Sports analytics from Laura Albert at the University of Wisconsin-Madison College of Engineering

NCAA football playoff forecasts, week 8

This project is a tool created by a team at the University of Wisconsin-Madison College of Engineering for forecasting the first four team NCAA football playoff using Monte Carlo simulation and data analytics.

I am very excited to present our first NCAA playoff forecasts today.

As mentioned in the methodology section, we consider three variants of the ranking methodology we use in our simulation. Below, we present the results for our models as well as the ensemble model that average across the three models, with 10,000 replications for each model. Let’s start with the ensemble model. Comments on the results are below.

Ensemble model results

The top teams to make the playoff with the fraction of simulations in playoff.

  1. Florida State       : 0.303544
  2. Mississippi State : 0.289211
  3. Mississippi          : 0.260909
  4. Alabama             : 0.174873
  5. Georgia              : 0.163807
  6. Notre Dame       : 0.139638
  7. Minnesota          : 0.138238
  8. Michigan State   : 0.111237
  9. Auburn               : 0.11037
  10. Nebraska           : 0.10457
  11. Oregon              : 0.0891697
  12. Arizona              : 0.0735691
  13. Marshall            : 0.0639723
  14. Utah                  : 0.0629354
  15. Baylor               : 0.0569019
  16. Ohio State         : 0.0568352
  17. Duke                 : 0.0516018
  18. Texas Christian : 0.0446682
  19. Kansas State     : 0.0359679
  20. Louisiana State  : 0.030501
  21. Louisville            : 0.0259342
  22. Arizona State     : 0.0253008
  23. Washington        : 0.0249675
  24. Colorado State   : 0.0240008
  25. Penn State          : 0.0215341

The top teams to just miss the playoff with the fraction of simulations ranked fifth.

  1. Florida State        : 0.0397347
  2. Mississippi State  : 0.0362014
  3. Georgia                : 0.0360347
  4. Notre Dame         : 0.0335011
  5. Minnesota            : 0.0323011
  6. Alabama               : 0.0307678
  7. Mississippi            : 0.0303681
  8. Michigan State     : 0.0277009
  9. Nebraska             : 0.0274343
  10. Marshall               : 0.0270344

The fraction of playoffs with at least one team from each major conference:

  • ACC: 0.495
  • Big12: 0.231
  • B1G: 0.528
  • PAC12: 0.387
  • SEC: 0.845

Comments:

We see that the SEC will almost certainly have a team (probably 2) in the playoff. The other conferences do not look as good. Based on inter-conference play, it looks like teams from the PAC12 looks weaker than it should (there are no undefeated teams) and B1G looks stronger than it should. Likewise, it looks like too many B1G teams will make the playoff in the simulation. Again, this happens because we don’t have humans in the loop at all – our model does not even “see” expert rankings such as the AP Top 25 or the USA Today Coaches polls based on the games played thus far. In the future, we will explore how additional information can be used to “vote” for teams in our model.

After eight weeks of footballs, the team-to-team network created by wins and losses is still pretty sparse. As a result, some of the forecasts made are too bold (I’m looking at you Minnesota!). We will have a more connected network when more games have been played, and with more data will come more refined forecasts.

 

Also keep in mind that the strength of schedule for future games is not equal between teams, and therefore, some teams are downgraded based on tough future matchups. However, there is still a lot of football to play – this should become more apparent a couple of weeks before the end of the season.

 

See you next week! Don’t forget to follow us on twitter @badgerbrackets

.

 

The results from the three models used in the ensemble model are below.

Results for the wins only model.

The top teams to make the playoff with the fraction of simulations in playoff.

  1. Florida State       : 0.5385
  2. Mississippi State               : 0.3947
  3. Georgia                : 0.2706
  4. Mississippi          : 0.251
  5. Alabama              : 0.2139
  6. Michigan State  : 0.2115
  7. Auburn : 0.2039
  8. Oregon : 0.1897
  9. Nebraska             : 0.1839
  10. Minnesota          : 0.1578

The top teams to just miss the playoff with the fraction of simulations ranked fifth.

  1. Auburn : 0.0504
  2. Georgia                : 0.0496
  3. Mississippi State               : 0.0474
  4. Mississippi          : 0.0472
  5. Florida State       : 0.0464

 

Results for the wins and losses model.

The top teams to make the playoff with the fraction of simulations in playoff.

  1. Mississippi          : 0.5022
  2. Mississippi State               : 0.4298
  3. Florida State       : 0.4166
  4. Alabama              : 0.2536
  5. Notre Dame       : 0.2103
  6. Minnesota          : 0.1996
  7. Georgia                : 0.1811
  8. Auburn : 0.1405
  9. Utah      : 0.1359
  10. Oregon : 0.1321

The top teams to just miss the playoff with the fraction of simulations ranked fifth.

  1. Florida State       : 0.0797
  2. Georgia                : 0.0627
  3. Mississippi State               : 0.0504
  4. Alabama              : 0.0479
  5. Marshall               : 0.0477

Results for the full model.

The top teams to make the playoff with the fraction of simulations in playoff.

  1. Mississippi          : 0.5288
  2. Mississippi State               : 0.4726
  3. Florida State       : 0.3742
  4. Alabama              : 0.3115
  5. Notre Dame       : 0.2779
  6. Minnesota          : 0.2605
  7. Georgia                : 0.21
  8. Auburn : 0.1437
  9. Arizona : 0.132
  10. Michigan State  : 0.1301

The top teams to just miss the playoff with the fraction of simulations ranked fifth.

  1. Florida State : 0.0782
  2. Mississippi State  : 0.066
  3. Notre Dame       : 0.0606
  4. Georgia                : 0.0589
  5. Minnesota          : 0.0589