Zoltar is my NFL football prediction computer program. It uses reinforcement learning and a neural network. Here are Zoltar’s predictions for week #17 of the 2022 season.

Update: My original posted predictions were completely wrong — I used an experimental version by mistake. The predictions should have been:

Zoltar:     cowboys  by    3  dog =      titans    Vegas:     cowboys  by   10
Zoltar:     falcons  by    1  dog =   cardinals    Vegas:     falcons  by  3.5
Zoltar:       lions  by    6  dog =       bears    Vegas:       lions  by  5.5
Zoltar:     jaguars  by    0  dog =      texans    Vegas:     jaguars  by    4
Zoltar:      chiefs  by   16  dog =     broncos    Vegas:      chiefs  by 13.5
Zoltar:    patriots  by    4  dog =    dolphins    Vegas:    dolphins  by  1.5
Zoltar:      giants  by    6  dog =       colts    Vegas:      giants  by    3
Zoltar:      eagles  by   10  dog =      saints    Vegas:      eagles  by  6.5
Zoltar:      ravens  by    6  dog =    steelers    Vegas:      ravens  by    3
Zoltar:  buccaneers  by    7  dog =    panthers    Vegas:  buccaneers  by    3
Zoltar:  commanders  by    6  dog =      browns    Vegas:  commanders  by  2.5
Zoltar: fortyniners  by    3  dog =     raiders    Vegas: fortyniners  by  5.5
Zoltar:    seahawks  by    5  dog =        jets    Vegas:        jets  by  2.5
Zoltar:     vikings  by    0  dog =     packers    Vegas:     packers  by    3
Zoltar:    chargers  by    6  dog =        rams    Vegas:    chargers  by  6.5
Zoltar:       bills  by    0  dog =     bengals    Vegas:       bills  by    1

Zoltar theoretically suggests betting when the Vegas line is “significantly” different from Zoltar’s prediction. For this season I’ve been using a threshold of 4 points difference but in some previous seasons I used 3 points.

At the beginning of the season, because of Zoltar’s initialization (all teams regress to an average power rating) and other algorithms, Zoltar is very strongly biased towards Vegas underdogs. I probably need to fix this. For week #17 Zoltar likes three Vegas underdogs:

Titans, Patriots, Seahawks

For example, a bet on the underdog Titans against the Cowboys will pay off if the Titans win by any score, or if the favored Cowboys win but by less than 10.0 points (i.e., 9 points or less). If a favored team wins by exactly the point spread, the wager is a push. This is why point spreads often have a 0.5 added — called “the hook” — to eliminate pushes.

At the end of the season, predictions get very dicey. Injuries have accumulated and teams that have nothing to gain often sit key players to avoid injuries.

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 #16, against the Vegas point spread, Zoltar went 3-3 using 4.0 points as the advice threshold. Dang. If only one game result had been different, Zoltar would have gone 4-2 instead of just 50% correct.

For the season, against the spread, Zoltar is 54-32 (~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 #16, just predicting the winning team, Zoltar went 11-5 which is good but not great. Vegas was about the same at 10-6 when just predicting the winning team.

My prediction system is named after the Zoltar fortune teller machine that can be found in arcades. Coin operated fortune tellers have been around for well over 100 years. Center: The Sega Astrodata fortune teller was manufactured in 1971. The user input their birthdate and then the machine would punch holes in a card to indicate seven predictions for things like “Your financial prospects will be” (poor, favorable, good, very good, excellent). Right: The Egyptian Seeress was a one-of-a-kind fortune teller (date looks to be 1960s, manufacturer unknown).