Zoltar is my NFL football prediction computer program. It uses a neural network and a type of reinforcement learning. Here are Zoltar’s predictions for week #19 (wildcard playoff games) of the 2024 season.

Zoltar:      texans  by    4  dog =    chargers    Vegas:    chargers  by    3
Zoltar:      ravens  by    6  dog =    steelers    Vegas:      ravens  by  9.5
Zoltar:       bills  by    6  dog =     broncos    Vegas:       bills  by    9
Zoltar:      eagles  by    6  dog =     packers    Vegas:      eagles  by  4.5
Zoltar:  buccaneers  by    2  dog =  commanders    Vegas:  buccaneers  by    3
Zoltar:     vikings  by    0  dog =        rams    Vegas:     vikings  by    2

Zoltar theoretically suggests betting when the Vegas line is “significantly” different from Zoltar’s prediction. For the beginning and end of the season, I use a relatively conservative threshold of 4.0 points difference. For the middle of the season, I use a relatively aggressive threshold of 3.0 points difference.

For week #19, five of Zoltar’s six predictions are less than 4.0 points difference from the Vegas point spreads and so Zoltar has no wagering recommendations on those games. But strangely, there’s a big discrepency between Zoltar and Las Vegas for the Chargers at Texans game. Las Vegas picks the Chargers to win by 3 points, but Zoltar thinks the Texans will win by 4 points — a big 7-point difference. Therefore, Zoltar suggests a wager on the Texans.

A hypothetical wager on the Texans would win if the Vegas underdog Texans by any score, or if the favored Chargers win but by less than 3.0 points (i.e., 2 points or less). If the Chargers win by exactly 3 points, all wagers are a push. The annoying possibility of pushes are why point spreads often have a “point five”, such as 3.5, because then pushes aren’t possible.

Theoretically, if you must bet $110 to win $100 (typical in Vegas) then you’ll make money if you predict at 53% accuracy or better. But realistically, you need to predict at 60% accuracy or better.

In week #18, the final regular season week, against the Vegas point spread, Zoltar went 0-0 using 4.0 points as the advice threshold, because there were so many games where one team had nothing to gain because the playoff spots were mostly determined, and many teams sat many or all of their normal starting players on the bench.

For the season, against the spread, Zoltar is 45-27 (~62% accuracy).

Just for fun, I track how well Zoltar does when just trying to predict just which team will win a game. This isn’t useful except for parlay betting. In week #18, just predicting the winning team, Zoltar went 8-8 which is poor, but again, many teams sat starters. Vegas was better, going 10-6 at just predicting the winning team, because Vegas indirectly takes subjective factors into account.

My prediction system is named after the Zoltar fortune teller machine you can find in arcades. Fortune telling with a crystal ball has been around for over 2,000 years. It was described by ancient Roman Pliny the Elder as “crystallum orbis”. Almost all cultures have some form of fortune teller. Here are three results of AI-generated images, but I doubt that the costumes are historically accurate Left: “Chinese fortune teller”. Center: “Egyptian fortune teller”. Right: “Russian fortune teller”.