clock menu more-arrow no yes

Filed under:

National Championship Game Statistical Predicted Margin of Victory

New, 163 comments

Model shows Noles should roll

Kim Klement-USA TODAY Sports

Simply, the margin of victory is how many points one team is better (worse) than the other. Brian Fremeau has done great work on identifying a more competitive margin of victory that excludes "first-half clock kills and end-of-game garbage drives and scores."

Below are game splits from FSU's and Auburn's games this season against FBS opponents (12 games). I've ordered the opponents each team faced by their current F/+ overall ranking in descending order. Also included is the non-garbage time final score and the margin of victory.

Auburn
Opponent F/+ Rank NG Final NG MOV
Bama 2 34-28 +6
Mizzou 17 59-42 +17
LSU 18 21-35 -14
UGA 20 43-38 +5
aTm 22 45-41 +4
Ole Miss 35 30-22 +8
MSU 43 24-20 +4
WSU 58 31-24 +7
FAU 73 45-10 +35
Tenn. 75 48-23 +25
Ark. 87 35-17 +18
Ark. St. 93 31-9 +22

Some table takeways:

  • Auburn's average opponent ranking was 45th.
  • In all of Auburn's games, their average margin of victory was +11 points.
  • In Auburn's four games against their four best opponents (~14th rank), their average margin of victory was +4 points.
  • In Auburn's four games against their four worst opponents (~82nd rank), their average margin of victory was +25 points.
FSU
Opponent F/+ Rank NG Final NG MOV
Clemson 15 41-7 +34
Miami 29 35-14 +21
Duke 36 38-0 +38
UF 46 34-7 +27
BC 54 48-34 +14
Pitt 59 41-13 +28
UMd 65 42-0 +42
Syracuse 76 45-0 +45
Wake 79 52-0 +52
Nevada 85 45-7 +38
NCSU 89 42-10 +32
Idaho 116 56-7 +49

The table takeaways:

  • FSU's average opponent ranking was 62nd.
  • In all of FSU games, their average margin of victory was +35 points.
  • In FSU's four games against their four best opponents (~32nd rank), their average margin of victory was +30 points.
  • In FSU's four games against their four worst opponents (~92nd rank), their average margin of victory was +43 points.

We could compare Auburn's MOV vs. Mizzou to FSU's over Clemson, as they are similar opponents (17th & 15th, respectively). The difference suggests that FSU is +17 points better than Auburn (i.e., +34 minus +17).

The disparity in the margins of victory is striking, with FSU +24 points better than Auburn's. However, we also see the disparity in the quality of opponent faced, with Auburn' schedule on average 17 ranks tougher than FSU's.

Can we estimate the average margin of victory for each team conditional on opponent ranking? Below is a plot for both Auburn and FSU's opponent ranking and their respective margins of victory.

24ood1v_medium

It turns out that both teams' MOV are linearly related to the quality of opponents faced. The relationship is represented by the blue line fitting each set of points for each team. Statistically, the relationship is significant for Auburn (p<0.02) and marginally significant for FSU (p<0.07). But the trend is consistent and coherent: Each team's MOV is larger (smaller) when playing worse (better) teams. The gray area is the 95% confidence interval around the mean (i.e., the blue line).

From this simple model, we can issue predictions.

Auburn is the 4th-ranked F/+ team. According to FSU's regression model, FSU's predicted MOV over Auburn is +22 points, a sound victory for the Noles. Using the 95% confidence interval, the upper range on that is +49 points -- a blowout win -- and the lower range is -4 points -- a close loss.

FSU is the 1st-ranked F/+ team. According to Auburn's regression model, Auburn's predicted MOV over FSU is -1 points, a narrow loss to FSU. The upper range for them has Auburn winning by +24 points while the lower range would net them a loss of -26 points.

Visibly, the models are different. This is because of the sample size problem that both infuriates and intrigues college football purveyors. There are 125 teams, and these teams have played 13 of them with no common opponents.

However, the model predictions themselves can be compared. Let's can compare the models directly by subtracting Auburn's predicted values from FSU's to show the net difference in terms of FSU's MOV.

Prediction Comparison
Paramater FSU Auburn Net Difference
Upper-bound +49 +24 +25
Mean +22 -1 +23
Lower-bound -4 -26 +21

After comparing the models relative to each team's performance vs. their respective opponents, the prediction comparison yields an FSU victory by 21-25 points -- a sound victory.

Acknowledgements: The analysis was carried out using R: A Language and Environment for Statistical Computing. The graphic was produced using ggplot2.