What a key number actually is
A key number is a final-margin outcome that occurs materially more often than a smooth probability distribution would predict. In the NFL, three is the textbook example: roughly one of every seven regular-season games finishes with exactly a three-point margin. That single fact reshapes how the entire spread market is priced. A line of -3 is not the same product as a line of -2.5 or -3.5; it is a discontinuity in probability space, and every serious bettor and every sportsbook treats it as such.
The term entered the public sports-betting lexicon in the 1970s through Roxborough-era Las Vegas oddsmaking, but the underlying mathematics was already understood by Hilton race-and-sports manager Lem Banker and the early computer-group sharps. What changed in the 2000s was the public availability of large game-result datasets, which let researchers like Andrew Kovalchick and Andrew Sokol publish the frequency distributions that retail bettors now treat as common knowledge.
The NFL frequency curve
The shape of the NFL margin distribution is dictated by the sport's scoring grammar. A touchdown plus extra point is seven. A field goal is three. A safety is two. A two-point conversion adds two. A missed PAT subtracts one. Stack a season of those events across hundreds of games and the resulting margin histogram is not smooth — it has tall spikes at 3, 7, and 10 with smaller spikes at 4, 6, 14, and 17.
| Margin | NFL frequency | NCAAF frequency | Notes |
|---|---|---|---|
| 3 | 14.5% | 9.6% | The king of all key numbers |
| 7 | 9.4% | 7.8% | One-TD game, second-most pricing weight |
| 10 | 6.1% | 5.4% | FG + TD differential |
| 6 | 5.5% | 4.7% | Two FGs differential |
| 4 | 5.0% | 4.9% | TD - FG differential |
| 14 | 4.6% | 4.2% | Two-TD game |
| 17 | 3.7% | 3.8% | Two TDs + FG |
| 1, 2, 5, 8, 9 | 2.0-3.0% each | 2.5-3.5% each | Structurally rare |
The takeaway is not that other margins never happen — they do, all the time — but that the cost of half-point movement is not constant across the board. Moving a spread across 3 or across 7 is far more expensive in true EV terms than moving it across, say, 8.5 to 9.5. Books know this. Sharp bettors know this. The asymmetry is the central feature of football spread pricing.
Buying half-points: the math under the hood
The half-point market exists because key numbers exist. A bettor who likes a +2.5 underdog can pay a few cents of vig to upgrade the bet to +3, and the question is whether the cost is worth the protection. The answer depends on how often the game lands exactly on 3 and what fraction of those landings would otherwise have been graded as losses or wins.
The standard framework: if a 3-point margin hits 14.5% of the time, and the favorite wins the game by exactly 3 in roughly half of those (the bettor is on the dog, so a 3-margin loss for the dog becomes a push instead of a loss), the half-point is worth approximately 7.25 cents of expected value. Books that charge 10-15 cents to make the move are charging slightly more than fair; books that charge 20-25 cents are charging materially more. The arithmetic flips for selling the half-point — taking a worse price to go from +3 to +3.5 — where the bettor is essentially being paid to give up the push protection.
The same logic, scaled down, applies across 7 (worth about 4.5 cents of EV) and across 10 (worth about 3 cents). At 4 and 6 the half-point is worth roughly 2.5 cents — borderline against the typical 10-cent spread the book charges. Past the major key numbers, buying half-points is almost always a loss for the bettor.
The two-point swing: across 3 versus across 2.5
The most expensive half-point in football is from +2.5 to +3, because it crosses the highest-frequency key number from the dog's side. The second most expensive is from -3 to -2.5, which is the symmetric trade from the favorite's side. The pricing should be roughly the same in both directions, but recreational US books frequently misprice the favorite-side buy at a discount because their pricing models focus on dog-side push protection. Sharp bettors will buy -3 to -2.5 at a US book and sell +2.5 to +3 at an offshore book to capture the asymmetry — though the practice is harder to execute as US books have tightened their key-number pricing since 2022.
Totals key numbers (less famous, still real)
Spreads get most of the key-number attention but totals have them too. In the NFL the most common total-side key numbers are 41, 43, 44, 47, and 51, with smaller spikes at 37 and 54. The driver is the same scoring grammar applied to two teams' scores summed together rather than subtracted. A total of 44 is not the same product as a total of 43.5 or 44.5; games land on it more often. The market response is the same — books charge an extra cent or two of vig at total key numbers — but the effect is smaller than on the spread side because total variance is wider and the probability of landing on any single number is lower in absolute terms.
Basketball key numbers, in their flatter form
The NBA produces key numbers but they are flatter and less actionable than football's. The dominant margins are 5, 7, 4, and 3, each hitting between 5% and 7% of games. The driver is the rhythm of late-game possessions: a one-possession game decided at the buzzer typically ends within a 1-to-5-point margin, and a two-possession game typically ends within 6-to-10. None of these spikes are sharp enough to justify buying half-points the way you would in the NFL. The practical implication for basketball bettors is to focus on spread shopping and live-line entries rather than half-point manipulation.
College basketball is messier still: the higher possession count smooths the margin distribution and key numbers barely show up. NCAAB spreads are best priced as if the distribution were continuous.
How books defend the key number
A sportsbook's biggest single positional risk on an NFL Sunday is being on the wrong side of a 3-point margin. Imagine a game with 80% of the spread handle on the favorite at -3. If the favorite wins by exactly 3, the book pushes the public side and keeps the dog losses — a windfall. If the favorite wins by 4 or more, the book pays out the entire public position and may be net negative on the game. The asymmetry is exactly what books fortify against.
Three defensive tactics dominate. Extra vig at the key number — turning -110/-110 into -115/-105 or worse when the line sits on 3 or 7. Sticky lines — refusing to move off 3 even under heavy one-sided action, forcing the bettor to take or pass at the current price. Alt spread markets — offering +2.5, +3.5, +4 as separate products with their own (carefully priced) odds, so sharps cannot freely buy half-points without paying a separate margin. Together these tactics keep the book's exposure at the key number controlled even when public action is lopsided.
Reverse line movement and the key-number signal
The most-watched sharp signal in NFL betting is reverse line movement that crosses or settles on a key number. The classic pattern: a game opens Patriots -3 (-110), the public hammers New England, but the line moves to Patriots -2.5 (-115) instead of -3.5. The book has decided to absorb the public position because respected money has come in on the dog at +3, and rather than push the line through 3 (where the book's positional risk is greatest), the book has moved it the other way. This is the cleanest non-insider signal a retail bettor can read: respected money on the dog, public money on the favorite, line crossing through 3 to land at 2.5.
The signal is strongest on Sunday morning, weakest at Monday-night open. The reliability is not perfect — books sometimes move lines for reasons unrelated to sharp money, including injury news and weather — but as a heuristic, RLM at a key number outperforms RLM at a non-key number by a wide margin in published sharp-following studies.
Alt spreads at key numbers
Alt spread markets give the bettor a menu of priced lines around the headline number. For a game posted at -3, the alt menu might run from -10 (deep favorite, plus-money price) through +10 (deep dog, heavy minus price). Each step crosses a key number with its own embedded EV cost. The bettor who understands the frequency table can spot which steps are mispriced.
The most common alt-spread inefficiency is at the +3 to +3.5 step, where some books charge only 15 cents to move a +3 favorite-side dog to +3.5, materially less than the 7.25 cents of true EV the move is worth doubled. The least common inefficiency is at the -7 to -6.5 step, where books have learned to price the touchdown-margin defense aggressively. A bettor who routinely scans alt-spread menus will find a few cents of EV per week with no model required.
Practical playbook
- Memorize the NFL frequency curve: 3, 7, 10, 6, 4, 14, 17. Anything beyond that is noise.
- Buy through 3 only when the cost is materially below 7 cents of EV — usually 4-6 cents at the sharpest books.
- Never sell off the 3 for a discount. The discount is rarely enough to compensate for the lost push protection.
- Watch RLM at key numbers on Sunday morning as a sharp-money proxy.
- Scan alt-spread menus for mispriced half-point moves, especially on dog sides.
- Apply the same logic to totals at 41, 44, 47, 51 — smaller effect, same direction.
Sources & further reading
- Kovalchick, A. & Sokol, A. "Modeling NFL Game Margins Using Final Score Distributions." Journal of Quantitative Analysis in Sports, 2017 — the canonical academic reference for NFL key-number frequencies.
- Levitt, Steven D. "Why are gambling markets organised so differently from financial markets?" Economic Journal, 2004 — the foundational analysis of bookmaker pricing strategy.
- Sharp Football Analysis — Warren Sharp's annual updates to the NFL margin frequency tables (2018-2024).
- Pinnacle Betting Resources — public documentation on half-point pricing, alt-spread construction, and key-number vig.
- Stanford Wong, Sharp Sports Betting, 2001 (revised 2009) — the practitioner's manual on key-number buying and selling, still the textbook reference.