Last time, we put together the formula for what I would refer to as "TV competition." But in doing the viewing and TV competition phases it became clear that I also needed what I would refer to as "non-TV competition"; in other words, on weekends and holidays there's a different kind of "competition" with TV viewing that better-adjusted people than I might call "real life." Y'know, "I'm not watching TV because I'm going out or traveling or something."
That kind of competition is somewhat approximately explained by a simple comparison of ratings to overall viewing, but the problem is that we need a number that goes even beyond that to account for the fact that there's very little "TV competition." Worst case scenario, I could do what I did last year: simply create two separate baselines for expected competition, one for weekends/holidays and one for weeknights. But I wanted a number that could actually pick out a viewing-depressed period so I wouldn't have to.
So I began a rather grueling brainstorm for numbers that appeared to stick out in holiday situations. Here were the three most serious candidates:
1) bc vs. Predicted. Very early on in the summer, I was working on a calculation of expected competition that essentially tried to predict what the overall broadcast ratings would be based on the overall viewing levels. Basically I was trying to come up with something that would connect the two competition baselines I used last year (23% PUT on weekends, 30% on weekdays). I ditched it pretty quickly, realizing that an expected competition should not be trying to "predict" the competition; if you "predict" there will be huge competition on Super Bowl Sunday, then you're not really handing out a competition bonus!
However, I thought the very rudimentary version of that equation might be good here. Essentially the number was pretty low in competitive situations and pretty high in non-competitive situations (like holidays!), but it's rarely a huge number. I came up with it by noticing that the bc's percentage of the TPUT was pretty close to the size of the actual TPUT itself. I eventually settled on the bc's percentage being 90% of the TPUT. Multiply both sides by TPUT and you have bc = 90% * TPUT^2. The bc vs. Predicted number equals that (90% * TPUT^2) minus the actual bc.
2) Cable. Another thing I noticed in my far-reaching brainstorm was that the "cable" number (overall viewing minus all the broadcast viewing) happened to be about the same on every night of the week. From Monday to Sunday it went 25.1 -> 24.6 -> 23.4 -> 23.5 -> 23.3 -> 23.4 -> 24.3. So the idea I ran with here was that that the cbl was a very good measure of competitiveness. It was about the same even in viewing depressed situations but it went way down in highly competitive spots and way up in highly uncompetitive spots. I thought that adding on essentially something along the lines of 24.0 minus cable would be a good way to offset the weakness of the other number.
In a sense, it was pretty decent at that, but it also didn't make much of an impact in terms of helping holiday numbers stand out, since (as I said before) the number is not really that big in viewing-depressed situations. And there were still some situations where the number was deeply negative, preventing me from doubling or tripling the number to add to the effect on holidays. Still, a combination of numbers one and two were for a long time the best thing I could possibly find. Until last weekend, when I got another bright idea...
3) PUT vs. Predicted. I decided to try coming up with a "prediction" of the PUT based on the bc number; essentially a reverse version of the Tendency PUT I use in the viewing section. Start with a "normal" Tendency PUT, add in 40% of the actual broadcast ratings and see how that compares to the actual TPUT. I figured that would do a good job of highlighting viewing depressions (where the PUT would appear lower than it was "supposed" to be) but (like the cable number) it would offset the "competitiveness" problem somewhat, since those highly competitive/uncompetitive situations typically have a fairly normal Tendency PUT. (That's the whole point of using Tendency PUT rather than normal PUT.)
This PUT vs. Predicted number is (Normal Tendency PUT + 0.4 * bc) minus the TPUT
I considered using just one constant number for "Normal Tendency PUT" but found that seemed to hurt shows that aired in two different timeslots. So I'm using the season average tendency PUT in each individual half-hour: 28.3 at 8:00, 29.5 at 8:30, 30.6 at 9:00, 31.3 at 9:30, 31.5 at 10:00, 29.8 at 10:30.
So far, adding numbers 1 and 3 is the best thing I've got. The simplest way I can think to explain it is this: these numbers ask "how much lower is the bc than it *should* be?" and "how much lower is the PUT than it *should* be?" and the idea is that weekends/holidays are the only time where both of those answers are a resounding, "Much lower." I'm a bit concerned about how the number looks in the aforementioned competitive/non-competitive situations, because #3 is not as good at offseting those situations as #1 is at creating the problems. But I haven't found some mix of the two numbers that really seems any better than just adding the two in their entireties together. While those numbers I pointed out to represent competitive situations when I first started this end up being still a little lower than I'd like, they're not nearly as bad as they were. And hey, maybe reality shows like Undercover Boss and Celebrity Apprentice are supposed to be a little "truly" weaker against award shows and huge football games. The uncompetitive situations mostly don't seem too bad because I think they offset what might be a bit of inflation in the Tendency PUT anyway.
The Viewing/Competition Semantics
An interesting question about this holiday number is whether it's really "competition." After all, it's essentially an attempt to find a "tendency not to view" and you would think I would tack that onto the Tendency PUT. I can't guarantee that I won't eventually change it to that, but for now I kind of like the way this looks. Tendency PUT takes a relatively smooth trajectory across the season, and the competition ends up appearing fairly consistent both in high-viewed spots and in low-viewed spots. So I like that consistency. Anyway, I can't really imagine anyone else caring about this, but it might become something of an issue if I start doing True2 "splits" in some kind of practical form (breaking down how each of the three steps affects the number). I guess I'll cross that bridge when I get to it.
Expected Competition
One real benefit of the Competition + Holiday number as currently construed is that it seems to barely have any relationship with actual viewing levels. Which was kind of the point! If anything, the correlation is somewhat inverse since the weekends have such big holiday numbers. I started off, like last year, looking for an expected competition number that was a certain percentage of the TPUT or of the Tendency PUT. But I've come to realize that may not really be necessary, since there's no real relationship there. So I'm instead just using the same expected competition for every day and timeslot, and it's equal to the average Competition + Holiday number on the big four for the full season: 9.82. The only exception is on the CW, where I multiply this times 4.00/3.25 (resulting in 12.08) to represent the fact that the CW faces four networks where the other nets only face 3.25 networks.
Sports as Competition
In the previous incarnation of this number, I subtracted half of a sports event's rating from the competition simply because it seemed like sports events did not make as much of an impact as their ratings suggested they should.
This time, I noticed the Holiday number was always very negative on evenings with sports events, because they have big numbers and cause a PUT inflation. So I'm gonna do the much easier thing and count sports as "full" competition this time.
A couple days ago as I was in the middle of writing these posts I checked on the True2 score for the Super Bowl and noticed it was a negative 5.88! Panic time began to set in. But it became quickly evident that all I needed to do was add half of the Super Bowl's rating to the competition score and the Holiday number was basically completely normal. So that's what I do in the True2 scores for sports events; they get half of their own rating added back in.
I think the way to look at this is not that sports count as full competition; effectively sports count as half-competition, but then I add back in half of the sports rating to normalize the Holiday number. That is essentially what I'm doing both for sports events and the shows they face; it's just that sports events don't "face themselves" so there's no sports competition to be halved. I think this problem has to do with the fact that the "bc" calculations in the Holiday Number count sports fully, so if I don't do that in the competition calculation it creates a problem. I could try to redo the bc vs. Predicted and PUT vs. Predicted equations with half-sports but at this point I think it might drive me clinically insane and I don't think it'd make much of a difference.
The Full Competition + Holiday Adjustment
Here's what we had last time:
Competition Impact = 3.75%
Competition = Sum of ratings for all OTHER broadcast shows + 10:00 Hour Adjustment + Impact * (A18-49 - 5.5% * TPUT)
Competition Adjustment = Impact * (Competition + Holiday Number - Expected Competition)
Now I can reveal the two things that weren't in the picture last time:
bc vs. Predicted (bcvP) = (90% * TPUT^2) - bc
PUT vs. Predicted (PvP) = (Normal Tendency PUT + 0.4 * bc) - TPUT
Holiday Number: bcvP + PvP (if sports, + 0.5 * A18-49)
Expected Competition = 9.82 (times 4.00/3.25 if CW)
Competition Adjustment = Impact * (Competition + Holiday Number - Expected Competition)
Tomorrow is the end of True2
True2 Week: Viewing | Lead-ins | Competition pt1 | Competition pt2 | Wrap-Up
No comments:
Post a Comment