NFL Favorites Beat Expectations as College Football Favorites Fell Short

Every week the Power Rankings publish an expected margin and a win probability, which is the chance each team wins, for the week's games before kickoff, and the next edition scores them against the results. This review gathers those scorecards for NFL Week 5 and college football Week 6, played Oct. 8 to Oct. 12, and sets each beside the rating's season record against the betting market's closing line.

Across both sports, favorites won 20 of 33 scored games, where the published probabilities implied 21.1. The sections below take each sport in turn, the misses with the hits.

NFL: Week 5 Expectations

Our NFL Week 5 edition, published October 6, 2026, set an expected margin and a win probability before kickoff for every game on the week's schedule, and 15 of those games have final scores. Favorites won 10 of the 15. Their published win probabilities added up to 8.5 expected wins, so favorites won 1.5 more games than those probabilities implied. One game ended in a tie, which counts as a game the favorite did not win. A gap of that size is within what chance alone produces in a week of 15 games, where the probabilities imply a spread of about 1.9 wins either way.

The average distance between a final margin and its expected margin was 10.6 points. The rating's typical error is 13.0 points, which implies an average miss of about 10.4 points, so this week's misses were in line with what the rating claims for itself.

The biggest upset came when the Buffalo Bills beat the Los Angeles Rams, 37 to 34. We had published the Los Angeles Rams as a 63% favorite, so the result was one a forecast at that level expects about 37% of the time.

Across the 46 NFL games of the 2026 season so far with a pre-kickoff rating and a closing line, the rating's log loss is 0.633 against the closing line's 0.658, where lower is better. Its Brier score is 0.221 against 0.233. So far the rating is ahead of the closing line. That is only 46 games, too few to separate skill from luck, so we read it as an early record, not a verdict.

The rating picked the winner in 67.4% of these games and the closing line in 56.5%. Picking winners ignores how confident each forecast was, so we weigh log loss more heavily.

In Week 5 alone, over 14 games with a closing line, the rating's log loss was 0.648 and the closing line's 0.648. That count covers every rated game with a closing line, so it can differ from the scorecard's.

The NFL Week 6 Power Rankings list every scored game, and the NFL ratings page keeps the record game by game.

College Football: Week 6 Expectations

Our college football Week 6 edition, published October 6, 2026, set an expected margin and a win probability before kickoff for the games involving a Top 25 team, and 18 of those games have final scores. Favorites won 10 of the 18. Their published win probabilities added up to 12.6 expected wins, so favorites won 2.6 fewer games than those probabilities implied. A gap of that size is within what chance alone produces in a week of 18 games, where the probabilities imply a spread of about 1.8 wins either way.

The average distance between a final margin and its expected margin was 12.3 points. The rating's typical error is 15.9 points, which implies an average miss of about 12.7 points, so this week's misses were in line with what the rating claims for itself.

The biggest upset came when North Carolina beat Pittsburgh, 21 to 17. We had published Pittsburgh as a 74% favorite, so the result was one a forecast at that level expects about 26% of the time.

Across the 170 college football games of the 2026 season so far with a pre-kickoff rating and a closing line, the rating's log loss is 0.553 against the closing line's 0.522, where lower is better. Its Brier score is 0.189 against 0.175. So far the rating trails the closing line, which remains the better forecast on this evidence.

The rating picked the winner in 72.4% of these games and the closing line in 72.9%. Picking winners ignores how confident each forecast was, so we weigh log loss more heavily.

In Week 6 alone, over 57 games with a closing line, the rating's log loss was 0.618 and the closing line's 0.600. That count covers every rated game with a closing line, so it can differ from the scorecard's.

The College Football Week 7 Power Rankings list every scored game, and the College Football ratings page keeps the record game by game.

What This Does and Does Not Show

These scorecards check whether the probabilities we published were honest about their own uncertainty. One week is a small sample. A favorite at 60% should lose four games in ten, so a week in which favorites fall short of the expected count is not by itself evidence that a rating is wrong, and a week in which they beat it is not evidence that it is right. The college football scorecard covers only the games involving a Top 25 team, not every game.

The season record is the better guide. On log loss, the NFL rating is ahead of the closing line over 46 games (a small sample), and the college football rating is behind the closing line over 170 games. A lower log loss than the closing line over part of a season is not a demonstrated edge over the betting market, and where a rating trails the close we say so as plainly as where it leads. We publish these numbers to show how accurate and how well calibrated the ratings are, not to suggest that they beat the market.