Zoltar is my NFL football prediction computer program. It uses a neural network and a type of reinforcement learning. The NFL 2025 season ended on Sunday, February 8, 2026 when the Seattle Seahawks beat the New England Patriots by a score of 29-13 in Super Bowl LX (60).

Zoltar had a good, but not great, season. For the 272 NFL games played during the regular season (i.e., not including the 12 post-season games and the Super Bowl), Zoltar predictions were 46-30 (~60% accuracy) against the Las Vegas point spread. Including all 13 post-season games, Zoltar was 50-30 (~62% accuracy) against the spread.

This is what Zoltar looks like (2025 week #18 predictions)

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.

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 to take into account logistics and things like manual data entry errors.

Before the season started, I asked Zoltar to predict the number of wins each team would reach (out of the 17 regular season games). I figured I’d check those predictions against the actual number of wins by each of the 32 NFL teams:

              Predicted  Actual
      Team       Wins     Wins   Diff
---------------------------------------------------
      bears       5       11      +6
    bengals       8        6      -2
      bills      12       12       0
    broncos       9       14      +5

     browns       3        5      +2
 buccaneers      10        8      -2
  cardinals       7        3      -4
   chargers      11       11       0

     chiefs      13        6      -7
      colts       7        8      +1
    cowboys       7        7       0
   dolphins       7        7       0

     eagles      12       11      -1
    falcons       7        8      +1
fortyniners       7       12      +5
     giants       3        4      +1

    jaguars       5       13      +8
       jets       6        3      -3
      lions      13        9      -4
    packers      10        9      -1

   panthers       5        8      +3
   patriots       5       14      +9
    raiders       5        3      -2
       rams      10       12      +2

     ravens      11        9      -2
 commanders      11        5      -6
     saints       5        6      +1
   seahawks      10       14      +4

   steelers       9       10      +1
     texans       9       12      +3
     titans       3        3       0
    vikings      12        9      -3

Not too bad. Zoltar was correct within plus-or-minus 2 games for 18 of the 32 teams.

The two biggest positive surprises to Zoltar were the Jaguars (+8 wins) and Patriots (+9 wins).

The biggest negative surprises were the Chiefs (-7 wins) and Commanders (-6 wins).

Well, Zoltar is going into hibernation for the next seven months, until the 2026 season starts in September.

Left: My prediction system is named after the Zoltar fortune teller machine you can find in arcades. Zoltar machines often make appearances at conferences too.

Center and Right: I got a very cool mini-Zoltar machine for Christmas this year.