College Football Notebook 2017

College Football Notebook 2017, a preseason analytics and predictions e-guide produced by McIllece Sports, is now available! Since the file is a 288-page PDF, it is recommended to right click and select “save as” to download a copy to your hard drive.

Link: College Football Notebook 2017

Preseason Analytics

  • Historical power and recruiting ratings
  • Rankings, records, and conference projections
  • Score predictions and win probabilities for every game
  • 2016 unit ratings and 2017 player weights
  • Unique stats and schedule ranks

In-Season Notebook

  • Track scores and injuries
  • Update strength of schedule and power ratings
  • Create custom score predictions for regular season, conference championship, and bowl games
  • Evaluate team trends with plus/minus ratings

Don’t agree with the Notebook’s ratings or predictions? Adjust them to create your own power ratings and score predictions for your favorite teams and biggest rivals! College Football Notebook 2017 puts the power of the analyst in your hands. Download and print your free copy today!

Note about rosters

Rosters are always in flux. Significant roster changes and injury information can be tracked in the workbook space at the end of each conference section. The roster information included in the 2017 Notebook was based on official team sites as of mid to late May and reflects the information used to make all 2017 McIllece Sports predictions.

While not exactly a roster change, the one exception to this rule was the retirement of Bob Stoops at Oklahoma. The 2017 Notebook does show Lincoln Riley as the new head coach. All other figures and information about the Sooners, however, were based on the assumption of Coach Stoops leading them in 2017.

2017 modifications

The following modifications were implemented with the release of the 2017 Notebook:

  • Projected offense, defense, and power ratings listed in +/- workbook area, making calculations quicker and easier
  • Expected win totals already populated (based on preseason win probabilities), making +/- wins calculations quicker and easier
  • Team-specific SOS already populated in the optional formula for updating SOS
  • Setting G=12 as the maximum for power ratings updates is noted in the workbook space, correcting an error for games 13+ from last season
  • “Talent Reload” renamed “Skill Talent Reload” to emphasize that this metric is specific to offensive skill players
  • “Coaching / Transition” redefined and renamed “Program Foundation”, reflecting the strength and continuity of the program over the past two years (teams with new head coaches tend to get lower values in this category)
  • For national “Schedule Advantage” rankings, BYU is no longer grouped with the Power 5 teams, due to playing a schedule more comparable to Group of 5 teams in 2017
  • Coastal Carolina (transitioning from FCS) and UAB (returning after two years of no competition) added
  • Injury report page at the end of each page is no longer separated by team, making it easier to note significant roster changes in the conference
  • Special teams changed from equal weights per category to predictive weights (more emphasis on kickers than returners)

Note about calculating “Game +/- Ratings”

In the right-hand page workbook space, the column for “Game +/- Ratings” describes the required calculation as “Results – Predictions”. It is important to use the preseason predictions in this calculation. Using updated weekly predictions here will not fully account for the difference between the preseason ratings and the in-season results.

Email analytics@mcillecesports.com with any questions about filling out the Notebook pages.

Power, Offense, and Defense Ratings

These are the three primary ratings that measure the quality (or predicted quality) of a team, in terms of points scored and points allowed. They are all schedule-adjusted, meaning that the quality of opposition faced is factored into the calculations.

  • Offense = The points scored value of a team’s offense (high is good)
  • Defense = The points allowed value of a team’s defense (low is good)
  • Power = Offense – Defense. Conceptually, this is the expected margin of victory (or defeat, if negative) versus an average FBS opponent on a neutral field. An average FBS team has a power rating of zero.

Therefore, for a simple estimate of how many points Team1 would score against Team2, add the Offense rating of Team1 to the Defense rating of Team2. This would be equal to the expected Points Scored (PS) for Team1. Analogously, to estimate how many points Team2 would score in that same game against Team1, add Team2’s Offense to Team1’s Defense.

Example

In 2015 Week 1, Alabama played Wisconsin at a neutral site. The final 2015 Offense and Defense ratings for these teams were:

Alabama Offense = 26.8
Alabama Defense = -3.1

Wisconsin Offense = 14.2
Wisconsin Defense = -0.2

Alabama (Expected) Points Scored = 26.8 + (-0.2) = 26.6 ≈ 27
Wisconsin (Expected) Points Scored = 14.2 + (-3.1) = 11.1 ≈ 11
Alabama (Expected) Margin of Victory = 27 – 11 = 16

Since 11 is an unusual point total, a reasonable score expectation for this matchup might be Alabama 27, Wisconsin 10. The final score in that game was 35 – 17, an 18 point Alabama win. Both teams basically scored an extra TD over the expected score line.

The right-hand notebook page for each team includes formulas that refine this a bit (using 0.984 as a multiplier and factoring in home-field advantage to give the home team a boost), but the concept is the same.

Archived Notebooks
College Football Notebook 2016

 

2017 Top 25 and Conference Predictions

McIllece Sports #1 over last three years

Over the past three years, McIllece Sports has posted the most accurate predictions in the country according to the Stassen Poll, long the standard of measuring college football prediction accuracy, outperforming outlets such as Phil Steele, ESPN, Lindy’s, Athlon, USA Today, The Sporting News, and the preseason media polls.

Additionally, since the beginning of the playoff era, McIllece Sports has correctly picked more College Football Playoff teams than anyone else in the country. Since the CFP began in 2014, 8 out of 12 teams we’ve listed in our preseason top 4 have made the College Football Playoff, more than any of the 20+ outlets annually included in the Stassen Poll. As a comparison, famed prognosticator Phil Steele has correctly picked 5 out of 12 playoff teams, illustrating just how difficult it can be to identify the final four in any given season.

A new way to review prediction accuracy

In the 2016 College Football Notebook, McIllece Sports published win-loss probabilities for every team, something no other outlet in the Stassen Poll provides. (ESPN, with their FPI system, produces projected win totals and conference championship probabilities.) With probabilities assigned to every possible team record, the ultimate test of accuracy can be evaluated: calculating the error in the preseason win prediction, then comparing the actual error distribution to the distribution expected by the preseason win-loss probabilities. If the expected distribution and the actual distribution closely match, then it is shown that the win-loss probabilities were accurate.

The 2016 comparison demonstrates that our win-loss probabilities were extraordinarily precise:

More than half the FBS teams were projected to within one win, and over three-fourths of teams were projected to within two wins. As a comparison, ESPN (the only other Stassen outlet to evaluate its 2016 win projections) was “within one win…for more than a third of FBS teams” and “within two wins for nearly two-thirds of teams.” The McIllece win projections were much more accurate, although it should be noted that ESPN included the eight conference championship games in their calculations, slightly altering the basis of comparison from 826 games to 834 games.

Most importantly, the expected number of teams to fall within certain error levels and the actual season results are nearly identical. In other words, the season record probabilities provided in the College Football Notebook are numbers you can trust, and you won’t find a breakdown as comprehensive anywhere else. The 2017 College Football Notebook is expected to be published by the end of June, so be sure to check back in a few weeks.

2017 Predictions

The McIllece Sports top 30 and divisional predictions are posted below. These predictions can also be found (without preceding text) on the Season Predictions page.

In the divisional standings tables, AvgRank is the team’s average finish from 100,000 full season simulations, and Championship % is how often they won they division in those simulations. The predicted order of finish in each division is based on AvgRank. Power 5 conferences are listed first.