Bouncing back from last week’s performance, the model flirted with a new all-time high for Week 5, correctly picking 83.7% of the winning teams (read more on our process here if you haven’t already). While we didn’t quite finish in that rare air, it did go down at the second best Week 5 on record. As you can see in the chart below, our average performance during this week of the season was around 78%.
We’ll continue to track the model’s performance each week and hopefully we will back it up with another week in the 80+% range.
Below, we’ll slice and dice the data from Week 5 to highlight some of the most interesting results.
Top 10 Week 5 Matchups
Sorted on top combined elos + expected margin of victory of two touchdowns or less
Top 10 Biggest Upsets
Sorted by largest pre-game win probabilities of losing teams. For example, Katy Jordan had a 97% probability to lose this week, but won by four touchdowns. As a result, Jordan gained 356 Elo for outperforming expectations.
Week 5 Top 10 Largest Elo Points Added
Week 5 Top 10 Expected winners who outperformed
Filtered on expected win percentage > 50 %, sorted by most elo added
These are teams that came into the week expected to win their respective games, but still surprised to the upside by well outperforming their expected margin of victory (MOV).
ALL Week 5 Results
All game results below, including model performance, plus Elo rating changes (in green next to team name/record) from pre-game to post-game for every team.