Friday, November 13, 2015

The SpotVault's Timeslot Stats/Projections Table

Here is what's in the SpotVault's newly overhauled Timeslot Stats table, now also including projections of future averages:

# Date A18-49 True Sitch Timeslot Projections (24)
Total Tot% View Comp Lead PUT BC eLI y2d Proj Plus
1 9/22/2015 2.5 2.68 -0.18 -7% -0.13 +0.02 -0.08 26.7 10.4 1.0 2.50 2.09 136
2 9/29/2015 2.2 2.30 -0.10 -4% -0.07 +0.04 -0.07 27.4 9.4 1.1 2.35 1.98 128
3 10/6/2015 2.2 2.28 -0.08 -3% -0.04 +0.03 -0.07 27.8 9.7 1.1 2.30 2.05 134
4 10/13/2015 2.1 2.08 +0.02 +1% +0.10 -0.03 -0.06 29.6 10.4 1.1 2.25 2.04 135
5 10/20/2015 2.2 2.15 +0.05 +2% +0.04 +0.08 -0.06 28.8 8.8 1.1 2.24 2.12 141
6 10/27/2015 2.1 2.05 +0.05 +2% +0.13 -0.03 -0.05 30.0 10.5 1.2 2.22 2.08 140
7 11/3/2015 2.3 2.10 +0.20 +10% +0.09 +0.16 -0.06 29.5 7.5 1.2 2.23 2.16 145
8 11/10/2015 2.3 2.21 +0.09 +4% +0.11 +0.03 -0.05 29.7 9.7 1.2 2.24 2.15 144
AVERAGES:A18-49 TrueTotalTot%ViewCompLeadPUT BC eLIHeatSkewMale
2.24 2.23+0.01+0%+0.03+0.04-0.0628.7 9.5 1.1-4%17%42%

Three Sitches

This year's version of the table has retained only a few of the most fundamental timeslot stats: overall 18-49 viewing (PUT), broadcast viewing (BC) and lead-in (eLI). These are all the "massaged" versions of these numbers that actually go into the True formula. Among the most important adjustments: PUT includes a very old adjustment installed when Nielsen changed its definition of viewing levels several years ago. BC includes an adjustment to try to normalize competition levels across the three hours, adding a significant value to 10:00 to account for unknown numbers from Fox/CW local programming and cable. And eLI accounts somewhat for compatibility, adjusting the lead-in rating based on the skew discrepancy between the show and its lead-in.

Overall, the goal here was to simplify these tables to focus on the net impact (Sitch) of each of the formula's three basic areas: overall viewing (View), competition (Comp) and lead-in (Lead).

In last year's table, Sitch was also split into three parts: Viewing/Competition, Lead-in and Seasonal Multiplier. Lumping viewing and competition together was far from ideal, since they are each significant parts of the formula on their own. But it's tough to pull these parts of the formula apart since they work in concert to some degree. I've worked a lot on how to communicate viewership and competition separately and am still not that satisfied with what I have now, but it's the best I've been able to come up with. The seasonal multiplier from last year is now part of View, and what I call "deflation" has been split among View and Comp to try to make each of those numbers most sensible.


Last spring, I introduced projections of final season averages for shows in the Plus Power Rankings, and I have made some improvements to those projections this season. Another addition this season is an ongoing track of these projections on each show's Vault page. Included in this section is the average of all episodes aired thus far (y2d), a projection of the final average (Proj) and a projection of the final A18-49+ (Plus), which uses not just the projected show average but also the projected league average at the end of each individual week.

Ideally, this ongoing track will serve as another way (along with True scores) of measuring whether a show is overachieving or underachieving as the season progresses. In other words, a show is getting stronger if its projected final average is continually increasing, and vice versa. I'm hopeful it'll be useful in this way even now, at least to some extent. But at this stage, this ongoing track will largely serve to expose flaws in the projection methods. Either way, I hope you will enjoy following along.

A few housekeeping notes on these projections:
  • In introducing the new projection methods for this season, I said that I would make some manual adjustments when a projection looked particularly awful. I want to note that this applies only to the Plus Power Rankings. The ongoing tracks in the Vault will always adhere strictly to the methods described, which will help expose when the models are at their worst.
    • To add to that, it's possible that the models might get upgraded before the end of the season... but if that happens, I will leave the old numbers up. There will be some clear indicator of what the new numbers are/when the model changed.
  • An important part of these projections is how many episodes are ordered. You can find these in parentheses next to the word "Projections" at the top of the table, and I should be finding a way to work them into the Plus Power Rankings soon. Episode orders always default to the initial order for new shows and to the previous season's order for returning shows, but obviously many have to be manually adjusted and I'm trying to stay on top of them. Please let me know if you see any that are wrong!
    • The one exception to "old projections will never change" rule is if an episode order gets clarified. Hopefully this is not going to be a significant issue at this point in the season, since most episode orders have been clarified, and in the future I will try to figure out a way to handle this better.
  • The most recent predictions are in bold. It should be noted that these pages will not update while a show is off the air, so you'll have to check the Plus Power Rankings if you want the very latest Plus projections. (In other words, these pages don't account for changes in the projected league average until the show is back on the air.)
One other addition: like many other tables around here, these are now sortable! Click the categories at the top and you can see when a show was Truly strongest, when its timeslot was toughest, and when the end-of-season projection was highest.

These tables are not yet installed on Vault pages, but that process will begin this weekend. They should be up for all currently airing shows next week.

No comments:

Post a Comment

© 2009-2022. All Rights Reserved.