It worked because Arrow’s Wednesday, 8PM viewership had steady and predictable ratings shifts across the season. I ran 15 other network (ABC, CBS, FOX, & NBC)
dramas with 5 or more seasons through my ratings model. The model lost a lot of predictive value for non CW shows. The model worked best for shows that were most like Arrow through S5: genre or procedural shows with consistent timeslots and number of episodes. Thus the model works for Grimm, OUAT & Chicago PD but not Scandal, Blacklist or Elementary.
Will the model continue to be predictive for Arrow on a new and night and time? I won’t know until the end of the season.
The Comparison Point?
In seasons past I have generally compared Arrow’s rate of change (the decline or increase in viewer from one season to the next) against an average rate of change of the prior two big CW dramas (Supernatural & Vampire Diaries) at the same point in their run. This would compare Arrow’s S5 rating rate of change against an average of the S5 rating rate of change for
Supernatural & Vampire Diaries. However, there is more variance possibility in the rate of change this season because of the new timeslot.
I have mapped out a few possible rate of change comparison points.
The “best case scenario” for ratings decline would be if S6 Arrow ratings were on par with the S5 average for “the final third” episodes (5x16-5x23) for about a -14.3% decline. It is reflected in the dark green above (sorry if the label is hard to read). This is about where both Supergirl and the Flash came in with their premiere numbers this week. Alternately, Arrow S6 could mirror an average of the Arrow last 2 season’s rate of change decline of about 20%. Either of those options would pretty good/okay because it would reflect the audience declines from last season had mostly stabilized.
The “as expected range” would be the average rate of change decline for CW S6 drama (-23.5%) or even the same decline from Arrow S5 (-29.7%).
The “less great scenarios” would be if Arrow S6 mirrored the viewership declines of the CW “Night Switchers.” While some shows on the CW have never really had a stable timeslot (The 100, iZombie, now even Supernatural moves every few years), some shows who once had a stable timeslot were downgraded to a less favorable night/time (Vampire Diaries, Reign, The Originals) which resulted in a sharp decline in their viewership. It should be noted that these were far more “on the bubble” shows than Arrow at the time of time slot switch but it is the only predictor we have of how ratings decline based on timeslot switch work on the CW. The chart above shows in orange (-34.6%) the average if you combine both
rate of change decline for CW S6 drama (in yellow) and the CW Night Switch rate of change decline (in red). Lastly, the red (sorry if the label is hard to read)
reflects the rate of change decline for the CW Night Switch alone (-45.7%).
Since my model’s predictive power is driven by the season premier rating we can view what these different scenarios may look like, IF my model still work for a new timeslot.
For each possible comparison point, we have rate of change percentage and a predicted S6 rating average. If the model works we can predict the S6 ratings average from the 6x01 premiere rating. In this scenario, for example, a 6x01 rating of around 1.51M would predict a 1.34M S6 average which would be on par with the CW S6 rate of change.
We will know tomorrow where the 6x01 premiere rating falls in this range, but as always we will know until the end of the season if the model created for Arrow in the prior Wednesday, 8PM timeslot will have predictive value for a Thursday, 9PM timeslot.
Signed, A Lady Who Still Makes Ratings Predictive Statistical Models For A Show I Don’t Watch Anymore Cause This Is How Math Nerds Relax.