Megan Ganz, the sneaky MVP of Community
How the episode model grades writers and directors.
As mentioned in the introductory article, all explanations are geared towards a non-technical audience. What you primarily need to know about the models is that writers and directors get credited for every episode in which they participate, while showrunners and cast members accumulate credit at the seasonal level. In all of these cases, credit is defined by the ratings of the episode (1–10). Audience numbers, etc. are completely out of the scope of this project. Let's start with screenwriter Megan Ganz.
Megan Ganz's first writing credit on a TV show was the meta bottle episode of Community in season 2 entitled "Cooperative Calligraphy." For writers and directors, the metric we care about is the rating of the episode, minus the ratings of other episodes that season that did not involve them.
"Cooperative Calligraphy" was one of the better episodes of a very good season: 9.238128 against a season average of 8.62. We predict how each episode will do against the rest of its season from the body of work of its writer and director. This article will focus on the iterative model process that creates the writer's and director's body of work.
It starts with the model prediction. As mentioned, this episode was Megan Ganz's first, so she had no body of work. Joe Russo had directed 26 episodes by then, across Arrested Development and Community. His body of work was slightly negative, as his episodes that early in his career underperformed their seasonal counterparts by a small amount. Ganz and Russo each get credited with their impact on the prediction, which was −0.06. That meant the model thought it would very slightly underperform season 2's average episode rating. However, the episode garnered a strong 9.238128, which was 0.86 points higher than the prediction. We'll discuss how the split of that +0.86 is decided in the next article. For now, all you need to know is that Ganz got +0.63 of that +0.86.
Fast forward to her fourth Community episode, "Basic Lupine Urology." By that point Megan was on an absolute heater, with three standout episodes under her belt. She now had a body of work. How was it calculated?
How a record is built
Notice that "Lupine Urology" came with a prediction attached and "Cooperative Calligraphy" didn't. That's the other half of the model: every episode a writer or director makes feeds the prediction for the next one. Three rules decide how.
1. The record only updates between seasons. Ganz's second season 2 episode, "Intermediate Documentary Filmmaking," still predicted 0, even though she'd just knocked it out of the park with "Cooperative Calligraphy." Within a season, episodes are written and shot in big batches, often before the first one even airs, so the model waits for the season to finish before it updates anyone.
2. If your body of work has come in a small sample, we push it towards average depending on how many episodes you've written. For the technically inclined, this is done via Bayesian shrinkage. Heading into season 3, her two episodes averaged +0.49, but her body of work the model sees was only +0.12. Two episodes on one show isn't much to go on, so the model only partially buys into her success. By season 4 she had four episodes averaging +0.53, and the model gave her a more confident +0.24. It takes about 10 episodes across 2 shows before a writer's record is taken at full value (8 across 2 for a director), and that bar is the same in 1975 as it is today.
3. What you've done on this show matters much more than what you've done elsewhere. The model sees both your body of work on the current show, as well as your body of work on other shows. Writing great Community episodes tells you a lot about her next Community episode, and a lot less about her first Modern Family episode.
Here's her whole writing career through that lens. Each row is a show-season, and the bar is what the model expected her episodes to add before they aired.
The pattern repeats on every show. She starts each one near zero, and her body of work there builds (or doesn't) from her results. By 2023 she'd built a strong record on It's Always Sunny (+0.19, predicting +0.15 an episode), while her Mythic Quest episodes went the other way (−0.18). Across 35 writing credits, her career total stands at +3.92, +2.28 of it from Community.
Ratings are IMDb user ratings from IMDb's public data files. Credits are from the model that built this page.
Next in the series: The better you are, the more credit you get (Part 2).