RAR Explained
What Ratings Above Replacement means, what a good number is, and a guide to the five-part series on how the models work.
Who's responsible for the success and failure of TV shows? How important is a writer versus a director? Who are the very best showrunners of all time? Which upcoming shows will actually be good? Questions like these propelled this entire project.
While analytics is a very divisive topic in sports, it allows the qualitative to at least be somewhat quantified. When it's done very well, it allows the viewer to appreciate the contributions of a previously anonymous player, such as the slick-fielding shortstop or the shutdown cornerback.
To start, there will be a series of articles walking through how the models work. The language is all geared towards a non-technical audience. As long as you know what an x and y axis are, I promise you'll be fine. I recommend you read through the entire series. If you have any questions, feel free to fire away below any of the articles and I'm happy to answer.
The series
- Megan Ganz, the sneaky MVP of CommunityHow the episode model grades writers and directors.
- The better you are, the more credit you getHow a surprise gets split among the people who made it.
- Eric Kripke, elite showrunnerFrom episodes to seasons, and what a showrunner is worth.
- Grey's Anatomy and the gift or curse of accumulationHow billing decides a cast member's share.
- Tying it all togetherHow the pieces become a forecast for every episode.
RAR: Ratings Above Replacement
Every credit in the series is measured against an average show: zero means a person did what the model expected for someone with their job. That's the correct way to build the model, but it undersells the fact that even just getting the opportunity to contribute to a TV show is indicative of being more talented than the rest of the people on Earth. Doing an average job as a writer for 200 episodes of television creates value.
So everywhere else on the site, people's contributions and projections are shown in terms of Ratings above Replacement (RAR). Replacement level is defined as the quality of a show made up of entirely freely available talent. We let this replacement level float over time, and use its time-weighted rolling average to avoid any sharp upticks or downturns. Each season's gap is split among its showrunners and cast by their standard seats, and each episode's among its writers and directors.
Megan Ganz from Part 1 is a good example. Over 35 written episodes she added +4.02 over average. The replacement piece adds +0.92 more, for +4.93 RAR. Her rank among writers slips from #147 of 16,817 to #155, because people with far more episodes collect far more of the piece: David E. Kelley, with 543 episodes, climbs from #8 to #2. Baseball's WAR makes the same trade: an everyday regular can pass an elite part-timer.
What's a good number?
Careers as the Record book counts them, among people who have really worked in the role (3+ seasons; writers and directors 10+ episodes). “Typical” is the middle of the pack. About 30% of them are below replacement over their careers.
| Career RAR | Typical | Top 25% | Top 10% | Top 1% |
|---|---|---|---|---|
| Showrunnersbest: David E. Kelley +19.4 | +0.56 | +1.53 | +3.39 | +10.06 |
| Actorsbest: Richard Dean Anderson +16.4 | +0.06 | +0.29 | +0.85 | +3.58 |
| Writersbest: J. Michael Straczynski +41.1 | +0.62 | +1.88 | +4.10 | +13.41 |
| Directorsbest: Martin Wood +16.3 | +0.46 | +1.34 | +2.80 | +6.85 |
| RAR per season or episode | Typical | Top 25% | Top 10% | Top 1% |
|---|---|---|---|---|
| Showrunnersper season | +0.12 | +0.27 | +0.46 | +0.88 |
| Actorsper season | +0.07 | +0.17 | +0.28 | +0.49 |
| Writersper episode | +0.03 | +0.08 | +0.13 | +0.28 |
| Directorsper episode | +0.01 | +0.03 | +0.06 | +0.13 |
Compare people within a role: writers and directors are counted episode by episode, so their totals grow faster than a showrunner's or an actor's. Totals reward long careers; the per-season and per-episode numbers show the rate.
Ratings Adjustments
The ratings presented on this site are a combination of season and series ratings. On IMDb, the people who rate individual episodes tend to be a show's biggest fans, so a show's episodes usually average higher than the rating IMDb voters give the show as a whole. Left alone, that would let a small, devoted fan base make a show, and everyone on it, look better than it is.
So for every show, we take the gap between its series rating and the average of its episode ratings (weighted by votes) and shift every episode and season by that amount. The order of episodes and seasons within a show never changes; only where the show sits against other shows does.
Joey is a good example. Its episodes average 7.19 on IMDb, but 46,149 people rated the series as a whole 6.1. The gap is 6.10 − 7.19 = −1.09, applied to both seasons:
| Joey | Episode average on IMDb | Adjustment | Rating on ScreenBall |
|---|---|---|---|
| Season 1 (2004) | 7.29 | −1.09 | 6.20 |
| Season 2 (2005) | 7.06 | −1.09 | 5.97 |
Season 2 is still a bit worse than season 1, just as the episode ratings say. But both now sit where the wider audience put the show, not where its most loyal fans did. Every rating, forecast and RAR number on the site uses these adjusted ratings, and adjusted ratings are shown between 0 and 10.
Two related details. Episode grades (how an episode did against the rest of its own season) don't change, since the adjustment moves every episode of a show by the same amount. And episodes with very few votes count only partially toward RAR: an episode's weight is its IMDb votes divided by a typical low count for its era (the 25th percentile of episodes from the previous five years), capped at 1. That keeps a single 50-vote episode from making a career. About one credit in five is below full weight, almost all of them from tiny-audience shows.