NFL Key Numbers: 27 Years and 7,276 Games of Margin Data

Everyone "knows" NFL final scores cluster on certain numbers — 3 and 7 above all. Almost nobody has checked that against their own database. So we analyzed 7,276 finished NFL games (1999-2025, 27 seasons) together with their real closing lines — margins, spreads, totals. According to our analysis, the winning team finishes ahead by exactly 3 points in 15.1% of all games — the single most common margin in the league. Here's the rest of what the numbers show. Unlike the usual "key numbers" articles that cite round textbook percentages, every figure and table below is computed directly from the actual final scores and closing lines in our own database — including the sample size behind every percentage, so you can judge for yourself how solid it is.

Method: exactly what we counted

This is the same database our NFL models run on, so the hygiene rules are strict, and every number below can be traced back cleanly to exactly these filters:

• Only games with status = FINISHED and both final scores, NFL (sport_id=7), 1999-2025 (27 seasons).
• Season assignment: January/February playoff games count toward the PRIOR season (a game in February 2016 belongs to the 2015 season).
• Spread convention: nflverse, positive `spread_line` = home team favored. Home covers if margin > line; away covers if margin < line; push on exact equality.
• Of those games, 100.0% have a captured closing line (spread and/or total) — every line-dependent section below uses exactly that subset.

The margin histogram: NFL games don't end at random scores

Across all 7,276 scored games, the distribution of final margins is far from smooth or random — it clusters hard on a small handful of numbers. The top margins, ranked by frequency:

• Margin of 3 — 15.1% of all games (1,100 of 7,276)
• Margin of 7 — 9.0% of all games (657 of 7,276)
• Margin of 6 — 6.0% of all games (436 of 7,276)
• Margin of 10 — 5.5% of all games (401 of 7,276)
• Margin of 4 — 4.9% of all games (355 of 7,276)
• Margin of 14 — 4.9% of all games (355 of 7,276)
• Margin of 1 — 4.2% of all games (303 of 7,276)
• Margin of 2 — 4.1% of all games (295 of 7,276)

Ties are extremely rare — 0.2% of all games (the NFL has allowed regular-season ties after one overtime period since 2012).

Why 3 and 7 dominate everything

A field goal is worth 3 points, a touchdown plus extra point is worth 7. Because so many NFL drives end in exactly one of those two outcomes, the final margin clusters at exactly those numbers — plus secondary bumps at multiples and combinations of them. This isn't a betting theory; it's just how the sport scores.

Secondary bumps: 4 (4.9%), 6 (6.0%), 10 (5.5%), 14 (4.9%), 17 (3.6%).

The secondary bumps are pure combinatorics of the same two numbers: 6 is two field goals (or a touchdown with a missed extra point), 10 is a field goal plus a touchdown, 14 is two touchdowns, and 17 is a field goal plus two touchdowns (or two field goals plus a touchdown). Almost the entire upper half of the distribution can be reconstructed from sums and differences of 3s and 7s — everything else (5, 9, 11, 12, 15, 16, 18, 19) requires a less common scoring sequence (a safety, a failed two-point conversion, a blocked-kick return) and shows up correspondingly less often.

The half-point value table

The most direct way to identify a key number is the push rate: how often does the final margin land exactly on the line? These are, empirically from real closing lines, the lines with the highest push rate in our database:

• Line 3.0 (n=1,153): pushes 9.0% of the time, favorite covers 47.6% when not a push
• Line 10.0 (n=213): pushes 6.6% of the time, favorite covers 44.7% when not a push
• Line 7.0 (n=491): pushes 6.5% of the time, favorite covers 50.3% when not a push
• Line 14.0 (n=90): pushes 4.4% of the time, favorite covers 44.2% when not a push
• Line 6.0 (n=354): pushes 3.7% of the time, favorite covers 49.6% when not a push
• Line 12.0 (n=33): pushes 3.0% of the time, favorite covers 31.2% when not a push

That's the entire math behind "buying points": a half-point that removes one of these pushes (e.g. -3 to -2.5, or -7 to -6.5) is worth real money — the same half-point on a thin number (e.g. -10.5 to -10) buys almost nothing. More on that in our key numbers guide.

Sportsbooks price this the same way: the extra charge to buy a half-point across 3 or 7 typically sits above standard vig (worse than -110, often -120 or -125), while the same half-point on a thin number often stays at -110 or even improves. The vig isn't flat across the spread spectrum — it's highest exactly where these numbers say the half-point is worth the most.

Did the 2015 extra-point rule change move the key numbers?

In 2015, the NFL moved the extra-point attempt from the 2-yard to the 15-yard line — the success rate dropped noticeably. If touchdown-plus-XP ends in exactly 7 points less often (a missed XP turns it into 6), that should show up in the margin distribution:

• 1999-2014 (before the rule change, n=4,248): margin 3 = 15.3%, margin 6 = 5.3%, margin 7 = 9.3%
• 2015-2025 (after the rule change, n=3,028): margin 3 = 14.8%, margin 6 = 6.9%, margin 7 = 8.7%

Totals key numbers: 41, 44, 47

Just like spreads, game totals cluster on certain numbers — sums that fall naturally out of field-goal and touchdown combinations. Across 7,276 games with a captured total line:

• Total 37.0 (n=185): over hits 42.7% of the time, push rate 3.8%
• Total 40.0 (n=198): over hits 51.0% of the time, push rate 5.1%
• Total 41.0 (n=322): over hits 47.8% of the time, push rate 3.4%
• Total 43.0 (n=306): over hits 47.4% of the time, push rate 3.9%
• Total 44.0 (n=362): over hits 50.6% of the time, push rate 4.1%
• Total 47.0 (n=262): over hits 43.9% of the time, push rate 2.7%
• Total 50.0 (n=80): over hits 36.2% of the time, push rate 3.8%
• Total 51.0 (n=97): over hits 46.4% of the time, push rate 4.1%

Overall, the total went over the line in 48.7% of games and under in 49.8% — close enough to 50/50 to show how efficiently the closing total is typically set.

Why exactly 41, 44 and 47? Because a typical NFL team score is itself built from 3s and 7s (20, 21, 23, 24, 27 are all common single-team scores) — add two such numbers together and the sum naturally lands right in that band. A total of 44, for example, is exactly what a 24-20 or a 27-17 final produces.

How sharp is the closing line, really?

Across 7,245 games with a clear favorite, the favorite covered 48.7% of the time and the underdog 51.3% (pushes: 2.7%). That sits close to 50/50 — exactly the signature of an efficiently priced market: the line is set so both sides cover at roughly even rates, regardless of who's favored.

The home/away split tells the same story from a different angle: home teams against the spread covered 49.0%, away teams 51.0% — both close to 50%. Unlike raw win share (where home advantage is real), the spread line is explicitly designed to price that home advantage out entirely — and by these numbers, it does so almost exactly.

What this means for bettors

None of this is a betting edge — it's description, not prediction. Sportsbooks know this distribution at least as well as we do, and they price half-point lines accordingly (the juice to buy a half-point across 3 or 7 is almost always higher than on thin numbers). What this data really gives you: an honest calibration of your own intuition before you read a line. Practically useful next steps are our teaser calculator (which uses exactly these key numbers to score whether a tease crosses one) and our no-vig calculator to turn any quoted price into a fair probability.

For teaser bets, the consequence is direct: a leg that doesn't cross a key number is the same mistake as buying a half-point on a thin number — you're paying for insurance against margins that barely happen. Our key numbers guide walks through the full cover-rate logic behind the classic 6-point Wong teaser, which is built specifically to cross both 3 and 7 in a single move.

See where this week's lines sit relative to these key numbers on our NFL 2026 schedule, full-season projections on NFL win totals, and how our model bets have performed overall on the public track record.

Honest limitations

• Results and closing lines only, not a live betting recommendation. This study says nothing about whether any of this is profitably bettable right now.
• 27 years are not 27 identical eras. Overtime rules (2010, 2012, 2017), kickoff rules (2024), and general pace/scoring trends shift the margin distribution alongside the 2015 extra-point move — we isolate only that one rule change, not everything at once.
• Key-number frequencies move with the scoring environment. A league with more field goals or more two-point conversions shifts the distribution slightly — the percentages shown here are averaged across the whole period.