[Skip to main content](#content)
In the current state of college football, the objective is to make the College Football Playoff. And unlike almost every other sport, playoff spots are decided by arbitrary rankings and human judgment, not by win/loss records or head-to-head results as in the NFL, NBA, or MLB. This is especially true for Group of Six teams trying to crash the party.One of the arguments made against G6 hopefuls is: “Yes, you’re undefeated or have one loss — but who have you played?” It’s the argument against Liberty, James Madison, and North Dakota State. Meanwhile, Texas loses to unranked Oklahoma, Alabama loses to unranked Vanderbilt and Oklahoma, and Ole Miss loses to unranked Kentucky and Florida and they all stay in the top 10.So how much does schedule strength actually drive the numbers the committee leans on? This week on Stats Corner, we examine the relationship between strength of schedule (SOS) and ESPN’s FPI (Football Power Index), using data from the 12 Mountain West teams across the 2022, 2023, 2024, and 2025 seasons. The tables of complete data can be found at the end.
**The Metrics**
For readers new to these terms, here’s what each one tells usSOR Strength of Schedule Rank: Reflects chance that an average Top 25 team would have team’s record or better, given the schedule.FPI Football Power Index: A model-based estimate of overall team strengthSOS Strength of Schedule:Rank among all FBS teams of games already played schedule strength, from perspective of an average Top 25 team.GC Game Control Rank: Reflects chance that an average Top 25 team would control games from start to end the way this team did, given the schedule.
AVGWP Average Win Percentage: Team’s average in-game win probability rank adjusted for chance that an average FBS team would control games from start to end the way this team did, given the schedule.
GC
0.94
0.91
0.82
0.87
0.89
AVGWP
0.92
0.92
0.77
0.76
0.84
SOR
0.94
0.78
0.74
0.9
0.84
SOS
\-0.41
0.17
0.45
\-0.08
0.03
The framework for interpreting the correlations:
**Overachievers:** strong SOR but weaker FPI, GC, and AVGWP good résumé, mediocre underlying strength.
**Underachievers:** weak SOR but stronger FPI, GC, and AVGWP strong team, unimpressive résumé.
**Consensus contenders:** teams that rank well across most or all metrics.Rank 1 means best. One important caveat before the results: each dataset contains only 12 Mountain West teams. These are descriptive correlations, not proof that one metric predicts future performance better than another, and the differences could partly reflect the particular teams in each sample. With that said, the patterns are worth your attention.
**The Four-Year Results**
For each season, I calculated Spearman rank correlations between FPI rankings and the other four metrics (average ranks used for tied values; correlations rounded to two decimal places).
**Game Control Is the Most Consistent Correlate of FPI**GC has the highest average correlation across all four years, at approximately +0.89, with annual values ranging from +0.82 to +0.94 strong in every single dataset. It was never the weakest correlate in any year, and it was the strongest in 2024.This suggests that teams rated highly by FPI also tend to control their games leading, or at least staying ahead in win probability, for long stretches. GC may capture something important about consistent team performance that win-loss records alone don’t reflect. If the goal is to understand which measures align most closely with FPI, GC is the strongest candidate for further investigation.
**AVGWP Is Consistently Strong but Weakened Recently**
AVGWP correlated with FPI at +0.92 in both 2022 and 2023, before falling to +0.77 in 2024 and +0.76 in 2025. That’s a noticeable slide, although the relationship remains strong throughout.One possible explanation: average in-game win probability and FPI capture somewhat different aspects of performance. AVGWP reflects how often a team is winning during games; FPI estimates overall team strength. Teams can dominate in-game situations without earning equally strong overall ratings and the data alone cannot establish why the relationship weakened. Interestingly, AVGWP is closely related to GC (they measure similar things), so part of the decline may simply be year-to-year noise in a 12-team sample.
**SOR Is Strong, But Variable**
SOR’s correlations range from +0.74 to +0.94: very close alignment with FPI in 2022 (+0.94) and 2025 (+0.90), but weaker in 2023 (+0.78) and 2024 (+0.74). It’s the only metric whose correlation swung more than 0.15 between consecutive years.
This variability makes conceptual sense. Strength of Record evaluates _accomplishments against a schedule_, while FPI estimates _underlying team strength_. Those two things often agree, but not always, and when they don’t, you get exactly the overachiever/underachiever divide described above. SOR is arguably the most _interesting_ metric on the list precisely because its disagreement with FPI is meaningful: it flags teams whose résumés outperform (or underperform) their actual strength. That’s the gap the playoff committee is asked to judge every December.
**SOS Behaves Like No Other Metric**
Here’s where it gets interesting and where the playoff argument comes home. SOS’s four-year average correlation with FPI is essentially zero (+0.03), but that average hides wild swings:
2022
\-0.41
2023
0.17
2024
0.45
2025
\-0.08
The relationship flipped sign twice in four years. In 2022 and 2025, teams with better FPI rankings actually tended to have _easier_ schedules (remember, lower SOS rankings mean tougher schedules, so a negative correlation means strong teams faced weaker opposition). In 2024, the relationship was moderately positive.The takeaway is not that schedule strength is unimportant. It’s that within a 12-team mid major conference sample, schedule difficulty doesn’t consistently track team quality, a hard schedule doesn’t make you good, and a good team doesn’t necessarily play a hard one. SOS is best used as context for interpreting the other metrics, not as a standalone measure of team strength.
**The Playoff Implication**
Now back to the argument that opened this piece. The committee’s “who have you played?” critique is an SOS argument and in four years of Mountain West data, SOS is the metric with the _least_ consistent relationship to team strength. Meanwhile, the metrics that actually track FPI closely are Game Control and Average Win Probability: measures of how thoroughly a team beats the opponents in front of it, not who those opponents happen to be.
That doesn’t prove the committee is wrong an undefeated G6 team genuinely has less information for evaluators to work with than a one-loss or two-loss SEC team (they wouldn’t let a thre3-loss team in would they?) But it does suggest the hierarchy is backwards: the quality of _performance_ (GC, AVGWP, SOR) tracks team strength far more reliably than the quality of _opposition_ (SOS). When a 12-0 team is left out because its SOS ranking was mediocre, the data says we’re weighting the noisiest number on the sheet.
**Important Limitations**Before treating these findings as definitive, three issues need to be considered:
**Small sample size:** Each year includes only 12 Mountain West teams. A single unusual season, a single dominant or terrible team, can substantially move these correlations.
**Related metrics:** GC, AVGWP, and FPI may capture overlapping aspects of team performance. Strong correlations do not demonstrate that one metric adds independent predictive value beyond the others.
**Correlation isn’t causation and it isn’t prediction:** These results describe how closely rankings align _within_ the same datasets. They don’t tell us which metric best predicts future wins.
**The Bottom Line**
Across four seasons, the hierarchy is clear:
GC
0.89
Consistent game control
AVGWP
0.84
Strong in-game winning position
SOR
0.84
Strong résumé relative to opponents
SOS
0.03
Schedule difficulty alone doesn’t track FPI
One final distinction worth making: a team can’t directly optimize its FPI _ranking_, because FPI is relative to the rest of the FBS. Even substantial improvement might produce little movement if other teams improve too. The practical goal is to improve the underlying qualities FPI evaluates opponent-adjusted offensive and defensive efficiency, special teams, and roster depth that holds up against quality opponents. Rankings follow strength; strength doesn’t follow rankings.
For G6 teams hoping to crash the playoff party, the message from four years of data is this: dominate the schedule you have. The committee asks who you played, but the numbers say who you beat, and by how much, tells us far more, but then again the committee has the final say.
New Mexico
49
78
91
67
50
Boise State
51
55
69
63
54
San Diego State
61
67
117
65
25
UNLV Rebels
62
69
125
92
64
Fresno State
64
85
122
83
53
Hawai’i
66
87
112
84
48
Utah State
86
80
73
76
79
Wyoming
112
110
103
119
115
Air Force
113
96
123
105
98
Nevada
120
123
90
118
118
San José State
125
119
98
117
109
Colorado State
129
115
89
124
133
Boise State
16
27
67
22
9
UNLV
31
40
87
37
17
Colorado State
71
105
77
75
53
San José State
93
91
119
94
70
Fresno State
102
81
98
67
45
Hawai’i
107
115
118
106
84
Air Force
108
106
117
115
100
New Mexico
109
107
123
122
106
Utah State
112
110
93
102
99
Wyoming
119
113
97
124
125
Nevada
123
103
86
107
108
San Diego State
124
122
109
123
113
Wyoming
46
74
79
59
44
Fresno State
58
79
124
77
40
Air Force
59
63
121
75
37
UNLV
63
73
62
82
52
Boise State
68
47
82
46
33
San José State
76
68
91
56
41
Utah State
101
105
117
120
103
Colorado State
106
112
130
112
87
San Diego State
111
110
102
105
93
Hawai’i
112
123
74
122
122
New Mexico
117
120
103
117
114
Nevada
131
128
109
122
131
Fresno State
43
57
83
52
27
Air Force
49
56
124
35
6
Boise State
54
54
102
65
33
San Diego State
80
88
81
60
34
Wyoming
82
102
103
106
87
San José State
83
89
126
93
64
Utah State
95
109
59
115
108
UNLV
107
106
105
113
104
Colorado State
121
125
71
120
122
Nevada
127
123
113
120
120
Hawai’i
128
129
66
120
116
New Mexico
129
128
97
120
114
All of your communities in one place.