I have sensed that there is some demand for a better way of looking at these historical-adjusted numbers as the season is ongoing. So here's a new set of rankings that will serve as a hub for tracking how everything will grade out in final, full-season A18-49+ stats. Say hello to the Plus Power Rankings!
|ABC Dramas||A18-49||y2y||Proj Avg||A18-49+||Proj +||Proj + y2y|
|2||How to Get Away with Murder||2.96||2.96||169||177|
|3||Once Upon a Time||2.49||+13%||2.47||142||147||+27%|
|5||Agents of SHIELD||1.69||-40%||1.43||97||86||-33%|
Three of the stats here have been available in the SpotVault all season: A18-49 (average of all 18-49 ratings through the listed date this season), y2y (a comparison of this season's 18-49 ratings with last season's 18-49 ratings through the same number of episodes) and A18-49+, which compares one show's season-to-date A18-49 with the season-to-date "league average."
The common thread among those three things is that they're all based entirely on numbers that are already in the books. What's new on these rankings is the idea of "projected A18-49+" (Proj +). I've used this number before on in-season War of 18-49 Updates, but all of those were with shows whose seasons had ended. So since all of the show's ratings were in the books, the only thing really being projected was the final "league average." The projected league average applies the current year-to-year trend to the previous season's final league average.
(For an ongoing look at the current and projected league averages used for these rankings, see the Climate Center: Basics page under the columns "LAy2d" and "ProjLA.")
In the Plus Power Rankings, we're not just projecting where the league average will end up but also where the individual show averages will end up (Proj Avg). How are these projections generated? There are a couple different models used here, both of which are extremely simple and at times very flawed. Here are some details on the three different ways of coming up with the Proj Avg:
1. For shows whose seasons have ended, Proj Avg = A18-49. Duh. But the Proj + numbers are still useful here because they account for the rest of the projected league average decline. I have to manually enter which shows have ended, so please let me know if you spot an ended show whose Proj Avg is different from the A18-49.
2. For new shows still airing, the projected average assumes that a show will get exactly its most recent rating for all the rest of the episodes in its order. This is meant to create a more realistic average in the early weeks of new series, when the premiere and early episodes cause inflated ratings that the show will probably never revisit. There are a few issues with this; it can be a bit volatile based on whatever the last episode's rating happens to be. And it's still pretty slow adjusting to post-premiere drops; namely, there has to be a second episode before it can make any projection at all, and it won't really be accurate until the show has settled.
3. For returning shows still airing, it's the same formula as the projected league average: the show's current year-to-year trend is applied to last season's final average. Basically, what this does is it assumes that each show will finish the season the same way compared to the current average as it finished last season.
For relatively stable shows, #3 should do a pretty good job of projecting how the average and A18-49+ will change the rest of the way, as it helps account for expected drops due to things like DST. But for shows that had a significant inflation or deflation in one section of a season, it can really be a detriment. The above Once Upon a Time and Agents of SHIELD projections are very good examples of this. Because Once did well in the second half of last season, its average is expected to stay almost exactly flat; but that's extremely unlikely, because it was so inflated early this season. And because Agents of SHIELD had such inflated early episodes in season one, its average dropped huge in the second half of the season; it's highly unlikely the average will drop that much again this season.
There will be at least a couple more upgrades to these rankings going forward. The endgame is that I hope to create somewhat more sophisticated ways for projecting final averages. I have a few concrete ideas on how to do this, but they will take some more work. It's not impossible this could get done by the end of the season, but by that time it'd probably be pretty close to May, and by then they wouldn't have that much effect since there wouldn't be much left to project. So this will probably be a summer project that gets introduced in 2015-16.
The more immediate fix, which could come within a week or two, is an emergency treatment for some of the shows where the Proj Avg is egregiously wrong, like the aforementioned Once and SHIELD, as well as shows like The Blacklist that have had a major timeslot change that is skewing the year-to-year trends. It seems like these kinds of shows would benefit noticeably from simply using #2 above - applying its most recent rating to all future episodes. So what I will probably do is cherry-pick a few returning shows that could really use that approach. I'll also probably add some sort of color coding that indicates which of the three above approaches applies in each case.
Here's version 1.0 of the Plus Power Rankings by network: ABC | CBS | NBC | Fox | CW