# A New National Football League Prediction Model

**AUTHOR’S NOTE**: This is my substantive post for today; I like sports forecasting because it offers immediate and irrefutable proof of the accuracy of one’s techniques.

For several years now I’ve been using the tarot to attempt prediction of the winning team in sporting events, particularly National Football League and Major League Baseball games in the US because their respective scoring models fit comfortably within the numerical range of the cards. However, I’ve been dissatisfied with the results because the “frame-by-frame” (quarter, inning, period, etc) numbers derived from the cards don’t always present a realistic picture. I had been pulling a single card to show each team’s scoring output in a comparative way; this worked best with football because it has four quarters and the number Four is a recurring theme in the tarot (baseball with its nine innings took some imaginative adjustment of the spread design). But by using the full 78-card deck I kept getting “odd” numbers that would almost never appear in an actual game and had to round up or down to bring them into precise alignment.

With football, the premise certainly wasn’t unreasonable. During each quarter, the maximum score that can be obtained from a single card is 21 (three touchdowns) and the minimum is zero. The problem is that scoring normally occurs in set increments of 3 and 7 and their multiples (with a rare 2 and 1 thrown in); there are too many “filler” cards in the deck that dilute the integrity of the model and yield irrational results in most cases: 1, 2, 4, 5 and 8 in particular, both as individual values and when added to the standard “base” scores (3 and 7) to produce uncommon ending numbers like 11, 12, 15, 18 and 19. Also, when low-numbered cards appear, using only one card per period for each team severely limits the opportunity to capture multiple scoring events.

The first solution that occurred to me was to treat the comparative single-card face values as showing the raw “point gap” between the teams at the end of each quarter rather than the actual scores, with the four-quarter roll-up reflecting the final “win/loss” margin for the outcome. This resembles the gambler’s “point spread” but is less sophisticated. I’m experimenting with that right now for 2022 season predictions, but I may have hit upon a better way involving use of a “prepared” deck that only includes cards that align with the normal scoring evolution in a game.

In total, the tarot contains five “3's” (representing a field goal); five “6's” (two field goals or one touchdown with failed extra-point conversion); five “7's” (a touchdown plus extra point); five “9's” (three field goals); five “10's” (a touchdown/extra point and a field goal); five “13's” (a touchdown/extra point and two field goals); five “14's” (two touchdowns/extra points); one “16” (a touchdown/extra point and three field goals); one “17” (two touchdowns/extra points and a field goal); one “20” (two touchdowns/extra points and two field goals) and one “21” (three touchdowns/extra points). Although the higher numbers are infrequent in a typical game, none of these scores are unheard-of quarterly totals for any team, and since the higher cards can only appear once in a reading they would not artificially bloat the score. (On the other hand, I believe there are enough high-numbered cards in the population to accommodate the occasional “blow-out” by a vastly superior team, or “runaway” scoring by two opponents with weak defensive lines.)

To address the likelihood of multiple scoring events, the five “1's” and five “2's” in the deck could be dealt separately from the main pull as “multipliers” for any single-card value (and perhaps as “divisors” for unreasonably large numbers); for example a “7” overlaid with a “1” would mean a single touchdown during a quarter, while accompanied by a “2” it would imply two touchdowns. (Since the extra-point conversion is almost a foregone conclusion following any touchdown, I’m going to assume it in the score.) I would only use this approach for the single-digit numbers “3, “6” and “7” since “9” and above already encompass the possibility of more than one score in a period. Using “1's” and “2's” as “divisors” would create a 50% chance to cut a very large quarterly scoring projection in half if the attempt seems justified, although any team is capable of double digits if they control the ball.

This concept seems to hold some promise. I’m going to apply it as I get into game-by-game predictions at the start of the next football season. I’ll have to think some more about how to come up with a reasonable model for baseball since its scoring most often occurs in smaller increments per inning spread over nine frames. However, I’ve had some remarkably successful results with World Series forecasts using my current methods, so I may not have to change much.

*Originally published at **http://parsifalswheeldivination.wordpress.com** on March 28, 2022.*