Zoltar is my computer program that predicts the results of NFL professional football games. Zoltar predicts which team will win in a game, and by how many points. This can be used to make wagers on the Las Vegas point spread. Zoltar does not predict how many points a team will score — just a margin of victory.

Zoltar recommends a bet when the Zoltar predicted margin of victory is significantly different from the Las Vegas point spread (which is essentially the Vegas predicted margin of victory). For example, if the Vegas line is “At Rams -3.5 Bears”, this means the home team Rams is favored by 3.5 points and a bet on them will pay off only if the Rams win by more than that 3.5 point spread.

If Zoltar has an advice threshold of 3 points, and Zoltar thinks the Rams will win by 7 points, then Zoltar would recommend a bet on the favored Rams.

To prepare for the upcoming 2025 NFL season, I analyzed the games from last year, by looking at how cautious or bold Zoltar’s advice should be. Early in the season, Zoltar needs to be very cautious because every team is changed a lot from the previous year. In the middle of the season, Zoltar needs to be bold to take advantage of as many betting opportunities as possible. Then late in the season, Zoltar needs to be cautious because injuries begin to have a big effect. And the last game of the season is a crap shoot because teams that have nothing to gain will often not play their starting players.

For the 2024 season, here is one example of cautious vs. bold betting advice strategy:

Weeks 1 -   3, set threshold = 5.0 pts : win 2,  lose 1
Weeks 4 -  10, set threshold = 2.0 pts : win 30, lose 24
Weeks 11 - 15, set threshold = 0.5 pts : win 39, lose 23
Weeks 16 - 17, set threshold = 5.0 pts : win 3,  lose 1
Week  18: no bets

Total: win 74, lose 49 = 60% accuracy

Good fun.

My NFL prediction system is named after the Zoltar fortune teller machine you can find in arcades. Here are two old newspaper articles that describe scenarios where a prediction of the future would have been helpful.