Two shows on the Humor Index tell jokes at exactly the same rate.
Fleabag fires 3.46 jokes per minute. The Simpsons fires 3.46 jokes per minute. Same number, to two decimal places, across 12 episodes and 552 episodes respectively. If joke density were the thing that made a comedy funny, these two would be neighbours on the leaderboard.
Fleabag scores 95.8. The Simpsons scores 79.0.
That is a 16.8-point gap between two shows telling the identical volume of jokes — and it is the cleanest illustration we have of something the data has been saying for months. Joke density is a quarter of the Humor Index and it decides almost nothing.
The setup
Every show on this site gets a Humor Index built from four components: craft (40%), impact (35%), peak density (15%) and weighted jokes-per-minute (10%). Density, in other words, is a quarter of the score by design. It is not a rounding error in the formula. We built it in deliberately, because a comedy that makes you laugh twice an hour is not doing the same job as one that makes you laugh twice a minute.
So the question is not whether density counts. It is: across real shows, does it ever decide anything?
It does not
Take the 13 shows currently scored on our canonical model — one scoring run, one scale, no cross-model contamination:
- Craft vs Humor Index: r = 0.96. Craft alone predicts 91% of the variance in the final score.
- Jokes-per-minute vs Humor Index: r = 0.28. Density predicts 8%.
- Craft vs jokes-per-minute: r = 0.13. The two are, for practical purposes, unrelated to each other.
That third number is the important one. If dense shows also happened to be well-written shows, density would look like it mattered when it was really craft wearing a disguise. It does not happen. Writing more jokes and writing better jokes are close to statistically independent activities.
Run the same three correlations across all 21 scored shows and you get r = 0.97, r = 0.36 and r = 0.26 — same story, slightly louder.
The two natural experiments
The clean way to see this is to find shows that hold density constant and let craft vary. We have two exact ties, and both fall entirely among canonically-scored shows:
| Jokes/min | Show | Craft | Humor Index |
|---|---|---|---|
| 3.52 | 30 Rock | 7.67 | 97.4 |
| 3.52 | Broad City | 7.16 | 86.9 |
| 3.46 | Fleabag | 7.62 | 95.8 |
| 3.46 | The Simpsons | 6.92 | 79.0 |
Identical joke rates. Ten-and-a-half points apart in the first pair, nearly seventeen in the second. In both cases the entire gap is carried by craft and the impact score that moves with it.
Now run it the other way. Flight of the Conchords averages 2.47 jokes per minute, one of the sparsest shows we score. The Simpsons averages 3.46, or 40% more jokes in every minute of television. Conchords scores 92.0 and finishes #7. The Simpsons scores 79.0 and finishes #16.
That pair is not a controlled comparison — Conchords also out-crafts The Simpsons by 0.66, which is most of the available range, and by the exchange rate below that alone would account for the gap. Which is the point. When density and craft pull in opposite directions, craft wins so decisively that the density difference never surfaces.
The exchange rate
Fit a line through the canonical 13 and you can price the two inputs directly:
- One tenth of a point of craft is worth about 2.3 Humor Index points.
- One additional joke per minute — for the entire run of the show — is worth about 3.9.
So a tenth of a craft point costs you roughly 0.6 jokes per minute to replace. That is one extra joke every hundred seconds, in every episode, forever, just to break even with a writers' room that got marginally sharper.
And here is the part that makes the exchange rate bite: nobody has that much craft to trade. Across the canonical 13, craft runs from 6.92 to 7.67 — a 0.75-point band on a ten-point scale. Density over the same 13 runs from 2.16 to 3.57, a 1.65x spread. Widen to all 21 shows and craft still only spans 0.92 while density spans 2.7x. Whichever set you use, the variable with the enormous range is the one that does not move the needle, and the variable that decides everything sits inside a band of about 13%.
That is not a quirk of our formula. It is a statement about the ceiling of the craft: television comedy writing is a narrow, brutally competitive skill where the distance between the best room we have measured and a perfectly respectable one is three quarters of a point. Priced at 23.5 points per craft point, that 0.75 is worth 17.6 Humor Index points — more than the entire canonical density range (1.41 jokes per minute) can buy at 3.9 points each, which is 5.5.
A show that ran the experiment on itself
The cross-show version of this always invites the objection that we are comparing incomparable things. So here is the within-show version, one show, one scale, one writers' room.
It's Always Sunny in Philadelphia did not get denser as it went on. The cult era (Seasons 1–7) averages 1.77 jokes per minute. The renaissance era (Seasons 13–17) averages 1.70. That difference is not statistically meaningful — the two eras tell jokes at effectively the same rate.
The renaissance era scores 3.1 points higher, and that gap is significant.
Same five leads, same core writers, same joke rate, better show. Whatever Sunny learned over fifteen years, it wasn't to talk faster.
What this means if you make comedy
The instinct in a room under pressure is to add jokes. More passes, more alts, more punch-ups, higher density. The data says that instinct is a bad trade at the margin.
Hold quality fixed and density does pay: run craft, impact and jokes-per-minute into the score together and each additional joke per minute is worth about 2.5 Humor Index points. It is not nothing. But the available range on density inside a single show's identity is small, and the range on craft, though narrower in raw units, is priced twenty-three points to the unit. Moving a tenth of a point on craft — 2.3 points — takes six-tenths of a joke per minute to match. Sustained, every episode, forever.
Volume is the cheap lever. It is cheap because there is so little of it left to buy.
What we still cannot see
The usual caveats, and one specific to this post:
- We read dialogue. Density is measured on scripted, spoken jokes. Shows that work visually — The Simpsons above all — are undercounted on both axes, not just one, so the comparison is fairer than it looks but still not perfect.
- Craft and impact move together (r = 0.95 across all 21, 0.90 across the canonical 13), so "craft explains the index" is substantially "the two quality dimensions explain the index." They are 75% of the formula between them. We are not claiming craft beats impact; we are claiming both of them beat density, comfortably.
- We correlated the wrong density number, strictly speaking. The formula's density quarter is peak density (15%) plus weighted JPM (10%). What we have published per show, and what is used above, is plain average JPM. Peak density is not in our public data, so the component carrying the larger share of the density weight is untested here. If peak density behaves very differently from average density, this finding would need revisiting — though it would be a strange result, since the two are mechanically related.
- This is a formula we built. Craft is 40% of the Humor Index because we decided it should be, so "craft predicts the index" is partly true by construction and we are not pretending otherwise. The finding that is not baked in: density, which we deliberately gave a quarter of the score, ends up sorting nobody — because real shows vary 2.7x in density and 13% in quality, and the 13% is where all the separation lives. That is a fact about comedy, not about our weights.
- Thirteen data points. The exchange rate is an ordinary least-squares line through 13 shows, on an outcome that craft is an input to. Read it as a description of this board, not as a causal estimate of what happens if a writers' room gets sharper.
The jokes-per-minute number stays on every show page. It is genuinely interesting — it tells you what kind of comedy you are about to watch, machine-gun or slow-burn. It just will not tell you whether it is any good.
Scale note: 13 of our 21 shows have been rescored on the current canonical model, and the headline correlations, both natural experiments and the exchange rate above use only those 13. Eight shows — The Office, Friends, Curb Your Enthusiasm, It's Always Sunny, Community, Futurama, Taxi and The Fresh Prince of Bel-Air — are still on the pre-2026 scale. The Sunny era comparison is internal to one show scored one way, so the mixed scale does not affect it.