Let’s now look at how ESPN’s other predictive algorithm, the newly released version of the Football Power Index (FPI), forecasts 2025 for the Tar Heels.
We dove into the results that ESPN’s SP+ predicted for the 2025 North Carolina football season last Friday. Let’s now look at how ESPN’s other predictive algorithm, the newly released version of the Football Power Index (FPI), forecasts 2025 for the Tar Heels. Since our annual prediction podcast is this week, these forecasts may be helpful to have.
“May be” is the correct phrase here, because this entire season – from the preseason perspective – is one giant maybe. In a typical year, it seems that the SP+ and Sagarin ratings are more accurate after about the midway point of the season when these models have actual on-field results, but the FPI holds up slightly better in the preseason. But with all the uncertainty created by a new staff and a substantially remade roster, how well any of these models will be able to predict UNC’s season is yet to be determined.
A couple of notes: the FPI does not rate FCS teams, so the Richmond outcome is a guess, and the asterisks below denote one-score games.
Without further ado …
FPI on UNC
TCU – 5.9-point underdog – 0-1*
@Charlotte – 10.8-point favorite – 1-1
[Richmond – N/A – 2-1]
@UCF – 8.2-point underdog – 2-2
Clemson – 9.4-point underdog – 2-3
@California – 1.2-point underdog – 2-4*
Virginia – 3.8-point favorite – 3-4 – *
@Syracuse – 2.3-point underdog – 3-5*
Stanford – 4.2-point favorite – 4-5*
@Wake Forest – 3.8-point favorite – 5-5*
Duke – 0.4-point underdog – 5-6*
@NC State – 3.7-point underdog – 5-7*
Don’t Shoot the Messenger
As this season gets closer and these types of models begin to proliferate, there is a huge disconnect between the amount of hype North Carolina football is receiving and the nuts and bolts of statistical analysis they generate.
The season, as forecasted above, seems quirky. UNC is forecasted to lose one-score games to TCU, California, Syracuse, Duke, and NC State, and to win one-score games over Virginia and Stanford, going 2-5 in one-score games. Losing three net one-score games is so uncommon as to be considered almost a feature of bad luck.
The probabilities of getting one-score outcomes correct are far less than those with wider spreads. A team favored by 3 points wins 60% of the time, when favored by 7 points they win 75% of the time, by 14 points they win 85% of the time, and 96% of the time if they are favored by 28 points. When over half of the team’s schedule is comprised of one-score games, the probability of getting the overall record correct diminishes greatly. Three of North Carolina’s FPI projected losses are by less than three points.
In any event, a 5-7 prediction for North Carolina is what the FPI’s algorithm kicked out, a game worse than the 6-6 season predicted by the SP+ this season. Stay tuned later this week when the IC staff makes their annual predictions – should be fun.
Category: General Sports
