For a few weeks this site had The Office ahead of Seinfeld on the Humor Index. A lot of you argued with that ordering — Seinfeld has better-written jokes, you said. You were right. We had a thumb on the scale, and we just took it off.
The Updated Numbers
With every silent format adjustment removed, here’s where the three scored shows land:
- Seinfeld: 83.9 (was 77.9, +6.0)
- The Office: 80.2 (was 81.0, −0.8 — essentially unchanged)
- Friends: 78.7 (was 72.8, +5.9)
Seinfeld now leads by a comfortable margin. Friends is competitive with Office on raw score. The old ordering — Office first, Seinfeld second, Friends a distant third — was partly an artifact of a 15–25% penalty we were applying to multi-cam sitcoms without telling you.
What Was the Old Penalty?
Our earlier methodology multiplied the impact score of every multi-cam episode by a coefficient below 1.0. The theory: a live audience / laugh track inflates perceived impact, so we should correct for it.
That theory isn’t crazy. But three things made the correction indefensible:
1. Confounding. With only three scored shows, the "format effect" can’t be separated from show-level differences. A 15% penalty applied to Seinfeld could just as easily be a 15% penalty on Seinfeld-the-show, and you’d never know. 2. Opaque calibration. The coefficient was a point estimate with no published confidence interval. The underlying calibration study was a small sample that we can’t re-run or audit. 3. Silent correction. You saw Friends at 72.8 and had no way to know that 15% of that came from a multi-cam tax, not from the comedy being weaker.
What We’re Doing Instead
Raw scores. Every show’s score is now whatever the joke-level data produces, full stop. We’ve added:
- Format tags next to every show and episode, so you can always see what kind of comedy you’re looking at.
- Format filters on every ranking page — compare multi-cam shows to each other, or filter to just single-cam.
- Bootstrap 95% confidence intervals on every Humor Index, so the noise in each score is visible.
- Show-relative percentiles on every episode, so an 85 on Seinfeld and an 85 on Friends are commensurable within their own shows.
This is the right way to do it. If we later run a rigorous calibration study — with blind mode on, matched content, proper sample sizes, and uncertainty estimates — we might reintroduce a format correction. For now, no silent corrections.
So Is Seinfeld Actually Funnier?
By raw Humor Index, on our scoring: yes, by about 3.7 points. By craft score per joke: yes. By joke density: yes. By peak density: yes.
The Office still has things Seinfeld doesn’t. The sustained cringe. The mockumentary format letting Jim carry a weak plot with a camera look. Michael as a performance. These are real and the scores don’t fully capture them.
What changed today is that we stopped pretending our scores did capture them. The data says Seinfeld’s jokes land harder. The old ordering was the format adjustment talking. Now you see what the numbers actually say.
The Takeaway
Methodologies should be transparent. When we make a choice that changes rankings by 5–6 points, that choice should be visible, defensible, and documented. The old format coefficient failed all three tests.
If you want the old, format-adjusted score for any show or episode, it’s still in our data as `humor_index_v1`. The UI just doesn’t display it anymore, because we don’t trust it.
That's the whole point of The Humor Index — we give you the data, you make the argument.
Explore the full data: [The Office](/shows/the-office) • [Seinfeld](/shows/seinfeld) • [Compare side by side](/compare/the-office-vs-seinfeld)