time instead of score more points in the latter part of a game.

Quick menu:

More about the model ...

Questions and answers ...

The purpose of this football web page is to provide an unbiased analysis of the teams: how they compare with each other and what sort of results to expect from upcoming games. It is not my intention, however, to encourage gambling with this information. I have purposely presented the information in a way that will minimize its usefulness in this regard while still providing useful entertainment, etc. Comments on how to avoid the misuse of these web pages for gambling are most welcome.

Enjoy!

2016 College Football Forecast (NCAA Div IA & IAA)

Bassett Model rankings:

The Stephenson Earned Rank shows what a team has earned from its wins. It ranks by how much better a given team's winning record is when compared against how all other teams would have faired if they had played the same schedule. (Available after the 5th week in the season.) The Power Rankings measure the strength of each team by how well they have played on average, going by point performance. The Current Model Power Rankings are the best indicator of how well a team will play in the future.

10 Jan 2017: Stephenson Earned Rank (NCAA Div I & IAA)
(Team wins subtracted by average number of wins if all other teams played their schedule as simulated by the Bassett Football Model)

10 Jan 2017: Average Model Power Rankings (NCAA Div I & IAA)
(average power ranking for entire season)

10 Jan 2017: Current Model Power Rankings (NCAA Div I & IAA)
(power ranking at final week of the season)

2016 Model Performance: 

All Div I games               forecast probabilities
                 50-59%  60-69%  70-79%  80-89%  90-99%    100%     all
total games:        103     640     178     319     251       0    1491
number right:        56     415     132     250     230       0    1083
number expected:   56.4   420.7   133.9   275.3   232.9     0.0  1119.1
right/expected:    0.99    0.99    0.99    0.91    0.99    0.00    0.97

Breakdown by Conference and Division:
Conference games              forecast probabilities
                 50-59%  60-69%  70-79%  80-89%  90-99%    100%     all
total games:         49     381     122     250     200       0    1002
number right:        25     232      87     194     182       0     720
number expected:   26.6   250.6    92.1   216.0   185.6     0.0   770.9
right/expected:    0.94    0.93    0.94    0.90    0.98    0.00    0.93

Non-Conference games          forecast probabilities
                 50-59%  60-69%  70-79%  80-89%  90-99%    100%     all
total games:         43     187      44      59      43       0     376
number right:        25     127      35      46      40       0     273
number expected:   23.6   122.9    32.9    50.6    39.8     0.0   269.7
right/expected:    1.06    1.03    1.07    0.91    1.01    0.00    1.01

Non-Division games            forecast probabilities
                 50-59%  60-69%  70-79%  80-89%  90-99%    100%     all
total games:         11      72      12      10       8       0     113
number right:         6      56      10      10       8       0      90
number expected:    6.1    47.3     9.0     8.7     7.5     0.0    78.5
right/expected:    0.98    1.19    1.12    1.16    1.07    0.00    1.15

actual/expected = 1 -- model right on
actual/expected > 1 -- model probabilities too low (more wins than expected)
actual/expected < 1 -- model probabilities too high (less wins than expected)

Past Seasons:

2016 NFL Forecast

9 Feb 2017: Average model rankings (NFL) (average rank over entire season)

9 Feb 2017: Current model rankings (NFL) (rank at final week in the season)

2016 Model Performance: 

All games                     forecast probabilities
                 50-59%  60-69%  70-79%  80-89%  90-99%    100%     all
total games:         44     152      33      38       0       0     267
number right:        24      89      22      32       0       0     167
number expected:   24.3    95.0    24.9    32.0     0.0     0.0   176.2
right/expected:    0.99    0.94    0.88    1.00    0.00    0.00    0.95

Breakdown by Division and Conference:
Division games                forecast probabilities
                 50-59%  60-69%  70-79%  80-89%  90-99%    100%     all
total games:         20      52      10      14       0       0      96
number right:        14      28       6      11       0       0      59
number expected:   11.2    32.6     7.4    11.8     0.0     0.0    62.9
right/expected:    1.26    0.86    0.81    0.93    0.00    0.00    0.94

Non-Division games            forecast probabilities
                 50-59%  60-69%  70-79%  80-89%  90-99%    100%     all
total games:         16      62      12      16       0       0     106
number right:         5      42       7      16       0       0      70
number expected:    8.8    38.5     9.1    13.4     0.0     0.0    69.8
right/expected:    0.57    1.09    0.77    1.19    0.00    0.00    1.00

Non-Conference games          forecast probabilities
                 50-59%  60-69%  70-79%  80-89%  90-99%    100%     all
total games:          8      38      11       8       0       0      65
number right:         5      19       9       5       0       0      38
number expected:    4.4    24.0     8.4     6.7     0.0     0.0    43.5
right/expected:    1.13    0.79    1.07    0.75    0.00    0.00    0.87

actual/expected = 1 -- model right on
actual/expected > 1 -- model probabilities too low (more wins than expected)
actual/expected < 1 -- model probabilities too high (less wins than expected)

Past Seasons:

Links of interest

Please email comments or questions to bfm@BassettFootball.net