As football begins again, sort of, I am reminded of one of the very best football prop angles of all time, first and second half team total under 10.5. This was a huge angle for me that really allowed me to snowball my bankroll due to the big advantage, simplicity / certainty that it was good meaning the rare losses that did occur did not bother me, and large volume of plays each week.

Readers here are almost certainly familiar with "key numbers" in football, where games are much more likely to fall on the margins of 3 and 7. What is lesser known is that there were also key numbers in full game totals, the biggest being 37 (four touchdowns and three field goals). However this key number effect greatly decays as you get into totals in the mid 40s, and in the modern game is even weaker due to the extra point rule and more two-point tries, to the point that it can basically be ignored. It is very rare a total as low as 37 comes up nowadays, and if the total is 37 in the 2020 NFL, the weather is probably horrible meaning one or more extra points are likely to be missed.

However, in quarter and half totals, and especially team totals, where the number of scoring events in the game and possibility of two point plays is limited, key numbers are still extremely important. In a single half team total, 10 is by far the most powerful key number in football. One field goal and one touchdown. Since 2009, 15.5% of all first half team totals have fallen 10.

A very basic model can be used to project the average score of a team in the first half based on the full game total and spread (using first half odds is very slightly better, and you can also adjust for things like the -3 being more important than -2, vig, and higher home field in the first half, but this is what I used at the time):

Coefficient | |

Intercept | 1.93 |

Points Favored | 0.2822 |

Over-Under | 0.209 |

Using this average projection as the input to a logistic regression, we can then determine the chance a team posts a total over 10.5 based on this average:

Coefficient | |

Intercept | -2.884 |

First Half Average Points | 0.2352 |

In this type of model, the chance of the game going over 10.5 will be 50% when (First Half Average Points) * .2352 = 2.884, meaning that for a team to have a 50-50 shot of going over, they need to average 12.27 points in the first half, or 1.77 points more than 10.5. And if a team was truly expected to score 10.5 points on average in the first half, they would only be expected to go over 10.5 39.7% of the time.

Initially the lines were set such that teams with an average expectation of 10.5, not 12.27, were a 50/50 shot. This meant that you would always bet the under if the total was 10.5. Usually you had to lay -115, so you had to win 53.7% of the time when in reality the prop would cover 1 - 39.7% or 60.3% of the time. Over the course of a weekend it was basically impossible to lose. Put another way, with two teams and two halves and assuming the game was priced at pick, betting every under 10.5 first half team total was like betting under 12.27*4 or under 49 when the actual game total was 10.5*4 or under 42. So you were getting a full touchdown of pure value without doing any handicapping or real work at all.

In college football, the key number effects are not quite as strong because there is more dispersion in scores around the average total, but the prop still covered around 57 or 58% in college. About two-thirds of the major books that paid around this time had these, and in any given week there were dozens of chances to bet under 10.5, plus there were other key numbers such as 17.5 in college which while not quite as strong still offered a very big edge.

One would have assumed the books adjusted right away to someone betting the same prop for the max every single game at every book that had it and winning, but I hammered these for a full season without noticing any adjustment. In the second year I bet these, lines began to come down to -125 or -130 in the NFL although college was still very strong as long as you bet as soon as the lines came out. Now there are few books that offer these in college, and the NFL lines are typically reasonable. However, if you are a smaller bettor satisfied with a $250 or $500 bet at a single book it is still possible to find +EV rogue lines on these and similar derivatives, especially in college.