The Quick Answer
The NFL's real key totals are 51, 37, 30, 55, 33, 41, 44 and 58. The single most common final total is 41, which has landed 264 times since 1999.
And the explanation you have read everywhere else: football scores in threes and sevens, so the key numbers are multiples of 3 and 7: is false. We measured it.
Free: The Weekly PM Market Brief, the week's biggest market-vs-model gaps, what's spiking next, and the boards worth watching. One email, Sundays. Sign up here
⭐ The 3s-And-7s Rule Is Backwards
Across 6,967 regular-season games, 1999 through 2025:
| Claim | what actually happens | what chance predicts |
|---|---|---|
| Total Is Divisible By 7 | 8.9% | 14.3% |
| Total Is Divisible By 3 | 33.4% | 33.3% |
Multiples of seven land less often than random. Not slightly, a third less often than you would get by drawing totals out of a hat.
Multiples of three land exactly as often as chance. 33.4% against 33.3%. There is no effect there at all.
⇒ The spikes in NFL totals are real and large, but they sit on specific numbers, not on multiples of anything. Every page telling you to respect 42 and 49 because they are 6×7 and 7×7 has the mechanism wrong. 42 is one of the worst totals on the board, it has landed 106 times, less than half as often as its neighbors.
The Key Numbers
"Lift" is how often the score lands on a total compared with the average of the six totals within three points of it. Above 1.00 means the number attracts scores beyond what the shape of the curve would predict.
| Total | games | lift vs neighbors |
|---|---|---|
| 51 | 262 | 1.81x |
| 37 | 257 | 1.58x |
| 30 | 187 | 1.51x |
| 55 | 172 | 1.45x |
| 33 | 208 | 1.45x |
| 41 | 264 | 1.41x |
| 44 | 264 | 1.37x |
| 58 | 138 | 1.37x |
51 is the strongest key number in football and almost nobody names it. It is not a multiple of 3 or 7.
The top ten totals cover 33.5% of all games across a range spanning 103 points. A third of every NFL game since 1999 finished on one of ten numbers.
Every Total, 1999–2025
The full distribution. Lift is shown only where it means something. See the note under the table.
| Total | games | share | lift |
|---|---|---|---|
| 13 | 22 | 0.32% | 1.63x |
| 16 | 50 | 0.72% | 2.38x |
| 17 | 27 | 0.39% | 0.94x |
| 19 | 53 | 0.76% | 1.45x |
| 20 | 46 | 0.66% | 0.90x |
| 21 | 35 | 0.50% | 0.60x |
| 22 | 57 | 0.82% | 0.91x |
| 23 | 129 | 1.85% | 2.13x |
| 24 | 61 | 0.88% | 0.69x |
| 25 | 50 | 0.72% | 0.53x |
| 26 | 114 | 1.64% | 1.21x |
| 27 | 147 | 2.11% | 1.50x |
| 28 | 54 | 0.78% | 0.44x |
| 29 | 123 | 1.77% | 1.03x |
| 30 | 187 | 2.68% | 1.51x |
| 31 | 117 | 1.68% | 0.83x |
| 32 | 96 | 1.38% | 0.62x |
| 33 | 208 | 2.99% | 1.45x |
| 34 | 176 | 2.53% | 1.10x |
| 35 | 121 | 1.74% | 0.70x |
| 36 | 164 | 2.35% | 0.94x |
| 37 | 257 | 3.69% | 1.58x |
| 38 | 143 | 2.05% | 0.73x |
| 39 | 142 | 2.04% | 0.73x |
| 40 | 227 | 3.26% | 1.18x |
| 41 | 264 | 3.79% | 1.41x |
| 42 | 106 | 1.52% | 0.48x |
| 43 | 245 | 3.52% | 1.25x |
| 44 | 264 | 3.79% | 1.37x |
| 45 | 177 | 2.54% | 0.90x |
| 46 | 141 | 2.02% | 0.68x |
| 47 | 222 | 3.19% | 1.24x |
| 48 | 197 | 2.83% | 1.08x |
| 49 | 139 | 2.00% | 0.74x |
| 50 | 155 | 2.22% | 0.88x |
| 51 | 262 | 3.76% | 1.81x |
| 52 | 143 | 2.05% | 0.89x |
| 53 | 94 | 1.35% | 0.60x |
| 54 | 142 | 2.04% | 0.98x |
| 55 | 172 | 2.47% | 1.45x |
| 56 | 60 | 0.86% | 0.47x |
| 57 | 134 | 1.92% | 1.23x |
| 58 | 138 | 1.98% | 1.37x |
| 59 | 91 | 1.31% | 0.96x |
| 60 | 52 | 0.75% | 0.51x |
| 61 | 96 | 1.38% | 1.18x |
| 62 | 87 | 1.25% | 1.14x |
| 63 | 63 | 0.90% | 0.84x |
| 64 | 59 | 0.85% | 0.82x |
| 65 | 96 | 1.38% | 1.68x |
| 66 | 60 | 0.86% | 1.02x |
| 67 | 32 | 0.46% | 0.56x |
| 68 | 42 | 0.60% | 0.79x |
| 69 | 60 | 0.86% | 1.59x |
| 70 | 23 | 0.33% | 0.57x |
| 71 | 47 | 0.67% | 1.37x |
| 72 | 23 | 0.33% | 0.65x |
| 73 | 37 | 0.53% | 1.43x |
| 74 | 21 | 0.30% | 0.77x |
| 75 | 23 | 0.33% | 1.08x |
33 further totals each occur fewer than 20 times, covering 194 games (2.8%).
⚠️ Why some lift figures should be ignored. Lift only means something where the neighboring totals have enough games to form a stable baseline: 32 of the 93 values, those whose neighbors average 100+ games. Out in the tails, a total with 20 games sitting beside neighbors with 3 produces an enormous ratio on almost no data. It looks like a key number and is not one. That is exactly how a "6.00x key number at 101 points" gets published elsewhere; we caught ours before it went out.
How The Kalshi Total Ladder Turns This Into A Price
Kalshi lists per-game totals under the series KXNFLTOTAL ("Pro Football Total Points"). It is worth knowing the shape before you use any of the table above against it, because the shape is what makes the table usable.
The ladder is nested, not disjoint. The New York vs Pittsburgh board carried 19 rungs, and every rung is phrased the same way: Over 18.5 points scored, Over 19.5, on up to Over 57.5. Each one resolves Yes if the two teams "collectively score more than" that number. Nothing on the board asks about an exact total, and because every threshold sits on a half point while every real final total is a whole number, no rung can push.
That half-point detail is the useful part. A nested ladder of Over N.5 rungs carries the exact-total distribution inside it as differences:
P(the total lands exactly on N) = P(Over N minus 0.5) less P(Over N plus 0.5)
So the spike at 51 in the table above is not an abstraction. It is the width of the step between the Over 50.5 rung and the Over 51.5 rung. The eight key numbers are the eight places where that step is widest, and the flat stretches in the table are where adjacent rungs should sit nearly on top of each other.
⚠️ This tells you where the steps belong, not that the board has them wrong. The efficiency check below is the part that matters: against the closing total the historical split is 48.8% over, 49.8% under, mean error 0.69 points. A structural map of the ladder is not an edge in it, and the measurement section is deliberate about that distinction.
⛔ What This Table Is Not
It is not a way to beat the number. Against the closing line across all 6,967 games:
| Over | 48.8% |
| Under | 49.8% |
| Push | 1.4% |
| Mean Error | +0.69 points |
| Median Error | 0.00 |
The closing total is a good estimate of the final score. It misses by less than a point on average and by nothing at all in the median game. Overs and unders split almost exactly evenly.
⇒ This is a lookup table, not an angle. Knowing that 51 is a magnet tells you something real about where scores cluster. It does not tell you the market has mispriced anything, because the market already knows.
What it is genuinely useful for:
- Judging A Half-Point. Moving a total from 50.5 to 51.5 crosses the single strongest key number in football. Moving 45.5 to 46.5 crosses almost nothing.
- Reading A Push. 1.4% of games land exactly on the number. On a total of 41 or 44 that risk is materially higher than on 42 or 53.
- Sanity-Checking A Claim. Anyone telling you 42 and 49 are key numbers is repeating a rule, not a measurement. Both are below-average totals.
How This Was Measured
Every regular-season NFL game from 1999 through 2025, a total of 6,967 games. The final combined score of each, counted. No model, no projection, no sample.
Divisibility was tested against chance, not against intuition. One in seven totals is divisible by seven, so 14.3% is the number to beat. The real rate is 8.9%. One in three is divisible by three, so 33.3% is the bar. The real rate is 33.4%.
- NFL 2026 on Kalshi: Every Board We Track
- NFL Key Numbers: Why 3 And 7 Rule The Point Spread
- How To Bet NFL On Kalshi: A Step-by-Step 2026 Guide
- NFL MVP Odds 2026: Kalshi Vs Polymarket Vs DraftKings
- AI NFL Predictions Week 1: 4 AIs Vs Kalshi On All 16 Games
FAQ
What is the most common NFL total? 41, with 264 games since 1999. 44 is tied on games (264) and just behind on lift.
What are the NFL key numbers for totals? 51, 37, 30, 55, 33, 41, 44 and 58. 51 is the strongest, landing 1.81 times as often as the totals around it.
Is 42 a key number because it is 6 x 7? No. 42 has landed 106 times, a lift of 0.48x — less than half as often as its neighbors. It is one of the weakest totals on the board.
How often do NFL games go over 50 points? Totals of 51 and above account for a substantial share of the distribution, and 51 itself is the most magnetic number in the table at 262 games. The full breakdown by exact total is above.
Do totals really land on multiples of 3 and 7? Multiples of three land at 33.4% against a chance rate of 33.3% — no effect. Multiples of seven land at 8.9% against 14.3% — meaningfully less often than random.
Can I use this to beat the total? No, and the data says so plainly. Against the closing line: over 48.8%, under 49.8%, push 1.4%, mean error 0.69 points. The market prices totals well.
These are model estimates, not predictions of fact and not financial advice. They are historical frequencies from completed games; past distributions do not determine future ones. Kalshi event contracts are CFTC-regulated derivatives traded on a designated contract market, not sportsbook wagers, and they can lose their full value. 18+, where legal.
OddsShopper covers betting markets as an independent analyst. No outcome here is certain, and the closing total is shown above to be an efficient price. The risk of loss is real.
