Weekly Predictions

Based on popular request, we will be posting updated game forecasts and tracking the accuracy of the McIllece Sports college football predictions all season long.

Early each week of the CFB season, the detailed table below will be updated with current week predictions for all FBS-only games. Lines, moneylines, and ESPN-FPI win probabilities will be drawn from the data posted at ESPN College Football’s daily lines page:

College Football Daily Lines | ESPN

The betting data will be loaded on Sundays except for Week 2, which will be loaded on Tuesday following the Labor Day Weekend games. In some cases, if lines for specific games aren’t yet posted by Sunday, the betting data will be filled in as soon as it becomes available at the link above.

In a departure from some other accuracy trackers, we’ll be grading two sets of McIllece Sports predictions:

  1. Preseason game forecasts from the 2022 College Football Atlas preview magazine
  2. Updated game projections after incorporating all the year-to-date results of the current CFB season

The Updated projections are a weighted combination of preseason expectations and in-season power ratings.

In Week 0, the Atlas predictions and the Updated predictions will be the same, since no games have yet been played.

Starting in Week 1, Updated predictions will begin to diverge (for those teams that played in Week 0).

By Week 2, the two sets of predictions will be different, as every FBS team will have played a game by then.

If the Lookup Table is not showing on mobile, please switch to desktop view.

2022 Weekly Predictions

Table Columns

  • Game Information: Week, Date, Time (Eastern), Away Team, and Home Team (*denotes a neutral-site game)
  • Point Spread: Point spread for the Away Team from ESPN daily lines page (linked above)
  • Atlas Line: Preseason point spread for the Away Team based on College Football Atlas power ratings
  • Updated Line: Updated point spread for the Away Team based on weighted combination of Atlas preseason power ratings and in-season, year-to-date game results
  • Final Net Score: Final margin of victory/defeat for the Away Team
  • Atlas ATS Pick: Against-the-spread pick based on the Atlas Line
  • Updated ATS Pick: Against-the-spread pick based on the Updated Line
  • Atlas ATS Record: Year-to-date record for the Atlas ATS Picks
  • Updated ATS Record: Year-to-date record for the Updated ATS Picks
  • Away/Home Moneyline (ML): Moneylines for the Away and Home Teams from ESPN daily lines page
  • Atlas Preseason Win%: Preseason Away Team win probability based on College Football Atlas power ratings
  • Updated Win%: Updated Away Team win probability based on weighted combination of Atlas preseason power ratings and in-season, year-to-date game results
  • ESPN-FPI Win%: Away Team win probability from ESPN-FPI prediction system
  • Atlas ML: Preseason moneyline for the Away Team based on College Football Atlas power ratings
  • Atlas ML Pick: Moneyline pick based on the Atlas preseason win probability
  • Updated ML Pick: Moneyline pick based on the Updated win probability
  • ESPN-FPI ML Pick: Moneyline pick implied by the ESPN-FPI win probability (ESPN-FPI does not explicitly list a ML pick)
  • Atlas ML Units: Year-to-date net unit gain/loss based on the Atlas ML Picks (assuming equal-size wagers of one unit)
  • Updated ML Units: Year-to-date net unit gain/loss based on the Updated ML Picks (assuming equal-size wagers of one unit)
  • ESPN-FPI ML Units: Year-to-date net unit gain/loss based on the implied ESPN-FPI ML Picks (assuming equal-size wagers of one unit)
  • Atlas Preseason Bias: Measure of the bias of the Atlas preseason win probabilities, calculated by comparing expected win/loss record to observed win/loss record, reported as a percentage (0% = perfectly unbiased; <0% means favorites’ chances of winning are overstated; >0% means underdogs’ chances of winning are overstated)
  • Updated Bias: Measure of the bias of the updated win probabilities, calculated by comparing expected win/loss record to observed win/loss record
  • ESPN-FPI Bias: Measure of the bias of ESPN-FPI’s win probabilities, calculated by comparing expected win/loss record to observed win/loss record, reported as a percentage
  • FBS-only Games: Number of FBS vs FBS games (with complete odds information) played

Note on Bias

An important feature of a good college football prediction system is the ability to generate unbiased win/loss probabilities. If the probabilities are biased to a significant magnitude, then any analyses based on them will be necessarily flawed, such as the chances of winning the division, winning the conference, or reaching the College Football Playoff.

This is a measurable attribute of the system over the course of the year. To do so, we can compute the expected record of the system’s favored teams versus their observed record during the season. Over several months of college football games, if the expected record and observed record are close, then the prediction system is approximately unbiased. (Some small variations can be explained as random error.) However, if the expected and observed records are far apart, then the prediction system suffers from demonstrable bias.

For example, if 100 college football games were played, each with an 80% win probability for the favored team, those favored teams would have an expected record of 80-20. If the favored teams’ observed record ended up 60-40, then there is clear evidence of bias in the prediction system. In this case, the chances of the favorites winning each game were overestimated, or biased upward.

The bias metrics computed here will allow you to see how ESPN-FPI, the College Football Atlas preseason magazine, and the McIllece Sports’ updated predictions fare in this critical feature all season long.

Early in the season, the bias metrics will likely fluctuate quite a bit due to the relatively small sample size. These will stabilize as more and more games are played.